tag:theconversation.com,2011:/au/topics/face-recognition-18822/articlesFace recognition – The Conversation2024-03-04T01:25:27Ztag:theconversation.com,2011:article/2246432024-03-04T01:25:27Z2024-03-04T01:25:27ZYour face for sale: anyone can legally gather and market your facial data without explicit consent<figure><img src="https://images.theconversation.com/files/579102/original/file-20240301-28-tzp738.jpg?ixlib=rb-1.1.0&rect=956%2C85%2C6119%2C4218&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">
</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/futuristic-technological-scanning-face-beautiful-woman-1554013514">Kitreel/Shutterstock</a></span></figcaption></figure><p>The morning started with a message from a friend: “I used your photos to train my local version of Midjourney. I hope you don’t mind”, followed up with generated pictures of me wearing a flirty steampunk costume.</p>
<p>I did in fact mind. I felt violated. Wouldn’t you? I bet Taylor Swift did when <a href="https://theconversation.com/taylor-swift-deepfakes-new-technologies-have-long-been-weaponised-against-women-the-solution-involves-us-all-222268">deepfakes of her hit the internet</a>. But is the legal status of my face different from the face of a celebrity?</p>
<p>Your facial information is a unique form of personal sensitive information. It can identify you. Intense profiling and mass government surveillance <a href="https://www.forbes.com/sites/kalevleetaru/2019/05/06/as-orwells-1984-turns-70-it-predicted-much-of-todays-surveillance-society/?sh=38a97b4e11de">receives much attention</a>. But businesses and individuals are also using tools that <a href="https://www.sbs.com.au/news/article/creepy-and-invasive-kmart-bunnings-and-the-good-guys-accused-of-using-facial-recognition-technology/h08q8evb1">collect</a>, <a href="https://www.afr.com/technology/how-clearview-ai-unleashed-a-global-dystopia-20230929-p5e8lc">store</a> and modify facial information, and we’re facing an unexpected wave of <a href="https://deepai.org/machine-learning-model/text2img">photos</a> and <a href="https://theconversation.com/what-is-sora-a-new-generative-ai-tool-could-transform-video-production-and-amplify-disinformation-risks-223850">videos</a> generated with artificial intelligence (AI) tools.</p>
<p>The development of legal regulation for these uses is lagging. At what levels and in what ways should our facial information be protected? </p>
<h2>Is implied consent enough?</h2>
<p>The Australian <a href="https://www.legislation.gov.au/C2004A03712/latest/text">Privacy Act</a> considers biometric information (which would include your face) to be a part of our personal sensitive information. However, the act doesn’t <em>define</em> biometric information. </p>
<p>Despite its drawbacks, the act is currently the main legislation in Australia aimed at facial information protection. It states biometric information cannot be collected without a person’s consent. </p>
<p>But the law doesn’t specify whether it should be <a href="https://www.ipc.nsw.gov.au/fact-sheet-consent">express or implied consent</a>. Express consent is given explicitly, either orally or in writing. Implied consent means consent may reasonably be inferred from the individual’s actions in a given context. For example, if you walk into a store that has a sign “facial recognition camera on the premises”, your consent is implied. </p>
<figure class="align-right zoomable">
<a href="https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="A poster at a supermarket that says camera technology trial in progress, partially obscured by a couple of bins." src="https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=237&fit=clip" srcset="https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=1067&fit=crop&dpr=1 600w, https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=1067&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=1067&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=1340&fit=crop&dpr=1 754w, https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=1340&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/578587/original/file-20240228-28-ns24xh.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=1340&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">An inconspicuous sign that flags camera technology trial is in progress counts as implied consent.</span>
<span class="attribution"><span class="source">Margarita Vladimirova</span></span>
</figcaption>
</figure>
<p>But using implied consent opens our facial data up to potential exploitation. <a href="https://www.choice.com.au/consumers-and-data/data-collection-and-use/how-your-data-is-used/articles/kmart-bunnings-and-the-good-guys-using-facial-recognition-technology-in-store">Bunnings, Kmart</a> and <a href="https://www.theguardian.com/business/2023/feb/19/woolworths-expands-self-checkout-ai-that-critics-say-treats-every-customer-as-a-suspect">Woolworths</a> have all used easy-to-miss signage that facial recognition or camera technology is used in their stores.</p>
<h2>Valuable and unprotected</h2>
<p>Our facial information has become so valuable, <a href="https://www.theguardian.com/australia-news/2023/oct/24/australian-federal-police-afp-pimeyes-facial-recognition-facecheck-id-search-engine-platform">data companies such as Clearview AI and PimEye</a> are mercilessly hunting it down on the internet <a href="https://onezero.medium.com/i-got-my-file-from-clearview-ai-and-it-freaked-me-out-33ca28b5d6d4">without our consent</a>.</p>
<p>These companies put together databases for sale, used not only by the police in various countries, <a href="https://www.theguardian.com/australia-news/2023/oct/24/australian-federal-police-afp-pimeyes-facial-recognition-facecheck-id-search-engine-platform">including Australia</a>, but also by <a href="https://www.clearview.ai/developer-api">private companies</a>. </p>
<p>Even if you deleted all your facial data from the internet, you could easily be captured in public and appear in some database anyway. Being in someone’s TikTok video <a href="https://www.abc.net.au/news/2022-07-14/tiktok-video-maree-melbourne-flowers/101228418">without your consent</a> is a prime example – in Australia this is legal.</p>
<p><div data-react-class="Tweet" data-react-props="{"tweetId":"1547794543726055425"}"></div></p>
<p>Furthermore, we’re also now contending with generative AI programs such as Midjourney, DALL-E 3, Stable Diffusion and others. Not only the collection, but the modification of our facial information can be easily performed by anyone. </p>
<p>Our faces are unique to us, they’re part of what we perceive as ourselves. But they don’t have special legal status or special legal protection.</p>
<p>The only action you can take to protect your facial information from aggressive collection by a store or private entity <a href="https://www.oaic.gov.au/privacy/privacy-complaints/lodge-a-privacy-complaint-with-us">is to complain</a> to the office of the Australian Information Commissioner, which may or may not result in an investigation.</p>
<p>The same applies to deepfakes. The Australian Competition and Consumer Commission will consider only activity that applies to trade and commerce, for example if a <a href="https://www.theguardian.com/technology/2022/mar/18/accc-takes-meta-to-court-over-facebook-scam-ads-depicting-australian-identities">deepfake is used for false advertising</a>. </p>
<p>And the Privacy Act doesn’t protect us from other people’s actions. I didn’t consent to have someone train an AI with my facial information and produce made-up images. But there is no oversight on such use of generative AI tools, either. </p>
<p>There are currently no laws that <em>prevent</em> other people from collecting or modifying your facial information.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/so-youve-been-scammed-by-a-deepfake-what-can-you-do-223299">So, you've been scammed by a deepfake. What can you do?</a>
</strong>
</em>
</p>
<hr>
<h2>Catching up the law</h2>
<p>We need a range of regulations on the collection and modification of facial information. We also need a stricter status of facial information itself. Thankfully, some developments in this area are looking promising.</p>
<p>Experts at the University of Technology Sydney have proposed a comprehensive legal framework for <a href="https://www.uts.edu.au/human-technology-institute/projects/facial-recognition-technology-towards-model-law">regulating the use of facial recognition technology</a> under Australian law.</p>
<p>It contains proposals for regulating the first stage of non-consensual activity: the collection of personal information. That may help in the development of new laws.</p>
<p>Regarding photo modification using AI, we’ll have to wait for announcements from the newly established government <a href="https://www.minister.industry.gov.au/ministers/husic/media-releases/new-artificial-intelligence-expert-group">AI expert group</a> working to develop “safe and responsible AI practices”.</p>
<p>There are no specific discussions about a higher level of protection for our facial information in general. However, the government’s recent <a href="https://www.ag.gov.au/rights-and-protections/publications/government-response-privacy-act-review-report">response to the Attorney-General’s Privacy Act review</a> has some promising provisions. </p>
<p>The government has agreed further consideration should be given to enhanced risk assessment requirements in the context of facial recognition technology and other uses of biometric information. This work should be coordinated with the government’s ongoing work on Digital ID and the National Strategy for Identity Resilience. </p>
<p>As for consent, the government has agreed in principle that the definition of consent required for biometric information collection should be amended to specify it must be voluntary, informed, current, specific and unambiguous. </p>
<p>As facial information is increasingly exploited, we’re all waiting to see whether these discussions do become law – hopefully sooner rather than later.</p>
<hr>
<p><em>Correction: we have amended a sentence to clarify Woolworths use camera technology but not necessarily facial recognition technology.</em></p><img src="https://counter.theconversation.com/content/224643/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Margarita Vladimirova does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Our facial information is sensitive – yet companies and individuals can collect, sell and manipulate it without our consent. Australian law must change to protect us all.Margarita Vladimirova, PhD in Privacy Law and Facial Recognition Technology, Deakin UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/2172262024-01-19T13:42:15Z2024-01-19T13:42:15ZFace recognition technology follows a long analog history of surveillance and control based on identifying physical features<figure><img src="https://images.theconversation.com/files/569962/original/file-20240117-29-ri412u.jpg?ixlib=rb-1.1.0&rect=0%2C0%2C5272%2C3598&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Today's technology advances what passport control has been doing for more than a century.</span> <span class="attribution"><a class="source" href="https://www.gettyimages.com/detail/news-photo/controll-of-passports-at-the-frontiers-between-beuthen-and-news-photo/548866047">ullstein bild via Getty Images</a></span></figcaption></figure><p>American Amara Majeed was <a href="https://www.bbc.com/news/world-asia-48061811">accused of terrorism</a> by the Sri Lankan police in 2019. Robert Williams was <a href="https://www.nytimes.com/2020/06/24/technology/facial-recognition-arrest.html">arrested outside his house</a> in Detroit and detained in jail for 18 hours for allegedly stealing watches in 2020. Randal Reid <a href="https://www.nytimes.com/2023/03/31/technology/facial-recognition-false-arrests.html">spent six days in jail</a> in 2022 for supposedly using stolen credit cards in a state he’d never even visited.</p>
<p>In all three cases, the authorities had the wrong people. In all three, it was face recognition technology that told them they were right. Law enforcement officers in many U.S. states are <a href="https://www.wired.com/story/hidden-role-facial-recognition-tech-arrests/">not required to reveal</a> that they used face recognition technology to identify suspects.</p>
<p>Face recognition technology is the latest and most sophisticated version of <a href="https://www.dhs.gov/biometrics">biometric surveillance</a>: using unique physical characteristics to identify individual people. It stands in a <a href="https://www.thalesgroup.com/en/markets/digital-identity-and-security/government/inspired/history-of-biometric-authentication">long line of technologies</a> – from the fingerprint to the passport photo to iris scans – designed to monitor people and determine who has the right to move freely within and across borders and boundaries.</p>
<p>In my book, “<a href="https://www.press.jhu.edu/books/title/12700/do-i-know-you">Do I Know You? From Face Blindness to Super Recognition</a>,” I explore how the story of face surveillance lies not just in the history of computing but in the history of medicine, of race, of psychology and neuroscience, and in the health humanities and politics.</p>
<p>Viewed as a part of the long history of people-tracking, face recognition techology’s incursions into privacy and limitations on free movement are carrying out exactly what biometric surveillance was always meant to do.</p>
<p>The system works by converting captured faces – either static from photographs or moving from video – into a series of unique data points, which it then compares against the data points drawn from images of faces already in the system. As face recognition technology improves in accuracy and speed, its effectiveness as a means of surveillance becomes ever more pronounced.</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="faces in a crowd highlighted and annotated with dates and times" src="https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/569964/original/file-20240117-15-h4ovvh.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Paired with AI, face recognition technology scans the crowd at a conference.</span>
<span class="attribution"><a class="source" href="https://www.gettyimages.com/detail/news-photo/live-demonstration-uses-artificial-intelligence-and-facial-news-photo/1080200068">David McNew/AFP via Getty Images</a></span>
</figcaption>
</figure>
<h2>Accuracy improves, but biases persist</h2>
<p>Surveillance is predicated on the idea that <a href="https://theconversation.com/surveillance-is-pervasive-yes-you-are-being-watched-even-if-no-one-is-looking-for-you-187139">people need to be tracked</a> and their movements limited and controlled in a trade-off between privacy and security. The assumption that less privacy leads to more security is built in.</p>
<p>That may be the case for some, but not for the people disproportionately targeted by face recognition technology. <a href="https://www.routledge.com/Histories-of-Surveillance-from-Antiquity-to-the-Digital-Era-The-Eyes-and/Marklund-Skouvig/p/book/9781032021539">Surveillance has always been designed</a> to identify the people whom those in power wish to most closely track.</p>
<p>On a global scale, <a href="https://doi.org/10.1080/21670811.2018.1493938">there are</a> <a href="https://longreads.tni.org/stateofpower/settled-habits-new-tricks-casteist-policing-meets-big-tech-in-india">caste cameras in India</a>, <a href="https://www.theguardian.com/world/2021/sep/30/uyghur-tribunal-testimony-surveillance-china">face surveillance of Uyghurs in China</a> and even <a href="https://mynbc15.com/news/spotlight-on-america/facial-recognition-technology-in-school-hallways-states-face-a-divisive-debate">attendance surveillance</a> <a href="https://dx.doi.org/10.7302/21934">in U.S. schools</a>, often with low-income and majority-Black populations. <a href="https://www.aclu.org/news/privacy-technology/how-is-face-recognition-surveillance-technology-racist">Some people are tracked more closely</a> than others.</p>
<p>In addition, the cases of Amara Majeed, Robert Williams and Randal Reid <a href="https://www.aclu.org/news/privacy-technology/how-is-face-recognition-surveillance-technology-racist">aren’t anomalies</a>. As of 2019, face recognition technology <a href="https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf">misidentified Black and Asian people</a> at up to <a href="https://www.nist.gov/news-events/news/2019/12/nist-study-evaluates-effects-race-age-sex-face-recognition-software">100 times the rate of white people</a>, including, in 2018, a disproportionate number of the <a href="https://www.aclu.org/news/privacy-technology/amazons-face-recognition-falsely-matched-28">28 members of the U.S. Congress</a> who were falsely matched with mug shots on file using Amazon’s Rekognition tool.</p>
<p>When the database against which captured images were compared had only a limited number of mostly white faces upon which to draw, face recognition technology would offer matches based on the closest alignment available, leading to a pattern of highly racialized – and racist – false positives.</p>
<p>With the expansion of images in the database and increased sophistication of the software, <a href="https://www.csis.org/blogs/strategic-technologies-blog/how-accurate-are-facial-recognition-systems-and-why-does-it">the number of false positives</a> – incorrect matches between specific individuals and images of wanted people on file – has <a href="https://bipartisanpolicy.org/blog/frt-accuracy-performance/">declined dramatically</a>. Improvements in pixelation and mapping static images into moving ones, along with increased social media tagging and <a href="https://www.penguinrandomhouse.com/books/691288/your-face-belongs-to-us-by-kashmir-hill/">ever more sophisticated scraping tools</a> like those developed by Clearview AI, have helped decrease the error rates.</p>
<p><a href="https://www.washingtonpost.com/technology/2019/12/19/federal-study-confirms-racial-bias-many-facial-recognition-systems-casts-doubt-their-expanding-use/">The biases</a>, however, remain deeply embedded into the systems and their purpose, explicitly or implicitly targeting already targeted communities. The technology is not neutral, nor is the surveillance it is used to carry out.</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="Pen and ink illustration of suited hands using calipers to measure a man's forehead to back of his head" src="https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=458&fit=crop&dpr=1 600w, https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=458&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=458&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=576&fit=crop&dpr=1 754w, https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=576&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/569966/original/file-20240117-21-awurl6.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=576&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Physiognomy went beyond recognition of an individual and tried to connect physical features with other characteristics.</span>
<span class="attribution"><a class="source" href="https://www.gettyimages.com/detail/illustration/head-royalty-free-illustration/1399373778">clu/DigitalVision Vectors via Getty Images</a></span>
</figcaption>
</figure>
<h2>Latest technique in a long history</h2>
<p>Face recognition software is only the most recent manifestation of global systems of tracking and sorting. Precursors are rooted in the now-debunked belief that bodily features offer a unique index to character and identity. This pseudoscience was formalized in the late 18th century under the rubric of the <a href="https://www.hup.harvard.edu/books/9780674036048">ancient practice of physiognomy</a>.</p>
<p>Early systemic applications included anthropometry (body measurement), fingerprinting and iris or retinal scans. They all offered unique identifiers. None of these could be done without the participation – willing or otherwise – of the person being tracked.</p>
<p>The framework of bodily identification was adopted in the 19th century for use in criminal justice detection, prosecution and record-keeping to allow governmental control of its populace. The intimate relationship between face recognition and border patrol was galvanized by the <a href="http://www.atlasobscura.com/articles/passport-photos-history-development-regulation-mugshots">introduction of photos into passports</a> in some countries including Great Britain and the United States in 1914, <a href="https://doi.org/10.1017/9781108664271">a practice that became widespread by 1920</a>.</p>
<p>Face recognition technology provided a way to go stealth on human biometric surveillance. Much early research into face recognition software was <a href="https://www.wired.com/story/secret-history-facial-recognition/">funded by the CIA</a> for the purposes of border surveillance.</p>
<p>It tried to develop a standardized framework for face segmentation: mapping the distance between a person’s facial features, including eyes, nose, mouth and hairline. Inputting that data into computers let a user search stored photographs for a match. These early scans and maps were limited, and the attempts to match them were not successful.</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="Woman looks at screen with her image on a vending machine" src="https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/569967/original/file-20240117-23-u3alzk.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">A customer pays via facial recognition at a smart store in China.</span>
<span class="attribution"><a class="source" href="https://www.gettyimages.com/detail/news-photo/nov-6-2018-a-visitor-tries-facial-recognition-payment-in-a-news-photo/1058496364">Huang Zongzhi/Xinhua News Agency via Getty Images</a></span>
</figcaption>
</figure>
<p>More recently, private companies have <a href="https://fortune.com/longform/facial-recognition/">adopted data harvesting techniques</a>, including face recognition, as part of a long practice of <a href="https://theconversation.com/data-brokers-know-everything-about-you-what-ftc-case-against-ad-tech-giant-kochava-reveals-218232">leveraging personal data for profit</a>.</p>
<p>Face recognition technology works not only to unlock your phone or help you board your plane more quickly, but also in promotional store kiosks and, essentially, in any photo taken and shared by anyone, with anyone, anywhere around the world. These photos are stored in a database, creating ever more comprehensive systems of surveillance and tracking.</p>
<p>And while that means that today it is unlikely that Amara Majeed, Robert Williams, Randal Reid and Black members of Congress would be ensnared by a false positive, face recognition technology has invaded everyone’s privacy. It – and the governmental and private systems that design, run, use and capitalize upon it – is watching, and paying particular attention to those whom society and its structural biases deem to be the greatest risk.</p><img src="https://counter.theconversation.com/content/217226/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Sharrona Pearl receives funding from Interfaith America.</span></em></p>Face recognition technology follows earlier biometric surveillance techniques, including fingerprints, passport photos and iris scans. It’s the first that can be done without the subject’s knowledge.Sharrona Pearl, Associate Professor of Bioethics and History, Drexel UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/2164652023-11-07T18:07:44Z2023-11-07T18:07:44ZAI-generated faces look just like real ones – but evidence shows your brain can tell the difference<figure><img src="https://images.theconversation.com/files/558021/original/file-20231107-29-2bgov3.jpeg?ixlib=rb-1.1.0&rect=0%2C0%2C2202%2C1103&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Can you tell which faces are real and which are synthetic? Answers are at the bottom of the article.</span> <span class="attribution"><span class="source">Robin Kramer</span>, <span class="license">Author provided</span></span></figcaption></figure><p>For a while, limitations in technology meant that animators and researchers were only capable of creating human-like faces which seemed a little “off”.</p>
<p>Films like 2004’s <a href="https://www.imdb.com/title/tt0338348/">The Polar Express</a> made some viewers uneasy because the characters’ faces looked almost human but not quite, and so they fell into what we call the “<a href="https://theconversation.com/uncanny-valley-why-we-find-human-like-robots-and-dolls-so-creepy-50268">uncanny valley</a>”. This is when artificial faces (or robots more generally) look increasingly human and get very close to resembling us while still showing signs of being artificial, they elicit discomfort or even revulsion.</p>
<p>Recent advances in artificial intelligence (AI) technology mean that we have well and truly crossed the valley. Synthetic faces now appear as real as genuine ones – if not more so.</p>
<figure>
<iframe width="440" height="260" src="https://www.youtube.com/embed/f0oyHtZiJJs?wmode=transparent&start=0" frameborder="0" allowfullscreen=""></iframe>
</figure>
<p>You may have come across the website <a href="https://www.thispersondoesnotexist.com/">ThisPersonDoesNotExist.com</a>. By repeatedly visiting the page, you can generate an unlimited number of images of faces, none of which belong to real people. </p>
<p>Instead, these synthetic faces are created by an AI algorithm known as a “generative adversarial network”. This is made up of two neural networks – essentially, computer models inspired by how neurons are connected in the brain. </p>
<p>These networks compete with each other. One generates new, plausible images (faces, in this case), while the other tries to discriminate real images from fake ones. Through a feedback loop, the generator learns to produce increasingly convincing images that the discriminator fails to spot as fake. </p>
<p>By using a large set of real photographs, along with the images produced by the generator, the system eventually learns to produce realistic, new examples of faces. The final generator is what’s producing the images you can see on the website.</p>
<p>Researchers have found that people shown synthetic faces mixed in with real ones <a href="https://www.pnas.org/doi/abs/10.1073/pnas.2120481119">struggle to tell the difference</a>. Participants classified the faces correctly only 48.2% of the time according to one study – slightly worse than random guessing (which would give 50% accuracy). They also rated synthetic faces as more trustworthy than real ones. </p>
<p>Another study found that synthetic faces were rated as <a href="https://theconversation.com/deepfakes-faces-created-by-ai-now-look-more-real-than-genuine-photos-197521">more real</a> than photographs of actual faces. This might be because these fake faces often look a little more average or typical than real ones (which tend to be a bit more distinctive) as a result of the generator learning that such faces are better at fooling the discriminator.</p>
<h2>Unconscious awareness in the brain</h2>
<p>In another <a href="https://www.sciencedirect.com/science/article/pii/S0042698922000852">recent study</a>, researchers in Australia delved deeper into our ability to tell the difference between real and synthetic faces. In their first experiment, online participants failed to distinguish between the two types of faces, and again perceived the synthetic faces as more real than the real ones.</p>
<p>However, their second experiment seemed to tell a different story. A new sample of participants, this time in the lab, were asked to wear electroencephalography (EEG) caps on their heads. The electrodes fitted to these caps then measured the electrical activity in the participants’ brains. </p>
<p>During the task, different faces were presented in a rapid sequence, and while this was happening, participants were asked to press a button whenever a white circle (shown on top of the faces) turned red. This ensured participants were focused on the centre of the screen where the images were being shown.</p>
<figure class="align-center ">
<img alt="A black cap covered in wires fitted over a mannequin's head." src="https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=399&fit=crop&dpr=1 600w, https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=399&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=399&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=501&fit=crop&dpr=1 754w, https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=501&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/557799/original/file-20231106-29-khqow.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=501&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Participants were asked to watch a sequence of faces while wearing a cap which measures brain activity.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/electroencephalogram-eeg-head-cap-flat-metal-602845460">Min Jing/Shutterstock</a></span>
</figcaption>
</figure>
<p>The results from the EEG test showed that brain activity differed when people were looking at real versus synthetic faces. This difference was apparent at around 170 milliseconds after the faces first appeared onscreen. </p>
<p>This N170 component of the electrical signal, as it’s known, is sensitive to the <a href="https://www.sciencedirect.com/science/article/pii/S0006899310027113">configuration of faces</a> (that is, the layout and distances between facial features). So one explanation might be that synthetic faces were perceived as subtly different to real faces in terms of the distances between features like the eyes, nose, and mouth.</p>
<p>These results suggest there is a distinction between how we behave and what our brains “know”. On the one hand, participants couldn’t consciously tell synthetic faces from real ones, but on the other, their brains could recognise the difference, as revealed by their EEG activity.</p>
<p>Although it may be surprising to think that our brains have access to information that is outside of our conscious awareness, there are many examples of this in psychology. </p>
<p>For instance, <a href="https://theconversation.com/blindsight-a-strange-neurological-condition-that-could-help-explain-consciousness-141625">blindsight</a> is a condition typically found in people who are blind in one half of their visual field. Despite this, they may be able to respond to objects placed on their blind side that they deny being consciously aware of. </p>
<p>Studies have also shown that <a href="https://www.pnas.org/doi/abs/10.1073/pnas.0605678103">our attention is drawn</a> to images of naked people, even when we’re unaware of seeing them. And we’ve all heard of the concept of subliminal advertising, although lab experiments <a href="https://www.sciencedirect.com/science/article/pii/S1053810018304306">fail to support</a> the idea that it actually works.</p>
<p>Now that synthetic faces are so easy to produce, and are as convincing as real photographs, we should be concerned about fake online profiles, fake news, and so on. Such advances in AI technology will have serious implications in the near future – there must be safeguards and other measures put in place to mitigate these dangers.</p>
<p>Perhaps the cues that our brains seem to use when spotting synthetic faces will prove useful in developing ways to identify these fakes in the coming years.</p>
<p><em>In the array of faces at the top of the article, the real and synthetic faces are as follows (from left to right):</em></p>
<p>R S S R S R</p>
<p>S R R S R R</p>
<p>R S R R R S</p><img src="https://counter.theconversation.com/content/216465/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Robin Kramer does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Scientists measured the brain activity of people trying to discern real from synthetic faces.Robin Kramer, Senior Lecturer in the School of Psychology, University of LincolnLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/2050402023-07-04T12:08:53Z2023-07-04T12:08:53ZFacial recognition technology could soon be everywhere – here’s how to make it safer<figure><img src="https://images.theconversation.com/files/524670/original/file-20230505-17-z59kxu.jpg?ixlib=rb-1.1.0&rect=442%2C22%2C3299%2C2132&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">shutterstock</span> </figcaption></figure><p>The recent coronation of <a href="https://www.theguardian.com/uk-news/2023/may/03/metropolitan-police-live-facial-recognition-in-crowds-at-king-charles-coronation">King Charles III</a> was a high-profile example of when facial recognition technology has been used to monitor a crowd, but there are plenty of others. The technology is <a href="https://www.met.police.uk/advice/advice-and-information/fr/facial-recognition-technology/">used by law enforcement</a> all over the UK and <a href="https://www.comparitech.com/blog/vpn-privacy/facial-recognition-statistics/">other countries</a>. </p>
<p>It’s now <a href="https://www.forbes.com/sites/willmcgough/2023/01/06/controversial-facial-recognition-software-has-been-implemented-at-airports-across-americahow-will-travelers-respond">common in US airports</a>. It’s being used to monitor refugees and identify dead bodies <a href="https://theconversation.com/facial-recognition-technology-how-its-being-used-in-ukraine-and-why-its-still-so-controversial-183171">in Ukraine</a>. Even Beyoncé fans have been <a href="https://news.sky.com/story/police-to-use-live-facial-recognition-in-cardiff-city-centre-for-start-of-beyonces-uk-tour-12882283">subjected to it</a>.</p>
<p>And there’s more to come. The UK government is <a href="https://www.telegraph.co.uk/news/2023/05/17/facial-recognition-police-body-cameras-chris-philip/">reportedly planning</a> to add facial recognition to the police’s body-worn devices, drones and numberplate cameras. It may soon be very difficult to leave your house without having your face scanned. </p>
<p>There are serious questions about whether the benefits of this technology outweigh such concerns. But steps could be taken to address the issues people are worried about.</p>
<h2>Uses and limits</h2>
<p>Facial recognition can be used by police to scan many faces in a crowd and compare them with a “watch list” of known criminals. This “live facial recognition” is used with the aim of <a href="https://www.pewresearch.org/internet/2022/03/17/ai-and-human-enhancement-americans-openness-is-tempered-by-a-range-of-concerns/ps_2022-03-17_ai-he_02-04/">reducing crime</a>. It can also be used retroactively on recorded CCTV footage.</p>
<p>In the UK, <a href="https://www.gov.uk/government/publications/update-to-surveillance-camera-code/amended-surveillance-camera-code-of-practice-accessible-version">the Protection of Freedoms Act 2012</a> provides a legal basis for the use of surveillance camera systems in a public place. </p>
<p>And according to the government’s <a href="https://www.gov.uk/government/consultations/surveillance-camera-code-of-practice">surveillance camera code of practice</a>, it’s justifiable to use facial recognition systems in decisions that could negatively affect people, such as whether to arrest them, so long as there is a human in the loop to supervise and make decisions.</p>
<p>So the use of facial recognition systems, or those for other types of biometric information, cannot be used for autonomous decision making, such as automatically tracking a suspect across multiple camera feeds. </p>
<h2>Problems with facial recognition</h2>
<p>But why should this be of concern to law-abiding citizens? <a href="https://www.libertyhumanrights.org.uk/issue/what-is-police-facial-recognition-and-how-do-we-stop-it">Civil liberties groups argue</a> facial recognition use in public places affects our privacy and freedom, particularly in terms of its ability to track individuals at mass gatherings and to potentially engage in racial profiling. </p>
<p>Security cameras have long captured us as we went about our daily lives. However, authorities easily being able to put a name to a face in the video footage is something we’re not so used to. </p>
<p>The technology creates a situation where many more people could get caught in the sights of the authorities than before. A person’s casual indiscretions or errors of judgement can now be easily tracked and linked to a name and address.</p>
<p>Those with a criminal record could be targeted in public based on their past, regardless of whether they intend to carry out any illegal activity. The technology could provide new opportunities <a href="https://www.brookings.edu/articles/police-surveillance-and-facial-recognition-why-data-privacy-is-an-imperative-for-communities-of-color/">for racial profiling</a>, where authorities track or suspect people based on their background, rather than because of specific information about them.</p>
<p>Facial recognition could also be used against people with no criminal past or plans to commit a crime but who the police simply want to stop, such as protesters. The Metropolitan Police may have announced that facial recognition would <a href="https://www.theguardian.com/uk-news/2023/may/03/metropolitan-police-live-facial-recognition-in-crowds-at-king-charles-coronation">not be used to target activists</a> at the coronation, but they also provoked outrage for arresting anti-monarchy demonstrators who were later <a href="https://www.theguardian.com/uk-news/2023/may/08/arrests-coronation-protesters-premeditated-republic-chief-police">released without charge</a>. </p>
<p>It’s also important to recognise facial recognition technology still suffers from inaccuracies, which can result in <a href="https://thebulletin.org/2021/11/its-time-to-address-facial-recognition-the-most-troubling-law-enforcement-ai-tool/#:%7E:text=False%20positives%2C%20on%20the%20other,with%20that%20of%20a%20criminal.">false positive matches</a> where an innocent person is mistaken for a known criminal. </p>
<p>With facial recognition posing such perceived threats, it could have a <a href="https://journals.sagepub.com/doi/10.1177/20539517211065368">chilling effect on free speech and demonstrations</a></p>
<h2>What can be done?</h2>
<p>However, there are ways that the technology could be used more safely. Law enforcement teams could perform two preliminary steps – <a href="https://en.wikipedia.org/wiki/Activity_recognition">activity recognition</a> or <a href="https://link.springer.com/article/10.1007/s11042-021-11864-2">event detection</a> – before they resort to face recognition. This approach can help minimise the potential for privacy violations and false positive matches. </p>
<p>Activity recognition refers to the process of identifying and categorising human activities or actions based on CCTV or other sensors. It aims to understand and recognise the activities of individuals or groups, which can include standard activities such as running, sitting or eating. </p>
<p>On the other hand, event detection focuses on identifying specific events or occurrences of interest within a given context. Events can range from simple events like a car passing by or a person entering a room to more complex events like accidents, fights, or more unusual behaviour. Event detection algorithms typically analyse CCTV and other sensors to detect and locate events. </p>
<p>Hence, activity recognition or event detection should be the first step before applying facial recognition to a surveillance camera feed.</p>
<p><a href="https://ieeexplore.ieee.org/abstract/document/9983830">Ensuring the data from cameras remains anonymous</a> can also enable police to study the activities of people in the crowd while preserving their privacy. Conducting regular <a href="https://ieeexplore.ieee.org/abstract/document/9166491">audits</a> and reviews can ensure that the collected data is handled responsibly and in compliance with <a href="https://ieeexplore.ieee.org/abstract/document/8793058">UK data privacy regulations</a>.</p>
<p>This can also help to address some of the concerns related to transparency and accuracy. By using activity recognition or event detection as a first step, it may be possible to give people more clarity – through signage, for example – about what exactly is going on during police surveillance in a public place.</p>
<p>It is the responsibility of the state to ensure the privacy and security of its citizens in order to foster a healthy society. But if facial recognition is implemented in a way that a significant proportion of citizens feel <a href="https://www.adalovelaceinstitute.org/report/beyond-face-value-public-attitudes-to-facial-recognition-technology/">infringes their rights</a>, it could create a culture of suspicion and a society where few people feel safe expressing themselves publicly.</p><img src="https://counter.theconversation.com/content/205040/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Nadia Kanwal does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Plans to use face recognition technology at public events raise issues about civil liberties.Nadia Kanwal, Senior Lecturer, Computer Science, Keele UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/2067012023-06-08T11:44:40Z2023-06-08T11:44:40ZHow the UK is getting AI regulation right<figure><img src="https://images.theconversation.com/files/530101/original/file-20230605-15-wc0auf.jpg?ixlib=rb-1.1.0&rect=722%2C0%2C5505%2C3720&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Regulation must protect AI innovation while addressing risks, but what's the right balance?</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/woman-holding-hologram-projection-displaying-biometric-1341188609">ra2 studio / Shutterstock</a></span></figcaption></figure><p>The latest generation of artificial intelligence (AI), such as ChatGPT, will revolutionise the way we <a href="https://www.forbes.com/sites/deborahlovich/2023/03/22/how-exactly-will-chatgpt-change-work/">live</a> and <a href="https://news.stanford.edu/2023/02/13/will-chatgpt-change-way-think-work/">work</a>. AI technologies could significantly <a href="https://www.brookings.edu/blog/future-development/2022/02/23/ai-for-social-protection-mind-the-people/">improve</a> education, healthcare, transport and welfare. But there are downsides, too: <a href="https://www.nytimes.com/2023/05/23/business/jobs-protections-artificial-intelligence.html">jobs automated out of existence</a>, <a href="https://carnegieendowment.org/2019/09/17/global-expansion-of-ai-surveillance-pub-79847">surveillance abuses</a>, and <a href="https://rm.coe.int/discrimination-artificial-intelligence-and-algorithmic-decision-making/1680925d73">discrimination</a>, including in <a href="https://hbr.org/2022/03/why-ai-failed-to-live-up-to-its-potential-during-the-pandemic">healthcare</a> and <a href="https://www.fairtrials.org/articles/news/fair-trials-tool-shows-how-predictive-policing-discriminates-and-unjustly-criminalises-people">policing</a>. </p>
<p>There’s general agreement that AI needs to be regulated, given its awesome potential for good and harm. The EU has proposed one approach, based on potential problems. The UK is proposing a different, <a href="https://www2.datainnovation.org/2021-aia-costs.pdf">pro-business</a>, approach. </p>
<p>This year, the UK government published a <a href="https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach/white-paper#fn:70">white paper</a> (a policy document setting out plans for future legislation) unveiling how it intends to regulate AI, with an emphasis on flexibility to avoid <a href="https://www.ft.com/content/a5970b6c-e731-45a7-b75b-721e90e32e1c">stifling innovation</a>. The document favours voluntary compliance, with five principles meant to tackle AI risks. </p>
<p>Strict enforcement of these principles by regulators could be added later if it’s required. But is such an approach too lenient given the risks?</p>
<h2>Crucial components</h2>
<p>The UK approach differs from the EU’s risk-based regulation. The <a href="https://www.europarl.europa.eu/meetdocs/2014_2019/plmrep/COMMITTEES/CJ40/DV/2023/05-11/ConsolidatedCA_IMCOLIBE_AI_ACT_EN.pdf">EU’s proposed AI Act</a> prohibits certain AI uses, such as <a href="https://www.amnesty.org/en/latest/news/2023/05/eu-ban-on-most-harmful-use-of-ai-moves-a-step-closer/">live facial recognition technology</a>, where people shown on a camera feed are compared against police “watch lists”, in public spaces. </p>
<p>The EU approach creates <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3896852">stringent standards</a> for so-called <a href="https://uploads-ssl.webflow.com/6305e5d52c28356b4fe71bac/638f76ec2912fe0355a3850e_Holistic-AI-Whitepaper-High-Risk-AI-Systems-Compressed.pdf">high-risk AI systems</a>. These include systems used to evaluate job applications, student admissions, eligibility for loans and public services.</p>
<p>I believe the UK’s approach better balances AI’s risks and benefits, fostering innovation that benefits the economy and society. However, critical challenges need to be addressed.</p>
<figure class="align-center ">
<img alt="Facial recognition in a crowd." src="https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=338&fit=crop&dpr=1 600w, https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=338&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=338&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=424&fit=crop&dpr=1 754w, https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=424&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/530036/original/file-20230605-19-10occ9.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=424&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">The EU’s AI Act would prohibit live face recognition by police forces in public spaces.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/cctv-ai-facial-recognition-camera-zoom-2198446519">Gorodenkoff / Shutterstock</a></span>
</figcaption>
</figure>
<p>The UK approach to AI regulation has three crucial components. First, it relies on existing legal frameworks such as privacy, data protection and product liability laws, rather than implementing new <a href="https://uk.practicallaw.thomsonreuters.com/w-039-2427?transitionType=Default&contextData=%28sc.Default%29">AI-centred legislation</a>. </p>
<p>Second, five general principles – each consisting of several components – would be applied by regulators in conjunction with existing laws. <a href="https://www.gov.uk/government/news/uk-unveils-world-leading-approach-to-innovation-in-first-artificial-intelligence-white-paper-to-turbocharge-growth#:%7E:text=Five%20principles%2C%20including%20safety%2C%20transparency,trust%20in%20this%20revolutionary%20technology.">These principles</a> are (1) “safety, security and robustness”, (2) “appropriate transparency and explainability”, (3) “fairness”, (4) “accountability and governance”, and (5) “contestability and redress”. </p>
<p>During initial implementation, regulators would not be legally required to enforce the principles. A statute imposing these obligations would be enacted later, if considered necessary. Organisations would therefore be expected to comply with the principles voluntarily in the first instance.</p>
<p>Third, regulators could adapt the five principles to the subjects they cover, with support from a <a href="https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1147045/uk_ai_regulation_impact_assessment.pdf">central coordinating body</a>. So, there will not be a single enforcement authority.</p>
<h2>Promising approach?</h2>
<p>The UK’s regime is promising for three reasons. First, it promises to use evidence about AI in its correct context, rather than applying an example from one area to another inappropriately. </p>
<p>Second, it is designed so that rules can be easily tailored to the requirements of AI used in different areas of everyday life. Third, there are advantages to its decentralised approach. For example, a single regulatory organisation, were it to underperform, would affect AI use across the board.</p>
<p>Let’s look at how it would use evidence about AI. As AI’s risks are yet to be fully understood, predicting future problems involves guesswork. To fill the gap, evidence with no relevance to a specific use of AI could be appropriated to propose drastic and inappropriate regulatory solutions.</p>
<p>For instance, some US internet companies use algorithms to <a href="https://arxiv.org/pdf/2009.11491.pdf">determine a person’s sex</a> based on facial features. These showed poor performance when presented with photos of <a href="https://proceedings.mlr.press/v81/buolamwini18a/buolamwini18a.pdf">darker-skinned women</a>. </p>
<p>This finding has been cited in support of a <a href="https://www.adalovelaceinstitute.org/wp-content/uploads/2022/06/The-Ryder-Review-Independent-legal-review-of-the-governance-of-biometric-data-in-England-and-Wales-Ada-Lovelace-Institute-June-2022.pdf">ban on law enforcement use of face recognition technology</a> in the UK. However, the two areas are quite different and problems with gender classification do not imply a similar issue with facial recognition in law enforcement.</p>
<p>These US gender algorithms work under relatively lower legal standards. Face recognition used by UK law enforcement undergoes <a href="https://theconversation.com/facial-recognition-why-we-shouldnt-ban-the-police-from-using-it-altogether-193895">rigorous testing</a>, and is deployed under <a href="https://www.judiciary.uk/wp-content/uploads/2020/08/R-Bridges-v-CC-South-Wales-ors-Judgment.pdf">strict legal requirements</a>.</p>
<figure class="align-center ">
<img alt="Driverless car." src="https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/530040/original/file-20230605-23-64ny45.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Some AI applications, such as driverless cars, could fall under more than one regulatory regime.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/man-autonomous-driving-test-vehicle-251998009">riopatuca / Shutterstock</a></span>
</figcaption>
</figure>
<p>Another advantage of the UK approach is its <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=653781">adaptability</a>. It can be difficult to predict potential risks, particularly with <a href="https://artificialintelligenceact.eu/wp-content/uploads/2022/05/General-Purpose-AI-and-the-AI-Act.pdf">AI that could be appropriated for purposes other than the ones foreseen by its developers</a> and machine learning systems, which improve in their performance over time.</p>
<p>The framework allows regulators to quickly address risks as they arise, avoiding lengthy debates in parliament. Responsibilities would be spread between different organisations. Centralising AI oversight under a single national regulator could lead to inefficient enforcement. </p>
<p>Regulators with expertise in specific areas such as transport, aviation, and financial markets are <a href="https://www2.datainnovation.org/2019-ftc-competition-consumer-protection.pdf">better suited</a> to regulate the use of AI within their fields of interest. </p>
<p>This decentralised approach could minimise the effects of corruption, of regulators becoming preoccupied with concerns other than the public interest and differing approaches to enforcement. It also avoids a single point of enforcement failure.</p>
<h2>Enforcement and coordination</h2>
<p>Some businesses could resist voluntary standards, so, if and when regulators are granted enforcement powers, they should be able to issue fines. The public should also have the right to seek compensation for harms caused by AI systems.</p>
<p>Enforcement needn’t undermine flexibility. Regulators can still tighten or loosen standards as required. However, the UK framework could encounter difficulties where AI systems fall under the jurisdiction of multiple regulators, resulting in overlaps. For example, transport, insurance, and data protection authorities could all issue conflicting guidelines for self-driving cars. </p>
<p>To tackle this, the white paper suggests establishing a <a href="https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1147045/uk_ai_regulation_impact_assessment.pdf">central body</a>, which would ensure the harmonious implementation of guidance. It’s vital to compel the different regulators to consult this organisation rather than leaving the decision up to them.</p>
<p>The UK approach shows promise for fostering innovation and addressing risks. But to strengthen the country’s position as a <a href="https://www.forbes.com/sites/edstacey/2022/02/16/achieving-the-national-ai-strategy-of-making-the-uk-a-global-science-superpower-will-require-the-creation-and-nurturing-of-an-ai-and-data-ecosystem/">leader</a> in the area, the framework must be aligned with regulation elsewhere, especially the EU.</p>
<p>Fine-tuning the framework can enhance legal certainty for businesses and bolster public trust. It will also foster international confidence in the UK’s system of regulation for this transformative technology.</p><img src="https://counter.theconversation.com/content/206701/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Asress Adimi Gikay does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>With fine-tuning, the approach correctly balances the risks with the need to encourage innovation.Asress Adimi Gikay, Senior Lecturer in AI, Disruptive Innovation and Law, Brunel University LondonLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1997612023-02-16T14:37:22Z2023-02-16T14:37:22ZHow to poll 93 million voters – the challenge of pulling off Nigeria’s presidential elections<figure><img src="https://images.theconversation.com/files/510325/original/file-20230215-29-2l84q.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Voters display their permanent voters card during the 2019 Presidential and National Assembly elections in Lagos. </span> <span class="attribution"><span class="source">Adekunle Ajayi/NurPhoto via Getty Images</span></span></figcaption></figure><p>Nigeria’s registered voters, which the <a href="https://www.inecnigeria.org/">Independent National Electoral Commission</a> has put at <a href="https://www.thecable.ng/preliminary-number-of-registered-voters-in-nigeria-now-93-5m-says-inec">93.5 million</a>, are expected to come out in their numbers in what will be Africa’s biggest election this year.</p>
<p>They will be electing the president and members of the National Assembly on 25 February and governors and members of the State Houses of Assembly on 11 March.</p>
<p>To vote in the elections, Nigerian citizens must be at least 18 years old and must have collected their permanent voter’s card by 5 February. The electoral commission has not yet released the number of people who have collected their cards. The number of collected cards will determine how many people that can be expected to vote.</p>
<p>The logistical challenges for the 2023 elections are huge, given the fact that 18 political parties are contesting, the security environment and the number of contestants at various levels. There are 18 presidential candidates, 1,101 candidates for the Senate and 3,122 candidates vying for federal constituencies in the House of Representatives. The elections will be conducted across 176,606 polling stations.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=600&fit=crop&dpr=1 600w, https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=600&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=600&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=754&fit=crop&dpr=1 754w, https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=754&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/511942/original/file-20230223-330-uixwc4.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=754&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Nigeria election infographic.</span>
<span class="attribution"><span class="source">Infographic: The Conversation Africa / Data: INEC Nigeria – Independent National Electoral Commission</span></span>
</figcaption>
</figure>
<p>But the presidential election, a three-horse race, could end in a runoff. Candidates of the ruling <a href="https://apc.com.ng/">All Progressives Congress</a>, <a href="https://peoplesdemocraticparty.com.ng/">People’s Democratic Party</a> and the <a href="https://labourparty.com.ng/">Labour Party</a> command a large national following, as shown by several <a href="https://www.stears.co/premium/article/stears-poll-predicts-nigerias-next-president/">pre-election polls</a>. </p>
<p>And the cost is huge. Nigeria spends about <a href="https://www.eisa.org/pdf/JAE11.1Aregbeyen.pdf">2% of its GDP</a> on elections. </p>
<p>Logistics, security challenges and malpractice in past elections have led to a focus on the <a href="https://www.researchgate.net/publication/350467727_Aiyede_Matters_Arising_from_the_2019_Elections_and_Electoral_Reform_Processes">reform of election administration</a>.</p>
<p>The <a href="https://www.yiaga.org/sites/default/files/portfolio/Signed-Version-of-the-Electoral-Act-2022-compressed.pdf">Electoral Act 2022</a> has given legal backing for any voter accreditation technology that the electoral commission uses. If any technical device used in the election fails to function and isn’t replaced, the election will be cancelled for that voting station and another scheduled within 24 hours. The law also allows the commission to transmit election results electronically. These steps greatly reduce the ability to rig results, compared with manual methods.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/nigerian-elections-eight-issues-young-people-want-the-new-government-to-address-199034">Nigerian elections: Eight issues young people want the new government to address</a>
</strong>
</em>
</p>
<hr>
<h2>The logistics</h2>
<p>A total of 1,265,227 officials have been trained and will be deployed for the elections. They include presiding, collation and returning officers, as well as 530,538 polling unit security officials.</p>
<p>The Independent National Electoral Commission will issue 1,642,386 identification tags for the polling and collation officers, and provide 176,846 <a href="https://dubawa.org/what-you-need-to-know-about-bimodal-voter-accreditation-system/">Bimodal Voter Accreditation System</a> (BVAS) devices and 17,618 BVAS machines for back-up. These devices verify the identity of voters by checking fingerprints and facial features electronically.</p>
<p>In December 2022, the electoral commission signed a memorandum of understanding with transport unions that will help deploy over one million personnel and large quantities of election materials to 774 local government areas, 8,809 electoral wards and 176,846 polling units across the country.</p>
<p>Over 100,000 vehicles and about 4,200 boats, accompanied by naval gunboats, will be used.</p>
<p>These have to be deployed under the current state of <a href="https://theconversation.com/nigerias-next-president-faces-a-collapsing-security-situation-five-things-he-can-do-188179">insecurity</a> as well as a scarcity of fuel. </p>
<iframe title="Nigeria Election 2023" aria-label="Map" id="datawrapper-chart-fzk8t" src="https://datawrapper.dwcdn.net/fzk8t/5/" scrolling="no" frameborder="0" style="width: 0; min-width: 100% !important; border: none;" height="800" data-external="1" width="100%"></iframe>
<h2>Challenges of 2023 polls</h2>
<p>The success of the 2023 general election will largely depend on the degree to which citizens can vote without impediments. But there are challenges. </p>
<p><strong>Voter apathy:</strong> Nigeria has a history of voter apathy, where a significant number of registered voters fail to show up on election day. For the 2019 general elections, the country had <a href="https://www.africanews.com/2019/01/07/84-million-nigerians-registered-to-vote-in-2019-polls-inec//">84 million registered voters</a>. Voter turnout in the presidential election was only <a href="https://punchng.com/2019-presidential-polls-only-35-of-voters-voted-inec/">35.66%</a>. In 2015 it was <a href="https://inecnigeria.org/wp-content/uploads/2019/02/Conference-Paper-by-Sakah-Saidu-Mahmud.pdf">43.65%</a>. These figures put Nigeria among the 10 countries with the <a href="https://www.icirnigeria.org/2019-election-nigeria-has-the-lowest-voter-turnout-in-africa/">lowest voter turnout</a> in the world. Rwanda recorded <a href="https://www.state.gov/report/custom/4572fd52ca/">98.15%</a> voter turnout in 2017, the highest in the world. </p>
<p><strong>Naira redesign and scarcity of fuel:</strong> The electoral environment has been bedevilled by <a href="https://www.premiumtimesng.com/news/headlines/576987-why-fuel-scarcity-persists-across-nigeria-marketers.html">scarcity of fuel</a> and <a href="https://theconversation.com/nigerias-currency-redesign-and-withdrawal-limits-questionable-policy-and-bad-timing-197813">naira notes</a>. The shortages have led to public <a href="https://dailypost.ng/2023/02/14/cash-crunch-protesters-shut-down-lagos-abeokuta-expressway/">demonstrations</a> and heightened tension which might deter some voters from coming out on election day.</p>
<p><strong>Insecurity:</strong> Fifty-two electoral commission offices were <a href="https://leadership.ng/arson-inec-records-50-attacks-in-4-years/">destroyed or burnt</a> between 2019 and 2022. Secessionist movements and militants from the southern regions and religious extremists and bandits in the north have <a href="https://theconversation.com/election-violence-in-nigerias-south-east-is-threatening-to-derail-voting-in-the-region-198610">besieged</a> electoral facilities. This may discourage prospective voters.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/africas-largest-democracy-goes-to-the-polls-amid-rising-insecurity-193960">Africa’s largest democracy goes to the polls amid rising insecurity</a>
</strong>
</em>
</p>
<hr>
<h2>The voting process</h2>
<p>There are four steps in the voting procedure to be followed on election day: accreditation, voting, sorting and counting, recording and announcement of results. </p>
<p><strong>Accreditation:</strong> Voters, armed with their permanent voter’s card, must be present at the polling unit where they are registered between 8.30 am and 2.30 pm. They need to queue up in an orderly manner for accreditation. Voters will present their card to the assistant polling officer, who will use the BVAS device to check that voters match their cards. Where the fingerprint fails to confirm the match, the BVAS will be used to verify the facial identity of the voter.</p>
<p><strong>Voting:</strong> After accreditation, voters will be given the ballot paper. They will go to the voting booth to make their choice on the ballot paper in secret by thumb printing. Then they put the ballot paper in the ballot box in full view of everyone present, without disclosing how they voted. Voting will be declared closed when the last voter in the queue has voted. Voters may remain at the polling unit to watch the vote.</p>
<p><strong>Sorting and counting:</strong> The ballots will be sorted and counted in full view of everyone at the polling unit. </p>
<p><strong>Recording and annoucement:</strong> The results will be filled into the result sheet and announced by the electoral commission officer at the polling unit. The results from the polling units will be taken to the various levels of collation. The sum of the results will be recorded and at the final level, the candidate who meets the criteria will be announced the winner. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/nigerias-2023-presidential-election-10-factors-that-could-affect-the-outcome-195247">Nigeria's 2023 presidential election: 10 factors that could affect the outcome</a>
</strong>
</em>
</p>
<hr>
<h2>Conclusion</h2>
<p>To vote validly, the voter must be aware of the process of voting beyond being registered to vote and collecting the permanent voter’s card. </p>
<p>Invalid ballots during the vote count will widely be attributed to inadequate civic and voter education. Thus, voter education is central to increasing voter turnout and reducing incidents of invalid votes. </p>
<p>For these reasons, the Independent National Electoral Commission made a <a href="https://www.inecnigeria.org/downloads-all/inec-strategic-plan-2017-2021/">plan</a> for voter education. It has a voter education manual and <a href="https://inecnigeria.org/voter-education/faqs/">frequently asked questions</a>. Observer <a href="https://www.wfd.org/where-we-work/nigeria">organisations</a> have also tried to help <a href="https://yiaga.org/election101">prepare</a> voters.</p>
<p>For a peaceful, free, fair and credible election, citizens must stay up to date, including knowing the location of their polling place before election day, and contribute to keeping the peace.</p><img src="https://counter.theconversation.com/content/199761/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Emmanuel Remi Aiyede does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Logistical challenges facing the 2023 elections remain huge given the number of political parties, the security environment and the number of contestants at various levels.Emmanuel Remi Aiyede, Professor of Political Institutions, Governance and Public Policy, University of IbadanLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1821002022-06-30T12:23:16Z2022-06-30T12:23:16ZPeople vary a lot in how well they recognize, match or categorize the things they see – researchers attribute this skill to an ability they call ‘o’<figure><img src="https://images.theconversation.com/files/471229/original/file-20220627-14-d5x5pz.jpg?ixlib=rb-1.1.0&rect=30%2C100%2C5854%2C4365&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Some people are inherently better at tasks like reading X-rays.</span> <span class="attribution"><a class="source" href="https://www.gettyimages.com/detail/photo/pediatrician-shows-concerned-father-foot-x-rays-royalty-free-image/1061001356">SDI Productions/E+ via Getty Images</a></span></figcaption></figure><p>Like snowflakes, no two people are exactly the same. You’re probably used to the idea that people differ substantially in personality and in cognitive abilities – skills like problem-solving or remembering information.</p>
<p>In contrast, there’s <a href="https://doi.org/10.1177/0963721417737151">a widely held intuition</a> that people vary far less in their ability to recognize, match or categorize objects. Many everyday tasks, hobbies and even critical jobs – like interpreting satellite imagery, matching fingerprints or diagnosing medical conditions – rely on these perceptual skills. The common expectation is that smart and motivated people who receive the appropriate training should eventually be able to excel at occupations that require hundreds of perceptual decisions every day.</p>
<p><a href="https://scholar.google.com/citations?user=zMFcCjEAAAAJ&hl=en&oi=ao">We</a> <a href="https://scholar.google.com/citations?user=dxEzLKAAAAAJ&hl=en&oi=ao">are</a> psychologists who measure how people compare on challenging perceptual tasks. Our research has found that this intuition that everyone has the same capacity for perceptual skills is not supported by the evidence. </p>
<p>It’s not a problem if you choose to spend every weekend bird-watching without ever getting very good at it – you may still get some fresh air and have fun. But when perceptual decisions influence safety, health or legal outcomes, there’s a case for seeking people who can achieve the best possible performance. Our research suggests some people are just better than others at learning to discriminate things perceptually, whatever the objects may be.</p>
<h2>A general ability to recognize things</h2>
<p><a href="https://doi.org/10.1037/11491-006">Classic psychological studies</a> at the turn of the 20th century discovered that performance across a range of cognitive tasks designed to test memory, math and verbal skills is correlated. In real life, this means someone who is great at sudoku is also likely to be good at memorizing their shopping list. This finding led to the modern notion of general intelligence, describing a collection of faculties that together predict a wide range of outcomes, from <a href="https://doi.org/10.1037/a0015497">income</a> to <a href="https://doi.org/10.1111/j.0963-7214.2004.01301001.x">health and longevity</a>.</p>
<p>In a similar way, our studies reveal that those who are the <a href="https://doi.org/10.1037/xge0001100">best at bird recognition may also excel at plane identification</a>,
and they may also be the best at learning to spot tumors in <a href="https://doi.org/10.1002/acp.3460">chest X-rays</a>. In other research, the same ability predicted better performance in <a href="https://doi.org/10.3758/s13414-021-02349-3">reading musical notation</a> or <a href="https://www.visionsciences.org/presentation/?id=4101">recognizing images of prepared food</a>.</p>
<p>Of course, people vary in their experience with birds or medical images. The more familiar you are with them, the <a href="https://doi.org/10.1186/s41235-017-0073-4">better you are at recognizing them</a>. Experience and training have an important role in how people make decisions based on visual information. But does everyone start on the same footing when they begin training?</p>
<h2>Does everyone start at square one?</h2>
<p>We were interested in whether everyone starts at about the same baseline of perceptual talent. To investigate this question, we measured people’s abilities with artificial objects they had never seen, to prevent any advantage due to different levels of experience.</p>
<p>In <a href="https://doi.org/10.1037/rev0000129">one large study</a>, we assessed 246 people for 13 hours each, testing them on several tasks with six categories of computer-generated artificial objects. We asked people to remember and recognize objects, to match them, or to make judgments about some of their parts.</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="images of abstract objects, a chest X-ray, four versions of a prepared food and four imaginary robots" src="https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=581&fit=crop&dpr=1 600w, https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=581&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=581&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=730&fit=crop&dpr=1 754w, https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=730&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/469066/original/file-20220615-9175-6vr9hj.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=730&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Examples of tasks that tap into o, from top left: 1) Are these two objects identical despite the change in viewpoint? 2) Which lung has a tumor? 3) Which of these dishes is the oddball? 4) Which option is the average of the four robots on the right?
Answers: 1) no 2) left 3) third 4) fourth.</span>
<span class="attribution"><span class="source">Isabel Gauthier</span>, <a class="license" href="http://creativecommons.org/licenses/by-nd/4.0/">CC BY-ND</a></span>
</figcaption>
</figure>
<p>Our results across tasks like these repeatedly reveal that people vary as much in perceptual abilities as they do in cognitive skills. Using <a href="https://doi.org/10.1146/annurev.psych.53.100901.135239">statistical methods</a> historically applied to intelligence and personality tests, we found that over 89% of the differences between people in their performance with these different tasks and categories could be explained by a general ability. We called this ability “o” for object recognition and in honor of the “g” factor, which stands for similar statistical evidence for general intelligence. </p>
<p>In <a href="https://doi.org/10.1037/xge0001100">follow-up studies</a>, we’ve found that o applies in the same way to artificial and real objects, and that people with high o are better at computing summary statistics for groups of objects (such as estimating the “average” of several objects) and also better at <a href="https://doi.org/10.1007/s00426-021-01560-z">recognizing objects by touch</a>. You can compare yourself to others in <a href="https://jasonkchow.github.io/ov_demo/">this short demo</a>.</p>
<h2>o is a distinct ability</h2>
<p>Since it is so general, is o just another name for general intelligence? We don’t think so.</p>
<p>In one study, we found that <a href="https://doi.org/10.1016/j.cognition.2017.05.019">neither IQ nor SAT scores predict recognition</a> of novel objects. <a href="https://doi.org/10.1037/rev0000129">In other work</a>, we found that o was distinct from g, but also from the personality trait of conscientiousness. This means that book smarts may not be enough to excel in domains that rely heavily on perceptual abilities.</p>
<p>We tested this idea by measuring how good people with or without expertise in radiology were at detecting lung nodules in chest X-rays. Those with the highest o were better at this task, even after controlling for intelligence and experience in radiology. This finding demonstrates the added value of measuring o. Even when medical students are selected to be smart and provided with training, it may not guarantee the highest levels of performance in specializations that rely on perceptual skills.</p>
<p>Many doors open when you demonstrate that you’re cognitively talented, which seems only fair. But it is fair only to the extent that general intelligence is the best way – or even a sufficient way – to predict success in a given domain. Many have raised warnings that intelligence testing can lead to inequities in hiring or career placement tied to race, gender or socioeconomic status.</p>
<p>Over the years, many thinkers have downplayed innate talents to emphasize environmental influences. They argued that success can be shaped through years of <a href="https://doi.org/10.1111/j.1553-2712.2008.00227.x">deliberate practice</a>, programs to change one’s <a href="https://doi.org/10.1080/00461520.2012.722805">attitudes about learning</a>, or even <a href="https://doi.org/10.1038/nrn3135">hours of playing video games</a>. </p>
<p>But the evidence in favor of the influence of innate talents remains strong, and denying them or overpromising on the efficacy of environmental factors <a href="https://doi.org/10.1177/0963721418797300">may sometimes be harmful</a>. People can waste time and resources that could be better invested, and may run the risk of experiencing stigma if their efforts do not succeed because of factors they cannot control.</p>
<p>One answer to this problem is to learn more about talents beyond those related to intelligence and then to make better use of them. Classical notions of intelligence may be just one factor of many that determine overall ability. An increased focus on perceptual abilities, specifically those that are general, could help reduce inequities. For instance, while differences in experience can drive <a href="https://doi.org/10.1016/j.visres.2016.10.003">sex differences in the recognition of objects in some familiar categories</a>, we’ve found <a href="https://doi.org/10.1037/xge0001100">no such differences in the general ability o</a>.</p><img src="https://counter.theconversation.com/content/182100/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>The authors do not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>To achieve perceptual expertise, you may need more than smarts and hard work. Research suggests there’s a general ability that may help you succeed in jobs that depend on perceptual decisions.Isabel Gauthier, David K. Wilson Professor of Psychology, Vanderbilt UniversityJason Chow, Ph.D. Student in Psychological Sciences, Vanderbilt UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1851262022-06-15T07:07:57Z2022-06-15T07:07:57ZBunnings, Kmart and The Good Guys say they use facial recognition for ‘loss prevention’. An expert explains what it might mean for you<figure><img src="https://images.theconversation.com/files/468915/original/file-20220615-14-ex57sp.jpg?ixlib=rb-1.1.0&rect=348%2C128%2C4215%2C2524&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption"></span> <span class="attribution"><a class="license" href="http://creativecommons.org/licenses/by/4.0/">CC BY</a></span></figcaption></figure><p>Once the purview of law enforcement and intelligence agencies, facial recognition is now being used to identify consumers in Australian stores. </p>
<p>If you’ve seen the movie Minority Report, you’ll remember how Tom Cruise’s character John Anderton is identified through iris recognition to perform his duties, and later tracked with it when he’s a wanted man. When he replaces his eyes to evade identification, Anderton is bombarded with advertisements targeting his new assumed identity.</p>
<p>This once-futuristic idea from a movie could soon be a reality in our lives. An investigative report published by consumer magazine <a href="https://www.choice.com.au/consumers-and-data/data-collection-and-use/how-your-data-is-used/articles/kmart-bunnings-and-the-good-guys-using-facial-recognition-technology-in-store">Choice</a> reveals three major retailers (out of 25 queried), Kmart, Bunnings and The Good Guys, have admitted using facial recognition technology on customers for “loss prevention”. </p>
<p>The companies say they advise consumers of the use of the technology as a condition of entry. But do consumers really know what this entails, and how or where their images could be used or stored?</p>
<h2>What is facial recognition and why do we care?</h2>
<p>We’ve grown accustomed to our phones and cameras using facial detection software to put our faces into focus. But facial <em>recognition</em> technology takes this a step further by matching our unique identifying information to a stored digital image.</p>
<p>Facial recognition has come a long way. It was initially used in 2001 to identify relationships between gamblers and employees in Las Vegas casinos, where there was suspected collusion. </p>
<p>The United States government would eventually use <a href="https://www.infoworld.com/article/2628017/innovation-that-matters--jeff-jonas-connects-the-invisible-dots.html">the same</a> technology to <a href="https://www.nationalgeographic.com/science/article/140505-jeff-jonas-big-data-gambling-computers-technology-ibm">identify the 9/11 hijackers</a>. It’s now widely adopted by law enforcement and intelligence communities.</p>
<p>Currently, software such as Clearview AI and PimEyes are being used in highly sophisticated ways, including by Ukrainian and Russian forces to <a href="https://www.washingtonpost.com/technology/2022/04/15/ukraine-facial-recognition-warfare/">identify combatants in Ukraine</a>. </p>
<h2>But what is this technology doing in Bunnings?</h2>
<p>As with its early use in casinos, Kmart, Bunnings and The Good Guys told Choice their facial recognition software is used for “loss prevention”.</p>
<p>Images captured on store surveillance devices and body cameras could be used to identify in-store individuals engaged in theft, or other criminal activities. Real-time identification could allow law enforcement to quickly identify shoppers with unpaid tickets, outstanding warrants, or existing criminal complaints.</p>
<p>Bunnings chief operating officer Simon McDowell told SBS News the technology was used “solely to keep team and customers safe and prevent unlawful activity in our stores”. Both The Good Guys and Kmart told <a href="https://www.theguardian.com/technology/2022/jun/15/bunnings-kmart-and-the-good-guys-using-facial-recognition-technology-to-crack-down-on-theft-choice-says">news outlets</a> they were using it for the same reasons, in a select number of stores – and that customers were notified through signage. </p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=354&fit=crop&dpr=1 600w, https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=354&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=354&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=445&fit=crop&dpr=1 754w, https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=445&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/468933/original/file-20220615-25-71yxl3.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=445&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Choice supplied this photo of a sign, which it said was taken at a Kmart in Marrickville, NSW.</span>
<span class="attribution"><span class="source">CHOICE</span></span>
</figcaption>
</figure>
<p>Choice confirmed there were some signs disclosing use of the technology – but reported these signs were small and would be missed by most shoppers. </p>
<p>The news has stoked shoppers’ fears of how their image data may be used. As in Minority Report, images captured in a store could theoretically be used for targeted advertising and to “enhance” <a href="https://www.wired.com/2011/11/malls-track-phone-signals/">the shopping experience</a>.</p>
<p>It’s likely images and video collected through standard in-store surveillance are either matched immediately against a remote database using specialised facial recognition software, or analysed against a database of tagged and catalogued images later on. Ideally, the images would be encoded and stored in a file that’s readable only by the algorithm specific to the device or software processor.</p>
<h2>Potential for misuse</h2>
<p>We have already seen online retailers use this tactic through <a href="https://theconversation.com/googles-scrapping-third-party-cookies-but-invasive-targeted-advertising-will-live-on-156530">cookies</a> and linking our purchase history on <a href="https://theconversation.com/smartphone-data-tracking-is-more-than-creepy-heres-why-you-should-be-worried-91110">electronic devices</a>. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/is-your-phone-really-listening-to-your-conversations-well-turns-out-it-doesnt-have-to-162172">Is your phone really listening to your conversations? Well, turns out it doesn't have to</a>
</strong>
</em>
</p>
<hr>
<p>We have also seen companies correlate our social media profiles and our other online experiences across various websites. Australian stores employing facial recognition could use collected information internally to track:</p>
<ul>
<li>the number of visits by a person</li>
<li>the times of those visits</li>
<li>pattern or behavioural analysis (such as a consumer’s reaction to pricing or signage) and</li>
<li>associations with other shoppers (such as friends, family and anyone else with them). </li>
</ul>
<p>Retailers could also use this identity data to extract information from social media, where most people have images of themselves uploaded. They could then perform risk analysis based on the credit and financial reporting access of that specific shopper. </p>
<p>Externally, the images and associated consumer information could be merged with financial, economic, social and political data already collected by commercial data aggregators – adding to the already massive data aggregation market.</p>
<p>Current Australian privacy laws require retailers to disclose what data are being collected, retained and protected, as well as how it might be used outside of a loss prevention model.</p>
<p>A Bunnings spokesperson told The Guardian the technology was being used in line with the Australian Privacy Act. Choice has reached out to the Office of the Australian Information Commissioner to determine whether the use of the technology is indeed consistent with the Privacy Act.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/shadow-profiles-facebook-knows-about-you-even-if-youre-not-on-facebook-94804">Shadow profiles - Facebook knows about you, even if you're not on Facebook</a>
</strong>
</em>
</p>
<hr>
<h2>What to do?</h2>
<p>While the retailers highlighted in Choice’s investigation state consumers must agree to the collection of their images as a condition of entry, the reality is the collection, retention, and use of their images are not usually disclosed in any explicit way. </p>
<p>As far as data collection in retail settings goes, there should be a precondition for all stores to make sure consumers are made aware of:</p>
<ul>
<li>the specific information that is collected while they are visiting</li>
<li>how it might be aggregated and combined with other relevant information from third parties</li>
<li>how long the images or data will be retained, retrieved, or accessed and by whom, and </li>
<li>what security precautions are being used to secure the data.</li>
</ul>
<p>Furthermore, as with their online shopping experience, consumers should be given the option to opt-out of such data collection. </p>
<p>Until then, consumers may try to avoid collection by donning hats, sunglasses and face masks. But considering the rate at which facial recognition technology is advancing – and how large the personal data market has already grown – retail cameras may soon be able to see through these disguises, too.</p><img src="https://counter.theconversation.com/content/185126/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Dennis B Desmond previously received funding from the United States Department of Defense.</span></em></p>Australia’s consumer advocacy group Choice identified three Australian retailers who use facial recognition to identify consumers. What are the privacy concerns?Dennis B. Desmond, Lecturer, Cyberintelligence and Cybercrime Investigations, University of the Sunshine CoastLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1834472022-05-24T06:01:41Z2022-05-24T06:01:41ZPay ‘with a smile or a wave’: why Mastercard’s new face recognition payment system raises concerns<figure><img src="https://images.theconversation.com/files/464421/original/file-20220520-19-yabkx8.jpeg?ixlib=rb-1.1.0&rect=0%2C31%2C3943%2C2787&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">
</span> <span class="attribution"><span class="source">Shutterstock</span></span></figcaption></figure><p>Mastercard’s <a href="https://www.mastercard.com/news/press/2022/may/with-a-smile-or-a-wave-paying-in-store-just-got-personal/">“smile to pay”</a> system, announced last week, is supposed to save time for customers at checkouts. It is being trialled in Brazil, with future pilots planned for the Middle East and Asia.</p>
<p>The company argues touch-less technology will help speed up transaction times, shorten lines in shops, heighten security and improve hygiene in businesses. But it raises concerns relating to customer privacy, data storage, crime risk and bias. </p>
<h2>How will it work?</h2>
<p>Mastercard’s biometric checkout system will provide customers facial recognition-based payments, by linking the biometric authentication systems of a number of third-party companies with Mastercard’s own payment systems. </p>
<p>A Mastercard spokesperson told The Conversation it had already partnered with NEC, Payface, Aurus, Fujitsu Limited, PopID and PayByFace, with more providers to be named. </p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="The 'Fujitsu' logo in red is displayed on a building's side" src="https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/464953/original/file-20220524-22-ga0v7l.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Mastercard has partnered with Fujitsu, a massive information and communications technology firm offering many different products and services.</span>
<span class="attribution"><span class="source">Shutterstock</span></span>
</figcaption>
</figure>
<p>They said “providers need to go through independent laboratory certification against the program criteria to be considered” – but details of these criteria aren’t yet publicly available.</p>
<p>According to <a href="https://www.siliconrepublic.com/business/mastercard-facial-recognition-biometric-payments">media</a> reports, customers will have to install an app which will take their picture and payment information. This information will be saved and stored on the third-party provider’s servers. </p>
<p>At the checkout, the customer’s face will be matched with the stored data. And once their identity is verified, funds will be deducted automatically. The “wave” option is a bit of a trick: as the customer watches the camera while waving, the camera still scans their face – not their hand.</p>
<p>Similar authentication technologies are used on smartphones (face ID) and in many airports around the world, including “<a href="https://www.abf.gov.au/entering-and-leaving-australia/smartgates/arrivals">smartgates</a>” in Australia.</p>
<p><a href="https://www.theverge.com/2017/9/4/16251304/kfc-china-alipay-ant-financial-smile-to-pay">China</a> started using biometrics-based checkout technology back in 2017. But Mastercard is among the first to launch such a system in Western markets – competing with the “pay with your palm” <a href="https://techcrunch.com/2020/09/29/amazon-introduces-the-amazon-one-a-way-to-pay-with-your-palm-when-entering-stores/">system</a> used at cashier-less Amazon Go and Whole Foods brick and mortars in the United States.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/ai-facial-analysis-is-scientifically-questionable-should-we-be-using-it-for-border-control-155474">AI facial analysis is scientifically questionable. Should we be using it for border control?</a>
</strong>
</em>
</p>
<hr>
<h2>What we don’t know</h2>
<p>Much about the precise functioning of Mastercard’s system isn’t clear. How accurate will the facial recognition be? Who will have access to the databases of biometric data? </p>
<p>A Mastercard spokesperson told The Conversation customers’ data would be stored with the relevant biometric service provider in encrypted form, and removed when the customer “indicates they want to end their enrolment”. But how will the removal of data be enforced if Mastercard itself can’t access it?</p>
<p>Obviously, privacy protection is a major concern, especially when there are many potential third-party providers involved.</p>
<p>On the bright side, Mastercard’s <a href="https://www.investopedia.com/articles/markets/032615/how-mastercard-makes-its-money-ma.asp">customers</a> will have a choice as to whether or not they use the biometrics checkout system. However, it will be at retailers’ discretion whether they offer it, or whether they offer it exclusively as the only payment option.</p>
<p>Similar face-recognition technologies used in airports, and <a href="https://www.brookings.edu/research/police-surveillance-and-facial-recognition-why-data-privacy-is-an-imperative-for-communities-of-color/">by police</a>, often offer no choice. </p>
<p>We can assume Mastercard and the biometrics provider with whom they partner will require customer consent, as per most privacy laws. But will customers know what they are consenting to? </p>
<p>Ultimately, the biometric service providers Mastercard teams up with will decide how they use the data, for how long, where they store it, and who can access it. Mastercard will merely decide what providers are “good enough” to be accepted as partners, and the minimum standards they must adhere to. </p>
<p>Customers who want the convenience of this checkout service will have to consent to all the related data and privacy terms. And as reports have noted, there is potential for Mastercard to integrate the feature with loyalty schemes and make personalised recommendations <a href="https://www.cnbc.com/2022/05/17/mastercard-launches-tech-that-lets-you-pay-with-your-face-or-hand.html">based on purchases</a>. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/fingerprint-login-should-be-a-secure-defence-for-our-data-but-most-of-us-dont-use-it-properly-127442">Fingerprint login should be a secure defence for our data, but most of us don't use it properly</a>
</strong>
</em>
</p>
<hr>
<h2>Accuracy is a problem</h2>
<p>While the accuracy of face recognition technologies has previously been challenged, the current <em>best</em> facial authentication algorithms have an error of just 0.08%, according to tests by the <a href="https://github.com/usnistgov/frvt/blob/nist-pages/reports/1N/frvt_1N_report_2020_03_27.pdf">National Institute of Standards and Technology</a>. In some countries, even banks have <a href="https://techhq.com/2020/09/biometrics-the-most-secure-solution-for-banking/">become comfortable</a> relying on it to log users into their accounts.</p>
<p>Yet we can’t know how accurate the technologies used in Mastercard’s biometric checkout system will be. The algorithms underpinning a technology can work almost perfectly when trailed in a lab, but perform <a href="https://www.csis.org/blogs/technology-policy-blog/how-accurate-are-facial-recognition-systems-%E2%80%93-and-why-does-it-matter">poorly</a> in real life settings, where lighting, angles and other parameters are varied.</p>
<h2>Bias is another problem</h2>
<p>In a 2019 study, NIST <a href="https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf#page=5">found</a> that out of 189 facial recognition algorithms, the majority were biased. Specifically, they were less accurate on people from racial and ethnic minorities. </p>
<p>Even if the technology has improved in the past few years, it’s not foolproof. And we don’t know the extent to which Mastercard’s system has overcome this challenge.</p>
<p>If the software fails to recognise a customer at the check out, they might end up disappointed, or even become irate – which would completely undo any promise of speed or convenience.</p>
<p>But if the technology misidentifies a person (for instance, John is recognised as Peter – or <a href="https://www.youtube.com/watch?v=e8-yupM-6Oc">twins are confused</a> for each other), then money could be taken from the wrong person’s account. How would such a situation be dealt with?</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=617&fit=crop&dpr=1 600w, https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=617&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=617&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=776&fit=crop&dpr=1 754w, https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=776&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/464424/original/file-20220520-19-5hfuvx.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=776&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">There’s no evidence facial recognition technology is infallible. These systems can misidentify and also have biases.</span>
<span class="attribution"><span class="source">Shutterstock</span></span>
</figcaption>
</figure>
<h2>Is the technology secure?</h2>
<p>We often hear about software and databases being hacked, even in <a href="https://www.csoonline.com/article/2130877/the-biggest-data-breaches-of-the-21st-century.html">cases of</a> supposedly very “secure” organisations. Despite Mastercard’s <a href="https://wwmastw.cnbc.com/2022/05/17/mastercard-launches-tech-that-lets-you-pay-with-your-face-or-hand.html">efforts</a> to ensure security, there’s no guarantee the third-party providers’ databases – with potentially millions of people’s biometric data – won’t be hacked.</p>
<p>In the wrong hands, this data could lead to <a href="https://www.comparitech.com/identity-theft-protection/identity-theft-statistics/">identity theft</a>, which is one of the fastest growing types of crime, and financial fraud. </p>
<h2>Do we want it?</h2>
<p>Mastercard suggests 74% of customers are in favour of using such technology, referencing a stat from its <a href="https://www.mastercard.com/news/ap/en/newsroom/press-releases/en/2020/april/mastercard-study-shows-consumers-moving-to-contactless-payments-for-everyday-purchases/">own study</a> – also used by <a href="https://www.mastercard.com/news/ap/en/newsroom/press-releases/en/2020/october/mastercard-idemia-and-matchmove-pilot-fingerprint-biometric-card-in-asia-to-enhance-security-and-safety-of-contactless-payments">business partner</a> Idemia (a company that sells biometric identification products). </p>
<p>But the report cited is vague and brief. Other studies show entirely different results. For example, <a href="https://www.getapp.com/resources/facial-recognition-technology/#how-comfortable-are-consumers-with-facial-recognition-technology">this study</a> suggests 69% of customers aren’t comfortable with face recognition tech being used in retail settings. And <a href="https://www.securitymagazine.com/articles/93521-are-consumers-comfortable-with-facial-recognition-it-depends-says-new-study">this one</a> shows only 16% trust such tech.</p>
<p>Also, if consumers knew the risks the technology poses, the number of those willing to use it might drop even lower.</p><img src="https://counter.theconversation.com/content/183447/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Rita Matulionyte receives funding from Lithuanian Research Council for the research project 'Government Use of Facial Recognition Technologies: Legal Challenges and Possible Solutions' (2021-2023). She is affiliated with Australian Society for Computers and Law (AUSCL). </span></em></p>The technology is currently being trialled outside of Australia. It’s one of the first major attempts to bring it to western markets on a large scale.Rita Matulionyte, Senior Lecturer in Law, Macquarie UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1765892022-02-10T14:23:16Z2022-02-10T14:23:16ZFrom ‘super-recognisers’ to the ‘face blind’ – how tests reveal the underlying cognitive processes<figure><img src="https://images.theconversation.com/files/444773/original/file-20220207-23-1tnlfpp.jpg?ixlib=rb-1.1.0&rect=88%2C54%2C4477%2C3162&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">'Have we met?' Actor Brad Pitt thinks he is faceblind.</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/venice-italy-august-29-brad-pitt-1527973523">Matteo Chinellato/Shutterstock</a></span></figcaption></figure><p>The ability to recognise faces is important in many different real life contexts and fundamental to our social relationships and interactions. It allows us to identify people we care about, and respond appropriately to them. We may greet a friend differently to our boss, for example. But some people are simply better than others at recognising faces.</p>
<p>At one end of the spectrum, there are people called “<a href="https://theconversation.com/facial-recognition-research-reveals-new-abilities-of-super-recognisers-128414">super-recognisers</a>”. They find face recognition easy – often successfully identifying people even if they have only seen them once, briefly or a long time ago. Conversely, people with “<a href="https://theconversation.com/always-forget-a-face-so-does-brad-pitt-dont-just-blame-your-memory-50334">developmental prosopagnosia</a>” are at the opposite end of the face recognition scale. They find it tricky to recognise faces and can therefore struggle to socially interact with those around them. </p>
<p>People with this type of “face blindness” may not realise they have it until they reach their teenage years or later. For most of us though, our face recognition ability falls between these extremes – we aren’t super-good, but we aren’t particularly bad either. </p>
<p>It may seem that face recognition is a single, isolated skill. But psychologists know that the ability depends on a number of different cognitive processes that interact in complex ways.</p>
<h2>Measuring face recognition</h2>
<p>If we are to test face recognition accurately, it is important that we know what it is. And it turns out that <a href="https://doi.org/10.1186/s41235-018-0112-9">different tests</a> may produce different results.</p>
<p>One task used to measure how good we are at determining identity in unfamiliar faces is called “face matching”, captured, for example, by the <a href="https://doi.org/10.3758/BRM.42.1.286">Glasgow Face Matching Test</a>. Here you are presented with two faces and asked whether they belong to the same person or different people. Performance across 40 pairs, ranges from being just over chance (guessing) to 100% correct. </p>
<p>Being good at face matching is particularly important for some jobs. Indeed, passport control officers verify identity by matching the identity of a live person to a photo in a passport.</p>
<p>But being good at face matching doesn’t necessarily make you a super-recogniser. Indeed some matching tasks may not be difficult enough to allow super recognisers to show off their skills. <a href="https://pubmed.ncbi.nlm.nih.gov/16169565/">The Cambridge Face Memory Test</a> is an example of a “face learning task”. It measures your ability to learn and identify previously unknown faces. Specifically, you memorise the faces of different people and then try to pick them out of a line-up of three faces. The test starts very easy and gets progressively more difficult. Doing well on this task depends on your ability to see the visual differences between faces and to memorise them. </p>
<p>Finally, we can test familiar face recognition. When recognising familiar faces, we are tapping into our stored long-term memories of known people. We may know them as a family member or friend, or it may be someone from our favourite TV show. Most <a href="https://link.springer.com/article/10.3758/PBR.16.2.252">familiar face recognition tasks</a> ask people to try to recognise famous faces either from current footage or use a test based on celebrity faces from before they were famous.</p>
<h2>Performance across tasks</h2>
<p>Interestingly, performance across different face tasks may be associated. So, if you are good at one face recognition task then you may also be good at other face tasks too. Some researchers <a href="https://pubmed.ncbi.nlm.nih.gov/28077292/">have proposed a general face factor</a> that accounts for some correspondence in performance across tasks. </p>
<p>But this is not necessarily the case. You can be good at one task but impaired at another. And this is where the complexity lies. For example, if you have problems with working memory then you may struggle with face learning and familiar face recognition, but not necessarily with face matching. Similarly, being impaired at long term memory may make it particularly difficult to access memories of familiar people. Finally, problems with attention may affect your everyday face recognition but have less impact in the lab when you are fully focused on the task in hand.</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="A mosaic of hundres of faces." src="https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=346&fit=crop&dpr=1 600w, https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=346&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=346&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=435&fit=crop&dpr=1 754w, https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=435&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/444774/original/file-20220207-13-1bou5u9.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=435&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">We are bombarded by faces on a daily basis, which can be a challenge for those with poor memory.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/hundreds-multiracial-people-crowd-portraits-headshots-1908066802">Andrew Angelov/Shutterstock</a></span>
</figcaption>
</figure>
<p>Given the different measures of face recognition, we need to clearly define how we determine when someone is impaired (or a super-recogniser). Certainly, there is no single accepted test to measure face recognition ability. Instead, psychologists think that its best to measure face recognition using multiple tasks, exploring the different aspects of face recognition. On each task, performance from an individual is compared to that achieved by a large population of the general public. </p>
<p>We must also listen to a person’s report of their face recognition ability in their everyday lives. Some people may score normally on, for example, the Cambridge Face Memory Test in the lab, but struggle to recognise their friends and family in the street. Conversely, while poor performance across a number of face recognition tests is a good indicator for prosopagnosia, it is not necessarily a clear diagnosis of it. </p>
<p>More valid measures of face recognition that reflect real life still need to be developed. In real life, face recognition is much more complicated. Most of the existing face tasks use still images of faces, whereas in the real world faces move in complex and subtle ways. The full picture will rely on tests that can distinguish more carefully between different types of face impairment.</p><img src="https://counter.theconversation.com/content/176589/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Karen Lander does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>No single test can accurately determine whether you are a super-recogniser.Karen Lander, Senior Lecturer in Experimental Psychology, University of ManchesterLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1650052021-08-05T06:29:59Z2021-08-05T06:29:59ZWhen faces are partially covered, neither people nor algorithms are good at reading emotions<figure><img src="https://images.theconversation.com/files/414452/original/file-20210804-14-bjw0w8.jpg?ixlib=rb-1.1.0&rect=0%2C0%2C5101%2C3403&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">
</span> <span class="attribution"><span class="source">Shutterstock/Sergey Tinyakov</span></span></figcaption></figure><p>Artificial systems such as homecare robots or driver-assistance technology are becoming more common, and it’s timely to investigate whether people or algorithms are better at reading emotions, particularly given the added challenge brought on by face coverings.</p>
<p>In our recent <a href="https://psyarxiv.com/mnsyx/">study</a>, we compared how face masks or sunglasses affect our ability to determine different emotions compared with the accuracy of artificial systems. </p>
<figure class="align-center ">
<img alt="Our study used full and partial masks and sunglasses to obscure parts of the face." src="https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=116&fit=crop&dpr=1 600w, https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=116&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=116&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=145&fit=crop&dpr=1 754w, https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=145&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/414016/original/file-20210731-13-wst32o.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=145&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">The study used full and partial masks and sunglasses to obscure parts of the face.</span>
<span class="attribution"><span class="license">Author provided</span></span>
</figcaption>
</figure>
<p>We presented images of emotional facial expressions and added two different types of masks — the full mask used by frontline workers and a recently introduced mask with a transparent window to allow lip reading. </p>
<p>Our findings show algorithms and people both struggle when faces are partially obscured. But artificial systems are more likely to misinterpret emotions in unusual ways.</p>
<p>Artificial systems performed significantly better than people in recognising emotions when the face was not covered — 98.48% compared to 82.72% for seven different types of emotion. </p>
<p>But depending on the type of covering, the accuracy for both people and artificial systems varied. For instance, sunglasses obscured fear for people while partial masks helped both people and artificial systems to identify happiness correctly. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/ai-is-increasingly-being-used-to-identify-emotions-heres-whats-at-stake-158809">AI is increasingly being used to identify emotions – here's what's at stake</a>
</strong>
</em>
</p>
<hr>
<p>Importantly, people classified unknown expressions mainly as neutral, but artificial systems were less systematic. They often incorrectly selected anger for images obscured with a full mask, and either anger, happiness, neutral, or surprise for partially masked expressions.</p>
<h2>Decoding facial expressions</h2>
<p>Our ability to recognise emotion uses the visual system of the brain to interpret what we see. We even have an area of the brain specialised for face recognition, known as the fusiform face area, which helps interpret information revealed by people’s faces. </p>
<p>Together with the context of a particular situation (social interaction, speech and body movement) and our understanding of past behaviours and sympathy towards our own feelings, we can decode how people feel. </p>
<p>A system of <a href="https://www.paulekman.com/facial-action-coding-system/">facial action units</a> has been proposed for decoding emotions based on facial cues. It includes units such as “the cheek raiser” and “the lip corner puller”, which are both considered part of an expression of happiness.</p>
<figure class="align-center ">
<img alt="Study team wearing face masks" src="https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=375&fit=crop&dpr=1 600w, https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=375&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=375&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=471&fit=crop&dpr=1 754w, https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=471&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/414017/original/file-20210731-17-1e2apxb.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=471&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Can you read the researchers’ emotion from their covered faces? Both artificial systems and people are compromised in categorising emotions when faces are obscured.</span>
<span class="attribution"><span class="license">Author provided</span></span>
</figcaption>
</figure>
<p>In contrast, artificial systems analyse pixels from images of a face when categorising emotions. They pass pixel intensity values through a network of filters mimicking the human visual system. </p>
<p>The finding that artificial systems misclassify emotions from partially obscured faces is important. It could lead to unexpected behaviours of robots interacting with people wearing face masks.</p>
<p>Imagine if they misclassify a negative emotion, such as anger or sadness, as a positive emotional expression. The artificial systems would try to interact with a person taking actions on the misguided interpretation they are happy. This could have detrimental effects for the safety of these artificial systems and interacting humans.</p>
<h2>Risks of using algorithms to read emotion</h2>
<p>Our research reiterates that algorithms are susceptible to biases in their judgement. For instance, the performance of artificial systems is greatly affected when it comes to categorising emotion from natural images. Even just the sun’s angle or shade can influence outcomes.</p>
<p>Algorithms can also be racially biased. As previous studies have found, even a small <a href="https://link.springer.com/chapter/10.1007/978-3-030-72699-7_30">change to the colour</a> of the image, which has nothing to do with emotional expressions, can lead to a drop in performance of algorithms used in artificial systems. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/face-masks-and-facial-recognition-will-both-be-common-in-the-future-how-will-they-co-exist-144417">Face masks and facial recognition will both be common in the future. How will they co-exist?</a>
</strong>
</em>
</p>
<hr>
<p>As if that wasn’t enough of a problem, even <a href="https://arxiv.org/abs/1702.02284">small visual perturbations</a>, imperceptible to the human eye, can cause these systems to misidentify an input as something else. </p>
<p>Some of these misclassification issues can be addressed. For instance, <a href="https://ieeexplore.ieee.org/document/9223510">algorithms can be designed</a> to consider emotion-related features such as the shape of the mouth, rather than gleaning information from the colour and intensity of pixels. </p>
<p>Another way to address this is by changing the <a href="https://psyarxiv.com/mnsyx/">training data characteristics</a> — oversampling the training data so that algorithms mimic human behaviour better and make less extreme mistakes when they do misclassify an expression.</p>
<p>But overall, the performance of these systems drops when interpreting images in real-world situations when faces are partially covered. </p>
<p>Although robots may claim higher than human accuracy in emotion recognition for static images of completely visible faces, in real-world situations that we experience every day, their performance is still not human-like.</p><img src="https://counter.theconversation.com/content/165005/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Will Browne receives funding from Science for Technological Innovation, Ministry of Business, Innovation and Employment. </span></em></p><p class="fine-print"><em><span>Harisu Abdullahi Shehu and Hedwig Eisenbarth do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Robots are more likely than people to misclassify emotions when reading faces that are partially covered. This could lead to unexpected behaviours when they interact with people wearing masks.Harisu Abdullahi Shehu, PhD Researcher, Te Herenga Waka — Victoria University of WellingtonHedwig Eisenbarth, Senior Lecturer in Psychology, Te Herenga Waka — Victoria University of WellingtonWill Browne, Professor in Artificial Cognitive Systems, Queensland University of TechnologyLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1596672021-04-30T15:00:56Z2021-04-30T15:00:56ZWhy we can still recognise people in face masks<figure><img src="https://images.theconversation.com/files/398086/original/file-20210430-18-pdqd8o.jpeg?ixlib=rb-1.1.0&rect=753%2C0%2C3505%2C3045&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">
</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/african-american-woman-wearing-medical-mask-1793657722">oneinchpunch/Shutterstock</a></span></figcaption></figure><p>The average person knows about <a href="https://royalsocietypublishing.org/doi/full/10.1098/rspb.2018.1319">5,000 faces</a> – from family and friends to the cashier at the local store. Most people can recognise familiar faces with ease, even from low quality images, or from photos that are many years old. We often recognise familiar faces even if we <a href="https://bpspsychub.onlinelibrary.wiley.com/doi/pdf/10.1111/j.2044-8295.1986.tb02199.x">can’t remember a person’s name</a> or how we know them. </p>
<p>Most of us take this ability to recognise familiar faces for granted – but when public health issues require our friends to mask up, covering their chins, lips, cheeks and noses, are our facial recognition skills scrambled? </p>
<p>We investigated this question in <a href="https://royalsocietypublishing.org/doi/pdf/10.1098/rsos.201169">our recent study</a> and compared the impact of masks (which cover the lower portion of the face) with that of sunglasses (which cover the eye region). Despite face masks covering a large proportion of our faces, we found that people find it surprisingly easy to recognise familiar faces behind masks – speaking to the remarkable versatility of this human skill.</p>
<h2>Familiar faces</h2>
<p>Identifying familiar faces is a useful day-to-day skill, but the identification of unfamiliar faces is also important in the context of <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150036">forensic investigations</a> and <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0103510">security scenarios</a>. Our study measured the recognition of both familiar and unfamiliar faces.</p>
<p>We presented our participants with pairs of face images, and asked them to decide whether the faces belonged to the same person or different people. One image of the pair was always presented with no concealment, and the other showed either no concealment, an image in sunglasses, or in a face mask. Participants completed the task for familiar faces (images of celebrities) and for unfamiliar faces.</p>
<p>Even though face masks cover a significant part of the face, we found that our participants identified familiar faces in masks with <a href="https://royalsocietypublishing.org/doi/full/10.1098/rsos.201169">around 90% accuracy</a> – no worse than the results for faces wearing sunglasses, and only slightly worse than unconcealed faces.</p>
<p>These results demonstrate just how robust familiar face recognition can be. And our task only involved comparisons of still images of faces. It’s possible that in the real world, information from the <a href="https://journals.sagepub.com/doi/abs/10.1177/0956797613492986?casa_token=VAPKa62kkjYAAAAA%253AESo_ejfvaYNwpXeSw69FLDAVif_F26PFAn_oC84118wPk4FVjLPNhyGTGfmn0NLLVJTbBiGLgX8&journalCode=pssa">body or gait</a> or from clothing may supplement the reduced information from the masked face, increasing accuracy further. </p>
<p>For unfamiliar faces, both masks and sunglasses reduced recognition accuracy further. Face masks reduced performance the most, but only a little more than sunglasses. But with or without masks and sunglasses, recognising unfamiliar faces generally tends to be difficult and <a href="https://link.springer.com/article/10.3758/BRM.42.1.286">error-prone</a>.</p>
<p>Still, some people are highly adept at this task. <a href="https://pure.hud.ac.uk/en/publications/what-is-a-super-recogniser">Super-recognisers</a> – people who excel at <a href="https://theconversation.com/facial-recognition-research-reveals-new-abilities-of-super-recognisers-128414">recognising faces</a> – were also recruited to complete the tasks by Professor Josh Davis from the University of Greenwich <a href="https://www.superrecognisers.com/">Face and Voice Recognition Lab</a> database. Super-recognisers were also impaired by masks, <a href="https://royalsocietypublishing.org/doi/full/10.1098/rsos.201169">but they performed far better than regular people</a> in all concealment conditions.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/facial-recognition-research-reveals-new-abilities-of-super-recognisers-128414">Facial recognition: research reveals new abilities of 'super-recognisers'</a>
</strong>
</em>
</p>
<hr>
<p>Given that familiar face identification ability was barely impaired when faces were masked, why is it that humans recognise familiar faces so well? Humans may be born with an <a href="https://www.jstor.org/stable/24937459?seq=1">innate preference</a> for face-like stimuli. We are so attuned to seeking out faces in our environment that we often pick out face-like patterns within objects or clouds – a phenomenon known as “<a href="https://theconversation.com/holy-grilled-cheese-sandwich-what-is-pareidolia-14170">face pareidolia</a>”. </p>
<figure class="align-center ">
<img alt="A cloud shaped like the profile of a face in a blue sky" src="https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/398107/original/file-20210430-18-5m48zt.jpeg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Our tendency to see faces in clouds and other objects shows how we’re programmed for recognition.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/clouds-shaped-like-human-faces-blue-1770647444">neenawat khenyothaa/Shutterstock</a></span>
</figcaption>
</figure>
<p>It has been suggested that face processing is adaptive – that our ancestors had an <a href="https://royalsocietypublishing.org/doi/full/10.1098/rstb.2010.0358#d3e2561">evolutionary advantage</a> if they could tell the difference between a friend and a foe, which would help them decide who to approach and who to avoid. </p>
<p>The ability to recognise familiar faces is attributed to <a href="https://journals.sagepub.com/doi/full/10.1080/17470218.2015.1136656?casa_token=t0KRo2EH5KcAAAAA%253A-pXWTPVutRsZ-ws9Zaur3Pixr82UbWmOzT9o5A0csBROqmIIzcnfCYsz_xSHR0QJngJ6TPpOs90">learning the different ways</a> that the same face can look over different encounters, and <a href="https://www.sciencedirect.com/science/article/pii/S0010027721000305?casa_token=96jwGsrw6QkAAAAA:wcZSCKn7olxYDiU5w1Q_2BRPP1Q7jTI0gXZ8h3j4E8sF64J-Ej0sTFfnVx5lb1ytZTbUSoWr">learning how the face differs to other known faces</a>. This makes unfamiliar face identification far <a href="https://www.sciencedirect.com/science/article/pii/S0010027711002022?casa_token=O9_vC1Hp1yQAAAAA:_hunFNp2knB8Ln3kkXqAXhtboxZ6iO0zsJ3VVipdv8jj749VlFzr3CcNwbO6Mxiw64-L_GGU">more challenging</a>, as these factors are unknown for a face that we have little experience with. For unfamiliar faces, we don’t know how a face varies across changes in pose, expression, lighting, or age – or how the face differs from other unknown faces.</p>
<h2>Expert recognisers</h2>
<p>How can this explain our surprisingly adept performance for identifying faces covered by masks? For familiar faces, we likely have enough experience with the face that we can make an identification based on the limited available information. We may have seen the face obscured before, or our representation of the full face is so strong that we can handle the reduction in visible features. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/face-masks-why-your-eyes-might-be-saying-more-than-you-realise-145076">Face masks: why your eyes might be saying more than you realise</a>
</strong>
</em>
</p>
<hr>
<p>In contrast, for unfamiliar faces, we cannot rely on experience with the face. Super-recognisers are the anomalies here and, while it’s unclear why they’re so good at identifying faces, there is evidence that face recognition abilities may be <a href="https://www.pnas.org/content/107/11/5238.short">genetic</a>.</p>
<p>There are currently 7.4 billion faces on the planet. Though we’ll only ever encounter a tiny fraction of them, our ability to remember and recognise familiar faces is an evolutionary skill hundreds of thousands of years in the making. Our research shows it’s a skill barely affected when the faces in question are hidden by a mask.</p><img src="https://counter.theconversation.com/content/159667/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Even though they cover most of our features, face masks are no match for our highly-evolved capacity to recognise friends.Eilidh Noyes, Senior Lecturer in Cognitive Psychology, University of HuddersfieldKatie Gray, Associate Professor, School of Psychology and Clinical Language Sciences, University of ReadingKay Ritchie, Senior Lecturer in Cognitive Psychology, University of LincolnLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1588092021-04-15T14:18:30Z2021-04-15T14:18:30ZAI is increasingly being used to identify emotions – here’s what’s at stake<figure><img src="https://images.theconversation.com/files/394516/original/file-20210412-23-qjo8dg.jpg?ixlib=rb-1.1.0&rect=0%2C0%2C4089%2C2152&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">AI can be biased.</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/emotion-detected-by-artificial-intelligence-ai-1898196328">aslysun/Shutterstock</a></span></figcaption></figure><p>Imagine you are in a job interview. As you answer the recruiter’s questions, an artificial intelligence (AI) system <a href="https://edition.cnn.com/2020/01/15/tech/ai-job-interview/index.html">scans your face</a>, scoring you for <a href="https://www.fastcompany.com/3064863/watch-this-ai-platform-assess-trumps-and-clintons-emotional-intelligence">nervousness</a>, <a href="https://edition.cnn.com/2020/01/15/tech/ai-job-interview/index.html">empathy</a> and <a href="https://www.wired.com/story/job-screening-service-halts-facial-analysis-applicants/">dependability</a>. It may sound like science fiction, but these systems are increasingly used, <a href="https://www.technologyreview.com/2021/02/05/1017388/ai-deep-learning-facial-recognition-data-history/">often without people’s knowledge</a> or consent. </p>
<p>Emotion recognition technology (ERT) is in fact a burgeoning <a href="https://www.marketresearchengine.com/reportdetails/emotion-detection-and-recognition-edr-market">multi-billion-dollar industry</a> that aims to use AI to detect emotions from facial expressions. Yet the science behind emotion recognition systems <a href="https://www.nature.com/articles/d41586-021-00868-5">is controversial</a>: there are biases built into the systems.</p>
<p>Many companies use ERT <a href="https://www.bbc.com/news/business-44799239">to test customer reactions</a> to their products, from cereal to video games. But it can also be used in situations with much <a href="https://www.wired.com/story/amazon-detect-fear-face-you-scared/">higher stakes</a>, such as in <a href="https://www.washingtonpost.com/business/2019/07/31/emotion-detection-ai-is-billion-industry-new-research-says-it-cant-do-what-it-claims/">hiring</a>, by <a href="https://www.washingtonpost.com/business/2019/07/31/emotion-detection-ai-is-billion-industry-new-research-says-it-cant-do-what-it-claims/">airport security</a> to flag faces as revealing deception or fear, in <a href="https://theconversation.com/ai-facial-analysis-is-scientifically-questionable-should-we-be-using-it-for-border-control-155474">border control</a>, in <a href="https://www.article19.org/wp-content/uploads/2021/01/ER-Tech-China-Report.pdf">policing</a> to identify “dangerous people” or in <a href="https://edition.cnn.com/2021/02/16/tech/emotion-recognition-ai-education-spc-intl-hnk/index.html">education</a> to monitor students’ engagement with their homework.</p>
<h2>Shaky scientific ground</h2>
<p>Fortunately, facial recognition technology is receiving public attention. The award-winning film Coded Bias, recently released on Netflix, documents the discovery that many facial recognition technologies do not accurately detect darker-skinned faces. And the research team managing ImageNet, one of the largest and most important datasets used to train facial recognition, was recently forced to blur 1.5 million images <a href="https://www.wired.com/story/researchers-blur-faces-launched-thousand-algorithms/">in response to privacy concerns</a>. </p>
<p>Revelations about algorithmic bias and discriminatory datasets in facial recognition technology have led large technology companies, including Microsoft, Amazon and IBM, to halt sales. And the technology <a href="https://www.newscientist.com/article/2251508-is-police-use-of-face-recognition-now-illegal-in-the-uk/">faces legal challenges</a> regarding its use in policing in the UK. In the EU, a coalition of more than 40 civil society organisations have <a href="https://reclaimyourface.eu/">called for a ban</a> on facial recognition technology entirely. </p>
<p>Like other forms of facial recognition, ERT raises questions about bias, privacy and mass surveillance. But ERT raises another concern: the science of emotion behind it is controversial. Most ERT is based on the <a href="https://onlinelibrary.wiley.com/doi/10.1002/0470013494.ch3">theory of “basic emotions”</a> which holds that emotions are biologically hard-wired and expressed in the same way by people everywhere. </p>
<p>This is increasingly being challenged, however. Research in anthropology shows that emotions <a href="https://journal.culanth.org/index.php/ca/article/view/ca32.2.01">are expressed differently</a> across cultures and societies. In 2019, the <a href="https://www.psychologicalscience.org">Association for Psychological Science</a> conducted a review of the evidence, concluding that <a href="https://journals.sagepub.com/doi/10.1177/1529100619832930">there is no scientific support</a> for the common assumption that a person’s emotional state can be readily inferred from their facial movements. In short, ERT is built on shaky scientific ground.</p>
<p>Also, like other forms of facial recognition technology, ERT is encoded with racial bias. A study <a href="https://theconversation.com/emotion-reading-tech-fails-the-racial-bias-test-108404">has shown</a> that systems consistently read black people’s faces as angrier than white people’s faces, regardless of the person’s expression. Although <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3281765">the study of racial bias</a> in ERT is small, racial bias in other forms of facial recognition is well-documented. </p>
<p>There are two ways that this technology can hurt people, says AI researcher Deborah Raji <a href="https://www.technologyreview.com/2020/06/12/1003482/amazon-stopped-selling-police-face-recognition-fight/">in an interview with MIT Technology Review</a>: “One way is by not working: by virtue of having higher error rates for people of color, it puts them at greater risk. The second situation is when it does work — where you have the perfect facial recognition system, but it’s easily weaponized against communities to harass them.” </p>
<p>So even if facial recognition technology can be de-biased and accurate for all people, it still may not be fair or just. We see these <a href="http://www.cs.yale.edu/homes/jf/BarocasSelbst.pdf">disparate effects</a> when facial recognition technology is used in policing and judicial systems that are already discriminatory and harmful to people of colour. Technologies can be dangerous when they don’t work as they should. And they can also be dangerous when they work perfectly in an imperfect world. </p>
<p>The challenges raised by facial recognition technologies – including ERT – do not have easy or clear answers. Solving the problems presented by ERT requires moving from AI ethics centred on abstract principles to AI ethics centred on <a href="https://blogs.lse.ac.uk/medialse/2020/02/13/beginning-just-ai-bravery-and-creativity-for-ethics-in-practice/">practice</a> and <a href="https://www.technologyreview.com/2021/02/26/1020007/what-is-an-algorithm/">effects</a> on people’s lives. </p>
<figure class="align-center ">
<img alt="Image of a man looking into a phone." src="https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/395223/original/file-20210415-21-f9k6nq.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">AI can be racist.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/businessman-bank-telephone-protection-has-facial-719919238">HQuality/Shutterstock</a></span>
</figcaption>
</figure>
<p>When it comes to ERT, we need to collectively examine the controversial science of emotion built into these systems and analyse their potential for racial bias. And we need to ask ourselves: even if ERT could be engineered to accurately read everyone’s inner feelings, do we want such intimate surveillance in our lives? These are questions that require everyone’s deliberation, input and action. </p>
<h2>Citizen science project</h2>
<p>ERT has the potential to affect the lives of millions of people, yet there has been <a href="https://ainowinstitute.org/AI_Now_2019_Report.pdf">little public deliberation</a> about how – and if – <a href="https://journals.sagepub.com/doi/full/10.1177/2053951720904386">it should be used</a>. This is why we have developed <a href="https://emojify.info">a citizen science project</a>.</p>
<p>On our <a href="https://emojify.info">interactive website</a> (which works best on a laptop, not a phone) you can try out a private and secure ERT for yourself, to see how it scans your face and interprets your emotions. You can also play games comparing human versus AI skills in emotion recognition and learn about the controversial science of emotion behind ERT.</p>
<p>Most importantly, you can contribute your perspectives and ideas to generate new knowledge about the potential impacts of ERT. As the computer scientist and digital activist <a href="https://www.media.mit.edu/people/joyab/overview/">Joy Buolamwini</a> <a href="https://twitter.com/jovialjoy/status/1373637633591623680">says</a>: “If you have a face, you have a place in the conversation.”</p><img src="https://counter.theconversation.com/content/158809/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Alexa Hagerty received funding from a Nesta Collective Intelligence grant for this research. She is a member of the Just AI network mentioned in the article and co-founder of Dovetail Labs, a public benefit research & design collaborative. </span></em></p><p class="fine-print"><em><span>Alexandra Albert received funding from a Nesta Collective Intelligence grant for this research. She is a member of Dovetail Labs, a public benefit research & design collaborative She is also a member of the Extreme Citizen Science research group at University College London.</span></em></p>Emotion recognition technology raises questions about bias, privacy and mass surveillance.Alexa Hagerty, Research Associate of Anthropology, University of CambridgeAlexandra Albert, Research Fellow in Citizen Social Science, UCLLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1568172021-03-12T12:24:57Z2021-03-12T12:24:57ZSuper-recognisers: some people excel at both face and voice recognition<figure><img src="https://images.theconversation.com/files/389240/original/file-20210312-21-14octoc.png?ixlib=rb-1.1.0&rect=49%2C16%2C2107%2C1419&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Can you recognise voices?</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-vector/vector-illustration-flat-style-businessmen-discuss-1079756876">Shutterstock/Viktoria Kurpas</a></span></figcaption></figure><p>How good do you think you are at recognising faces? What about voices? Some people are great at it, while others struggle. Researchers have found a large variation in people’s abilities to recognise the faces or voices of those completely unknown to them. </p>
<p>When it comes to face recognition, the ability ranges from those who struggle to recognise the faces of friends and family – a condition known as “prosopagnosia” <a href="https://theconversation.com/can-you-teach-recognition-to-the-face-blind-67211">or face-blindness</a> – to those who have demonstrated exceptional unfamiliar face recognition skills, <a href="https://theconversation.com/facial-recognition-research-reveals-new-abilities-of-super-recognisers-128414">called “super-recognisers”</a>. These super-recognisers have been successfully deployed in many agencies, including <a href="https://thepsychologist.bps.org.uk/volume-26/edition-10/i-never-forget-face">London’s police service</a>.</p>
<p>Similar findings have also been discovered for voice recognition. The condition “phonagnosia” describes those who struggle to recognise the voices of their friends and family. But whether or not someone can possess exceptionally good voice recognition ability remained uncovered, until now.</p>
<p><a href="https://onlinelibrary.wiley.com/doi/10.1002/acp.3813">In a new study</a>, we and our colleagues tested whether super-face-recognisers could transfer their abilities across to recognising voices, to explore the possibility of a super-voice-recogniser.</p>
<h2>Face and voice abilities</h2>
<p>Typically, to assess the ability levels of super-recognisers, researchers have often used two tests. First, the <a href="https://link.springer.com/article/10.3758/PBR.16.2.252">Cambridge face memory test</a> measures the ability to learn and remember a face. Then the <a href="https://link.springer.com/article/10.3758/BRM.42.1.286">Glasgow face matching test</a> is used to measure the ability to say whether two faces belong to the same person or two different people. </p>
<p>However, being good at recognising a face doesn’t necessarily mean someone is also good at face matching. <a href="https://onlinelibrary.wiley.com/doi/full/10.1002/acp.3260?casa_token=rGNeDOCD4BEAAAAA%3AMi8vP9BOhuKVgHMPMX6YW8k9X7Ivzg7waP9H52jtde_F8QnYnHCjuIFCdQfNMT2TB3-nB0r5U160nhs">Research has shown</a> even super-recognisers can be very good at face memory, but just as good as typical ability participants on face matching or vice versa. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/facial-recognition-research-reveals-new-abilities-of-super-recognisers-128414">Facial recognition: research reveals new abilities of 'super-recognisers'</a>
</strong>
</em>
</p>
<hr>
<p>Voice tests have also been designed, not to measure super-recognition skills, but rather to measure the general ability <a href="https://link.springer.com/article/10.3758/s13428-015-0689-6">to remember a voice </a>, and to decide whether two voices belong to the <a href="https://link.springer.com/article/10.3758/s13428-017-0985-4">same person or two different people</a>. But the extent to which super-recognisers can perform well on voice tests was not yet examined.</p>
<h2>Our study</h2>
<p>In our recent study, we tested four groups of participants, based on face memory and face-matching abilities. The participants completed a test of voice memory, a test of voice matching and a test for identifying celebrity voices. The study had a number of findings. </p>
<p>First, we found voice recognition ability varies substantially beyond the definitions found in current literature, which describes people falling into two categories, either “typical” or phonagnosic. We found people can perform very well at voice recognition, beyond the typical range abilities.</p>
<p>Second, we found those who possessed exceptional face memory skills, face-matching skills, or both, outperformed those with typical ability skills at voice memory and voice matching. </p>
<figure class="align-center ">
<img alt="A woman smiling with a green box drawn around her face." src="https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/389243/original/file-20210312-21-145hbsy.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Facial recognition.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/facial-recognition-system-concept-680761543">Shutterstock/metamorworks</a></span>
</figcaption>
</figure>
<p>Some participants managed to achieve consistently high scores across multiple tests. This hints at the possibility of a super-voice-recogniser. However further research is needed to properly substantiate this possibility. </p>
<p>Partly this is because the voice tests used were never originally designed to distinguish between the exceptional and the very good, so perhaps are unable to fully explore superior voice processing. As such, new voice tests specifically designed to focus on the upper end of the voice-recognition ability spectrum are required.</p>
<p>Our research provides tentative support to the idea that there may be some link between different mechanisms in the brain. These could be cross-modality (voices and faces) and cross-task (memory and perception) mechanisms that, working together, drive this kind of superior ability to recognise voices and faces.</p>
<p>It may be the case that these individuals can be useful in similar job positions that super-face-recognisers have also been employed in – such as police and security forces. This is particularly relevant when voice clips are the only available evidence, such as in cases of phone surveillance, kidnapping, fraud, blackmail and counter-terrorism operations. </p>
<p>Our work is the first to explore the potential abilities of super-voice-recognisers and ask whether those that possess exceptional face memory abilities, face-matching abilities or both can transfer their skills across to voice tests. </p>
<p>It also provides the first piece of work to suggest individuals with excellent voice-recognition abilities may be able to enhance policing and security operations. Screening for those with such an ability might be a useful tool during recruitment stages of these types of professions.</p><img src="https://counter.theconversation.com/content/156817/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>The people who are excellent at recognising faces might also excel at recognising voices.Ryan Jenkins, PhD Candidate, Psychology, University of GreenwichDavid James Robertson, Lecturer in Psychology, University of Strathclyde Josh P Davis, Reader, Applied Psychology, University of GreenwichLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1500892020-11-16T18:56:08Z2020-11-16T18:56:08ZAre you among Australia’s best facial super-recognisers? Take our test to find out<figure><img src="https://images.theconversation.com/files/369487/original/file-20201116-21-1483hyu.jpg?ixlib=rb-1.1.0&rect=47%2C124%2C3902%2C2836&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">
</span> <span class="attribution"><span class="source">Shutterstock</span></span></figcaption></figure><p>At just 11 years old, Nicole couldn’t understand why her classmates were struggling. The competition was straightforward: recognise teachers at the primary school from their baby photos. Most students struggled to recognise more than a few faces. Nicole easily named all 20 faces correctly.</p>
<p>It wasn’t until completing our free, online test — the UNSW Face Test — that she finally understood why she could do what her classmates couldn’t. Her results were clear: Nicole is one of the rare individuals who can call themselves a “super-recogniser”.</p>
<p>Most of us easily recognise the faces of people we know — our family, friends and colleagues — but recognising less familiar people can be surprisingly challenging. For instance, you might not notice the cashier you smiled at in the supermarket takes the same daily bus route as you. </p>
<p>That is, unless you’re a super-recogniser. In our study, <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0241747">published today</a> in PLOS One, we release the results of the first 25,000 people to complete the UNSW Face Test. It’s currently the most challenging test available for identifying super-recognisers. You can take it <a href="http://www.unswfacetest.com/">here</a>.</p>
<h2>What is a super-recogniser?</h2>
<p>A super-recogniser is someone who is exceptionally gifted at facial recognition. Super-recognisers remember faces much more accurately than the average person and often after many years, or very short encounters.</p>
<p><a href="https://www.pnas.org/content/107/11/5238">Research suggests super-recognition is genetic</a>, meaning these skills are coded in our DNA. However, researchers have yet to find any other abilities shared by super-recognisers. This is because facial recognition ability is independent from <a href="https://psycnet.apa.org/record/2010-15679-001">intelligence</a>, <a href="https://cognitiveresearchjournal.springeropen.com/articles/10.1186/s41235-018-0112-9">personality</a> and <a href="https://cognitiveresearchjournal.springeropen.com/articles/10.1186/s41235-018-0112-9">other cognitive and perceptual skills</a>. </p>
<h2>Searching for the Einstein of facial recognition</h2>
<p>Currently, the only way to identify a super-recogniser is by giving them facial recognition tests. There is no Olympic-level contest or world record for this. If there were, however, the UNSW Face Test could be the arena in which super-recognisers compete. </p>
<p>It’s the first online test designed to really challenge super-recognisers. It takes around 20 minutes to complete and features images of 80 different people. Our test is difficult because it requires people to correctly identify a face despite substantial changes in appearance from one encounter to the next. </p>
<p>This could include differences in the pictured person’s age, pose, lighting and expression. This is why the average person scores around 60%, which is lower than the average in other facial recognition tests. </p>
<p>Super-recognisers will typically score 70% or higher. And while this is a good score, it’s certainly far from perfect. In other tests it’s not uncommon for super-recognisers to <a href="https://onlinelibrary.wiley.com/doi/full/10.1002/acp.3608">score perfectly</a>. But this has yet to be achieved with the UNSW Face Test. </p>
<p>Now with more than 31,000 test results available, the highest anyone has scored is 97%, and only 11 people have scored 90% or higher.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/super-recognisers-accurately-pick-out-a-face-in-a-crowd-but-can-this-skill-be-taught-112003">Super-recognisers accurately pick out a face in a crowd – but can this skill be taught?</a>
</strong>
</em>
</p>
<hr>
<p>We’ve learned from analysing our test results that, although all super-recognisers are exceptional at facial recognition, some are better than others. This leaves room at the top for a better kind of super-recogniser to emerge: a super-<em>duper</em>-recogniser — the highest achievers among an already extraordinary group.</p>
<p>With more people completing the test, we hope to find the Einstein of facial recognition.</p>
<h2>Why are super-recognisers important?</h2>
<p>Many important tasks rely on recognising or matching images of unfamiliar faces. Examples include matching a traveller to their passport, or a CCTV image to a police mugshot. Despite advancements, facial recognition technology is <a href="https://theconversation.com/the-trouble-with-facial-recognition-technology-in-the-real-world-69685">often unable to execute</a> such tasks with complete accuracy.</p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=405&fit=crop&dpr=1 600w, https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=405&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=405&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=509&fit=crop&dpr=1 754w, https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=509&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/369526/original/file-20201116-21-15v0x11.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=509&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Passport officers have to decide if the person in front of them is the same as the person pictured on an identity document.</span>
<span class="attribution"><span class="source">Author provided</span></span>
</figcaption>
</figure>
<p>Through collaborations with the Australian Passport Office, we now know experience and <a href="https://theconversation.com/super-recognisers-accurately-pick-out-a-face-in-a-crowd-but-can-this-skill-be-taught-112003">training</a> alone may not reduce error rates in passport officers tasked with detecting fraudulent passport applications.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/passport-staff-miss-one-in-seven-fake-id-checks-30606">Passport staff miss one in seven fake ID checks</a>
</strong>
</em>
</p>
<hr>
<p>Instead, employing and assigning super-recognisers to such roles that would benefit from their skills is a promising solution. </p>
<p>In Australia, many police forces and government agencies, including <a href="https://www.facebook.com/watch/?v=2021695261472644">Queensland Police</a> and the <a href="https://newsroom.unsw.edu.au/news/science-tech/passport-officers-using-face-recognition-technology-better-detecting-fraud#:%7E:text=Passport%20officers%20using%20face%20recognition%20technology%20better%20at%20detecting%20fraud,-Twitter%20Facebook%20LinkedIn&text=Face%2Dmatching%20experts%20using%20automatic,detecting%20fraud%2C%20new%20research%20shows.">Australian Passport Office</a>, are now selecting people for facial recognition roles on the basis of their facial recognition ability. In the UK, London’s <a href="https://www.bbc.com/future/article/20150611-the-superpower-police-now-use-to-tackle-crime">Metropolitan Police</a> has done the same.</p>
<h2>Advancing knowledge on cognitive processes</h2>
<p>Uncovering what makes super-recognisers different to the average person is also of fundamental scientific interest. </p>
<p>Understanding why there’s so much variation in people’s ability to recognise faces could shed light on the cognitive processes and mechanisms that let certain people to become experts in visual tasks more generally. Currently, we know very little about these factors. </p>
<p><a href="https://journals.sagepub.com/doi/10.1080/17470218.2016.1161059">Some research</a> has suggested super-recognisers may process faces more “holistically” than other people, <a href="https://www.sciencedirect.com/science/article/abs/pii/S0010945216301186">combining individual facial features</a> into an overall picture. But very few studies have aimed to understand the underlying causes for this, so current knowledge is still evolving. </p>
<p>Our long-term research aim is to understand the underlying cognitive and perceptual processes that give rise to super-recognisers’ impressive abilities. But first, we have to find them. So if you think you might be one, take the <a href="http://www.unswfacetest.com/">UNSW Face Test</a> to help us understand your superpowers.</p><img src="https://counter.theconversation.com/content/150089/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>David White receives funding from the Australian Research Council and the Australian Passport Office. </span></em></p><p class="fine-print"><em><span>Alice Towler and James Dunn do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Psychologists are hoping a new, extra-difficult facial recognition test will help unearth more of Australia’s top performers in facial recognition — known as ‘super-recognisers’.James Dunn, Postdoctoral Research Associate, UNSW SydneyAlice Towler, Post-doctoral Research Fellow, UNSW SydneyDavid White, Scientia Fellow, UNSW SydneyLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1444172020-09-06T20:13:11Z2020-09-06T20:13:11ZFace masks and facial recognition will both be common in the future. How will they co-exist?<figure><img src="https://images.theconversation.com/files/356521/original/file-20200904-20-1h9t8oj.jpg?ixlib=rb-1.1.0&rect=22%2C14%2C1894%2C833&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">
</span> <span class="attribution"><span class="source">Pixabay</span>, <a class="license" href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA</a></span></figcaption></figure><p>It’s surprising how quickly public opinion can change. Winding the clocks back 12 months, many of us would have looked at a masked individual in public with suspicion. </p>
<p>Now, some countries have enshrined face mask use <a href="https://www.aljazeera.com/news/2020/04/countries-wearing-face-masks-compulsory-200423094510867.html">in law</a>. They’ve also been made <a href="https://www.health.gov.au/news/health-alerts/novel-coronavirus-2019-ncov-health-alert/how-to-protect-yourself-and-others-from-coronavirus-covid-19/masks">compulsory in Victoria</a> and are recommended in several other states.</p>
<p>One consequence of this is that facial recognition systems in place for security and crime prevention may no longer be able to fulfil their purpose. In Australia, most agencies are silent about the use of facial recognition. </p>
<p>But documents leaked earlier this year revealed <a href="https://www.buzzfeednews.com/article/ryanmac/clearview-ai-fbi-ice-global-law-enforcement">Australian Federal Police</a> and state police in <a href="https://theconversation.com/australian-police-are-using-the-clearview-ai-facial-recognition-system-with-no-accountability-132667">Queensland, Victoria and South Australia</a> all use Clearview AI, a commercial facial recognition platform. New South Wales police <a href="https://www.abc.net.au/news/2020-01-23/australian-founder-of-clearview-facial-recognition-interview/11887112">also admitted</a> using a biometrics tool called PhotoTrac.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/your-face-is-part-of-australias-national-security-weapon-should-you-be-concerned-47364">Your face is part of Australia's 'national security weapon': should you be concerned?</a>
</strong>
</em>
</p>
<hr>
<h2>What is facial recognition?</h2>
<p><a href="https://us.norton.com/internetsecurity-iot-how-facial-recognition-software-works.html">Facial recognition</a> involves using computing to identify human faces in images or videos, and then measuring specific facial characteristics. This can include the distance between eyes, and the relative positions of the nose, chin and mouth. </p>
<p>This information is combined to create a <a href="https://www.eff.org/pages/face-recognition">facial signature, or profile</a>. When used for individual recognition – such as to unlock your phone – an image from the camera is compared to a recorded profile. This process of facial “verification” is relatively simple.</p>
<p>However, when facial recognition is used to identify faces in a crowd, it requires a significant database of profiles against which to compare the main image. </p>
<p>These profiles can be legally collected by enrolling large numbers of users <a href="https://immi.homeaffairs.gov.au/help-support/meeting-our-requirements/biometrics">into systems</a>. But they’re sometimes collected through <a href="https://news.miami.edu/stories/2020/02/new-facial-recognition-software-scrapes-inventory-from-social-media.html">covert means</a>.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=500&fit=crop&dpr=1 600w, https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=500&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=500&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=628&fit=crop&dpr=1 754w, https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=628&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/353221/original/file-20200817-24-10qf4oa.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=628&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Facial ‘verification’ (the method used to unlock smartphones) compares the main image with a single pre-saved facial signature. Facial ‘identification’ requires examining the image against an entire database of facial signatures.</span>
<span class="attribution"><span class="source">teguhjatipras/pixabay</span></span>
</figcaption>
</figure>
<h2>The problem with face masks</h2>
<p>As facial signatures are based on mathematical models of the relative positions of facial features, anything that reduced the visibility of key characteristics (such as the nose, mouth and chin) interferes with facial recognition.</p>
<p>There are already many ways to <a href="https://www.businessinsider.com.au/clothes-accessories-that-outsmart-facial-recognition-tech-2019-10">evade or interfere</a> with facial recognition technologies. Some of these evolved from techniques designed to evade number plate recognition systems.</p>
<p>Although the coronavirus pandemic has escalated concerns around the evasion of facial recognition systems, <a href="https://theintercept.com/2020/07/16/face-masks-facial-recognition-dhs-blueleaks/">leaked US documents</a> show these <a href="https://www.documentcloud.org/documents/6989376-U-FOUO-in-Violent-Adversaries-Likely-to-Use.html">discussions</a> taking place back in 2018 and 2019, too. </p>
<figure>
<iframe width="440" height="260" src="https://www.youtube.com/embed/Xpu2MSmZkmU?wmode=transparent&start=0" frameborder="0" allowfullscreen=""></iframe>
<figcaption><span class="caption">This clip shows how fashion designers are outsmarting facial recognition surveillance / YouTube.</span></figcaption>
</figure>
<p>And while the debate on the <a href="https://theconversation.com/large-scale-facial-recognition-is-incompatible-with-a-free-society-126282">use</a> and <a href="https://theconversation.com/facial-recognition-technology-is-expanding-rapidly-across-australia-are-our-laws-keeping-pace-141357">legality</a> of facial recognition continues, the focus has recently shifted to the challenges presented by mask-wearing in public.</p>
<p>On this front, the US National Institute of Standards and Technology (NIST) coordinated a <a href="https://www.nist.gov/news-events/news/2020/07/nist-launches-studies-masks-effect-face-recognition-software">major research project</a> to evaluate how masks impacted the performance of various facial recognition systems used across the globe.</p>
<p>Its <a href="https://nvlpubs.nist.gov/nistpubs/ir/2020/NIST.IR.8311.pdf">report</a>, published in July, found some algorithms struggled to correctly identify mask-wearing individuals <a href="https://www.vox.com/recode/2020/7/28/21340674/face-masks-facial-recognition-surveillance-nist">up to 50% of the time</a>. This was a significant error rate compared to when the same algorithms analysed unmasked faces.</p>
<p>Some algorithms even <a href="https://edition.cnn.com/2020/07/28/tech/face-masks-facial-recognition/index.html">struggled to locate a face</a> when a mask was covering too much of it. </p>
<h2>Finding ways around the problem</h2>
<p>There are currently no usable photo data sets of mask-wearing people that can be used to train and evaluate facial recognition systems. </p>
<p>The NIST study addressed this problem by <a href="https://www.theverge.com/2020/7/28/21344751/facial-recognition-face-masks-accuracy-nist-study">superimposing</a> masks (of various colours, sizes and positions) over images of faces, as seen here:</p>
<p><div data-react-class="Tweet" data-react-props="{"tweetId":"1287876259586822144"}"></div></p>
<p>While this may not be a realistic portrayal of a person wearing a mask, it’s effective enough to study the effects of mask-wearing on facial recognition systems. </p>
<p>It’s possible images of real masked people would allow more details to be extracted to improve recognition systems – perhaps by estimating the nose’s position based on visible protrusions in the mask.</p>
<p>Many facial recognition technology vendors are already <a href="https://www.cnet.com/health/facial-recognition-firms-are-scrambling-to-see-around-face-masks/">preparing for</a> a future where mask use will continue, or even increase. <a href="https://www.dezeen.com/2020/02/27/face-recognition-masks-resting-risk-face/">One US company</a> offers masks with customers’ faces printed on them, so they can unlock their smartphones without having to remove it. </p>
<h2>Growing incentives for wearing masks</h2>
<p>Even <a href="https://qz.com/299003/a-quick-history-of-why-asians-wear-surgical-masks-in-public/">before the coronavirus pandemic</a>, masks were a common defence against air pollution and viral infection in countries including China and Japan. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/ive-always-wondered-why-many-people-in-asian-countries-wear-masks-and-whether-they-work-90178">I've always wondered: why many people in Asian countries wear masks, and whether they work</a>
</strong>
</em>
</p>
<hr>
<p>Political <a href="https://www.bbc.com/news/world-asia-china-49939173">activists</a> also wear masks to evade detection on the streets. Both the <a href="https://www.nytimes.com/2019/07/26/technology/hong-kong-protests-facial-recognition-surveillance.html">Hong Kong</a> and <a href="https://www.washingtonpost.com/technology/2020/06/12/facial-recognition-ban/">Black Lives Matter</a> protests have reinforced protesters’ desire to dodge facial recognition by <a href="https://www.theverge.com/2020/7/17/21328287/face-masks-facial-recognition-privacy-security-protests">authorities and government agencies</a>. </p>
<p>As experts forecast a future with more <a href="https://theconversation.com/this-isnt-the-first-global-pandemic-and-it-wont-be-the-last-heres-what-weve-learned-from-4-others-throughout-history-136231">pandemics</a>, <a href="https://www.theguardian.com/environment/2016/may/12/air-pollution-rising-at-an-alarming-rate-in-worlds-cities">rising levels</a> of <a href="https://www.who.int/health-topics/air-pollution#tab=tab_1">air pollution</a>, persisting <a href="https://freedomhouse.org/report/freedom-world/2020/leaderless-struggle-democracy">authoritarian regimes</a> and a projected <a href="https://theconversation.com/how-climate-change-is-increasing-the-risk-of-wildfires-99056">increase</a> in <a href="https://theconversation.com/climate-change-is-bringing-a-new-world-of-bushfires-123261">bushfires</a> producing dangerous smoke – it’s likely mask-wearing will become the norm for at least a proportion of us.</p>
<p>Facial recognition systems will need to adapt. Detection will be based on features that remain visible such as the eyes, eyebrows, hairline and general shape of the face. </p>
<p>Such technologies are already under development. Several suppliers are offering <a href="https://www.facewatch.co.uk/2020/05/11/facewatch-launches-facemask-recognition-upgrade/">upgrades</a> and <a href="https://www.prnewswire.com/news-releases/finally-a-biometric-solution-that-recognizes-users-wearing-face-masks-and-doesnt-require-touch-301069400.html">solutions</a> that claim to deliver reliable results with mask-wearing subjects.</p>
<p>For those who oppose the use of facial recognition and wish to go undetected, a plain mask may suffice for now. But in the future they might have to consider alternatives, such as a mask printed with a fake <a href="https://thispersondoesnotexist.com/">computer-generated face</a>.</p><img src="https://counter.theconversation.com/content/144417/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Paul Haskell-Dowland does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>With face masks now compulsory or recommended in various parts of the country, how are facial recognition systems functioning?Paul Haskell-Dowland, Associate Dean (Computing and Security), Edith Cowan UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1232052019-09-09T14:23:17Z2019-09-09T14:23:17ZIntrovert? You may just be bad at recognising faces<figure><img src="https://images.theconversation.com/files/291501/original/file-20190909-109962-xohbm.jpg?ixlib=rb-1.1.0&rect=3279%2C36%2C4943%2C2342&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">s</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/smiling-faces-happy-group-multiethnic-people-524577217">pathdoc/Shutterstock</a></span></figcaption></figure><p>Although most of us can distinguish between and remember hundreds of different faces, some people are better at it than others. “<a href="https://theconversation.com/how-an-army-of-super-recognisers-could-help-spot-criminals-and-missing-persons-67571">Super-recognisers</a>” can accurately identify faces even when they have only seen them briefly previously. At the other extreme, “<a href="https://theconversation.com/always-forget-a-face-so-does-brad-pitt-dont-just-blame-your-memory-50334">developmental prosopagnosics</a>” are significantly impaired at recognising faces in many everyday situations. </p>
<p>For the majority of us though, our face recognition ability falls between these extremes. But why are there such huge individual differences? How do these abilities affect us and where do they come from? Psychologists have started to investigate such questions, and found several answers. For example, we have discovered that it is linked to personality.</p>
<p>Face recognition differences may reflect processing or structural differences in the brain. For example, people with prosopagnosia <a href="https://www.ncbi.nlm.nih.gov/pubmed/28825896">may have reduced connectivity</a> between brain regions in the face processing network.</p>
<p>Another idea is that face recognition ability is related to other more general cognitive abilities, like memory or visual processing. Here, though, findings are mixed. Some research <a href="https://link.springer.com/article/10.3758/BF03193433">supports a link</a> between face recognition and specific abilities like visual processing. But other research <a href="https://psycnet.apa.org/record/2010-15679-001">has discounted this idea</a>.</p>
<p>Yet another possibility is that individual differences in face recognition reflect a person’s personality or their social and emotional functioning. Interestingly, face recognition ability <a href="https://psycnet.apa.org/record/2012-24435-027">has been linked</a> to measures of empathy and anxiety.</p>
<p>Empathy reflects a person’s ability to understand and share the feelings of another person. In 2010, researchers asked volunteers to try and remember the identity of a number of faces presented one at a time. They were later presented with the same faces mixed together with new faces and were asked to state whether each face was “old” (learnt) or “new”. The performance was measured by the number of learnt faces correctly identified as being familiar. The researchers found that those who rated themselves as high in empathy <a href="https://psycnet.apa.org/record/2012-24435-027">performed significantly better</a> at a face recognition memory task than those with low empathy skills.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/291503/original/file-20190909-109952-16b4enf.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Actor Brad Pitt struggles to recognise faces.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/cannes-france-may-21-2019-brad-1412672144?src=gihmorpJoTWxstHkXCYBwA-1-2">taniavolobueva/Shutterstock</a></span>
</figcaption>
</figure>
<p>Research has also found that people who report significantly <a href="https://psycnet.apa.org/record/1981-00476-001">lower levels of general anxiety</a> have better face recognition skills than those who are have higher anxiety. </p>
<p>Interestingly, more recent research has suggested the link between anxiety and face recognition ability may be <a href="https://kar.kent.ac.uk/32038/">more prominent for women</a>, and may be particularly <a href="https://www.ncbi.nlm.nih.gov/pubmed/22194916">related to anxiety in social situations</a> (social anxiety).</p>
<p>Situational anxiety may also play a role. For example, face recognition may be impaired when an eyewitness is asked to try and identify the face of a suspect viewed in a stressful situation. </p>
<h2>Personality</h2>
<p>In our own work we have considered the relationship between extroversion and individual face recognition ability. Extroverts are known to be superior at decoding social information and to be more involved in social activities than introverts. It may therefore be that extroverts are more skilled at recognising different identities. </p>
<p>In a previous study, researchers collected data from a group of 20 highly extroverted and 23 highly introverted volunteers (from an original sample of 339 volunteers). They found that extroverts <a href="https://www.ncbi.nlm.nih.gov/pubmed/20798810">performed significantly better</a> in a face recognition memory task compared with the introverts.</p>
<p>In our own work we looked at 100 volunteers <a href="https://journals.sagepub.com/doi/10.1080/17470218.2014.988737">with a range of extroversion levels</a>. The volunteers were shown famous faces and were asked to try and identify them by giving their name or some other identifying information. </p>
<p>Volunteers were also asked to say whether two unfamiliar faces belonged to the same person or different people – a task dubbed face matching. While there was no relationship between extroversion and face matching, there was a positive relationship between extroversion and famous face recognition. So to answer our own question, although there is individual variation, extroverts do tend to be better at recognising faces.</p>
<p>We do not yet understand the importance and reason for these findings, however. It may be that extroversion causes superior face recognition or that people who are better at identifying faces become more extroverted as a result. </p>
<p>If so, then a person’s inability to learn and recognise faces may lead them to become more introverted, to avoid potentially embarrassing social situations. Alternatively, introverted people may meet fewer people and therefore never develop good face recognition skills. </p>
<p>It may also work both ways. If you are slightly worse at recognising faces to start with you may end up meeting fewer people, and therefore becoming even worse at it over time. It could also be that both extroversion and face recognition are related to yet another factor that we still don’t know about.</p>
<p>In future work, we need to consider how our findings with extroversion fit together with research on empathy and anxiety. We also need to consider how much practical impact these issues have on face identification in applied situations – from identification by police officers to <a href="https://theconversation.com/how-an-army-of-super-recognisers-could-help-spot-criminals-and-missing-persons-67571">passport control</a>.</p>
<p>Our own ongoing work is looking at the impact of wider individual factors like altruism and optimism on face recognition. It may be that we soon find even more explanations for why some of us are just better at recognising faces than others.</p><img src="https://counter.theconversation.com/content/123205/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Dr Karen Lander has previously received funding from the ESRC, Leverhulme Trust, British Academy and Parkinson's UK. </span></em></p>Your ability to recognise faces may have some connection to your extroversion, empathy levels and anxiety.Karen Lander, Senior Lecturer in Experimental Psychology, University of ManchesterLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1189312019-06-18T12:16:23Z2019-06-18T12:16:23ZSurveillance cameras will soon be unrecognisable – time for an urgent public conversation<figure><img src="https://images.theconversation.com/files/279992/original/file-20190618-118505-1rkskn7.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Camera never lies.</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/businesswoman-on-blurred-background-using-futuristic-1087470950?src=8S5H3HpVCq7PWLDR8qK20Q-1-7&studio=1">sdecoret</a></span></figcaption></figure><p>It is often <a href="http://news.bbc.co.uk/1/hi/uk/8160757.stm">argued</a> that <a href="https://truepublica.org.uk/united-kingdom/britain-now-officially-the-most-surveilled-state-in-the-world/">the UK is the most surveilled</a> country on the planet. This may or may not have been the case in the past but there are <a href="https://www.disnetwork.co.uk/12-facts-cctv-surveillance-uk/">certainly now</a> millions of surveillance cameras in public spaces – not to mention private buildings and homes. Behind those lenses they are changing in ways that people are often barely aware of, with privacy implications that should be widely discussed as a matter of urgency. </p>
<p><a href="https://videosurveillance.blog.gov.uk/2019/03/21/the-debate-on-automatic-facial-recognition-continues/">Automatic face recognition</a> is currently the hot ticket in this industry, having been introduced in a number of cities around the world, in the <a href="https://www.wired.com/story/some-us-cities-moving-real-time-facial-surveillance/">US</a>, <a href="https://www.roboticsbusinessreview.com/ai/facial-recognition-cameras-5-countries/">China</a>, <a href="https://www.reuters.com/article/us-germany-security/german-police-test-facial-recognition-cameras-at-berlin-station-idUSKBN1AH4VR">Germany</a> and <a href="https://www.reuters.com/article/us-singapore-surveillance/singapore-to-test-facial-recognition-on-lampposts-stoking-privacy-fears-idUSKBN1HK0RV">Singapore</a>. The police argue that piloting such systems has allowed them to test the technology to help identify potential terrorists and other known offenders. Yet this has to be weighed against different concerns. The broadest is our expectation of privacy and anonymity in public places - and whether this is a step too far towards our every move being visible to the state. </p>
<p>Then there is the question of how well these face recognition systems work at present. Their success rate at recognising faces <a href="https://www.independent.co.uk/news/uk/home-news/met-police-facial-recognition-success-south-wales-trial-home-office-false-positive-a8345036.html">has been</a> shown to be as low as 2%. Linked to this is an inbuilt bias within the software that makes the technology <a href="https://www.nytimes.com/2018/02/09/technology/facial-recognition-race-artificial-intelligence.html">far less accurate</a> at <a href="https://www.libertyhumanrights.org.uk/resist-facial-recognition">identifying</a> darker skinned people and women. It therefore has the potential to exacerbate tensions between ethnic minorities and the police. </p>
<p>This could be compounded by another contentious issue, which is the police using so-called “watch list” databases of faces against which it is trying to match live images. Typically these databases include policing images of people taken in custody, who may never have been convicted of a crime and are unlikely to have consented to their data being used in this way. </p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/279995/original/file-20190618-118530-k8luh1.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Watching you not watching me.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/businesswoman-on-blurred-background-using-futuristic-1087470950?src=8S5H3HpVCq7PWLDR8qK20Q-1-7&studio=1">MY Stock</a></span>
</figcaption>
</figure>
<p>For these reasons the use of automatic face recognition software has been very controversial, and until the technology is more reliable we should probably be very cautious in how we use it. There have been two significant pilots in the UK in recent times, in the <a href="https://www.south-wales.police.uk/en/news-room/introduction-of-facial-recognition-into-south-wales-police/">south of Wales</a> and in <a href="https://www.met.police.uk/live-facial-recognition-trial/">London</a>. Both are the subject of judicial review actions, brought respectively by civil liberties organisations <a href="https://www.walesonline.co.uk/news/wales-news/facial-recognition-technology-south-wales-16305359">Liberty</a> and <a href="https://www.independent.co.uk/news/uk/home-news/facial-recognition-uk-police-london-inaccurate-legal-challenge-judicial-review-trials-stratford-a8463541.html">Big Brother Watch</a>, which are due to conclude in the coming months. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/setting-precedents-for-privacy-the-uk-legal-challenges-bringing-surveillance-into-the-open-117639">Setting precedents for privacy: the UK legal challenges bringing surveillance into the open</a>
</strong>
</em>
</p>
<hr>
<p>In the US, meanwhile, the city of San Francisco <a href="https://www.nytimes.com/2019/05/14/us/facial-recognition-ban-san-francisco.html">banned</a> the use of face recognition in its public systems in May. Other American cities are <a href="https://www.cnet.com/news/how-a-facial-recognition-ban-could-come-to-your-city-soon/">expected</a> to follow suit – with face recognition software currently <a href="https://www.wired.com/story/some-us-cities-moving-real-time-facial-surveillance/">being used</a> in the likes of Chicago, New York and Detroit. The technology has also <a href="https://www.thestar.com/news/gta/2019/05/28/toronto-police-chief-releases-report-on-use-of-facial-recognition-technology.html">generated much debate</a> in Canada, where it is in use in Toronto and some other cities. </p>
<h2>Tomorrow’s world</h2>
<p>Face recognition highlights bigger questions around which types of surveillance cameras and systems are acceptable to society. This question is complicated by the fact that surveillance cameras are becoming more sophisticated and computerised without necessarily looking much different. There is no signage or information that tells us about their enhanced capabilities, which means the activities behind them become less transparent. </p>
<p>As the technology has been miniaturised and costs have fallen, new types of cameras have emerged, including <a href="https://www.telegraph.co.uk/news/2018/11/30/traditional-police-notebook-gives-way-body-worn-video-police/">body-worn video devices</a>, <a href="https://www.pogo.org/analysis/2018/09/these-police-drones-are-watching-you/">drones</a>, <a href="https://www.driving.co.uk/news/features/revolutionary-dash-cam-portal-makes-sending-police-videos-dangerous-driving-doddle/">dash</a> and <a href="https://www.bbc.co.uk/news/uk-40920095">head cams</a>. At the same time, imaging and recording techniques have become more and more standardised. This has allowed for greater connectivity between systems and has raised quality to the point that images can be trustworthy evidence in legal proceedings. </p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=428&fit=crop&dpr=1 600w, https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=428&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=428&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=538&fit=crop&dpr=1 754w, https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=538&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/279996/original/file-20190618-118497-1cabm6u.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=538&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Body cams: already in use.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/london-uk-19th-april-2019-close-1397810531?src=BheoQpeNZ1VjYavPteecVA-1-26&studio=1">John Gomez</a></span>
</figcaption>
</figure>
<p>Besides face recognition, we are seeing the emergence of cameras <a href="https://nakedsecurity.sophos.com/2017/07/19/police-bodycams-get-tech-that-can-identify-faces-and-people/">capable of</a> object tracking and recognition, plus advances in <a href="https://www.bbc.co.uk/news/science-environment-20629671">noise</a> or <a href="https://thenextweb.com/the-next-police/2018/11/29/ai-scanner-detect-drugs-mail/">smell</a> analysis. Police forces in the US and UK <a href="https://www.newscientist.com/article/2186512-exclusive-uk-police-wants-ai-to-stop-violent-crime-before-it-happens/">have been</a> trialling systems that predict how likely individuals are to commit a crime. It is all a quantum leap away from the old CCTV cameras with which we are familiar. </p>
<p>Governance and regulation is having to evolve quickly to keep abreast of this environment. To this end, surveillance cameras in England and Wales are now regulated by the specialist office of the <a href="https://www.gov.uk/government/organisations/surveillance-camera-commissioner">Surveillance Camera Commissioner</a>; along with the <a href="https://ico.org.uk">Information Commissioner’s Office</a>, which has responsibility for overseeing data protection in the UK. Also relevant to the use of face recognition systems is the <a href="https://www.gov.uk/government/organisations/biometrics-commissioner">Office of the Biometrics Commissioner</a>.</p>
<h2>Surveillance Camera Day</h2>
<p>Most surveys <a href="https://www.le.ac.uk/oerresources/criminology/msc/unit8/page_16.htm">suggest that</a> the public are in favour of basic CCTV cameras, but the question for those who set the rules is whether citizens would still support these systems if they knew what they were becoming capable of. Judging by <a href="https://www.theguardian.com/commentisfree/2018/may/16/facial-recognition-useless-police-dangerous-met-inaccurate">most reactions</a> in the media to face recognition, it seems not. </p>
<figure class="align-right zoomable">
<a href="https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=237&fit=clip" srcset="https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=876&fit=crop&dpr=1 600w, https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=876&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=876&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=1101&fit=crop&dpr=1 754w, https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=1101&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/279998/original/file-20190618-118526-1hrz7rk.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=1101&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">The future once.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/cctv-surveillance-camera-375997810?src=rmfzlnrmQickyz0pxFCNaQ-1-3&studio=1">TUM2282</a></span>
</figcaption>
</figure>
<p>I suspect that most of the advances in technology could be used to improve the system if they were regulated properly, but cameras must be seen to be delivered in the interests of society and with the support of voters. So where should policymakers draw a line in the sand?</p>
<p>To help with this, a world first is about to take place in the UK on June 20: <a href="https://videosurveillance.blog.gov.uk/2019/04/23/get-involved-in-surveillance-camera-day/">Surveillance Camera Day</a>. This is not intended to be a celebration of surveillance cameras but to allow people to influence how they develop by raising awareness about their capabilities, merits and consequences. It will include everything from <a href="http://www.crisp-surveillance.com/blog/194144/announcing-national-surveillance-camera-day">open days</a> at a number of CCTV control centres to public factsheets to discussions in the media. Everyone can contribute to the conversation through #cameraday2019. </p>
<p>The direction of travel for surveillance cameras does not need to be towards a defined technological determinism where it inevitably becomes more and more intrusive. Surveillance Camera Day represents an opportunity for everyone to help shape the discussion. It will be interesting to observe how the general public and other players respond.</p><img src="https://counter.theconversation.com/content/118931/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>William Webster is an unpaid advisor to the Surveillance Camera Commissioner and is responsible for leading the civil engagement element of the national surveillance camera strategy.</span></em></p>If you thought police surveillance was mere CCTV, it’s time to catch up on what’s happening on the other side of the lens.William Webster, Professor and Director, Centre for Research into Information, Surveillance and Privacy, University of StirlingLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1145872019-04-04T19:13:25Z2019-04-04T19:13:25ZWhy do we mix up faces? Game of Thrones might help us find the answer<figure><img src="https://images.theconversation.com/files/266978/original/file-20190402-177199-qg32l.jpg?ixlib=rb-1.1.0&rect=73%2C73%2C4846%2C2899&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Maisie Williams plays Arya Stark in Game of Thrones and is one of the actors whose face is part of a new test of face recognition skills.</span> <span class="attribution"><span class="source">AAP</span>, <a class="license" href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA</a></span></figcaption></figure><p>In the <a href="https://www.hbo.com/game-of-thrones">Game of Thrones</a> universe, confusing a photograph of actor <a href="https://en.wikipedia.org/wiki/Jack_Gleeson">Jack Gleeson</a>, who played the popular HBO TV show’s <a href="https://www.theguardian.com/tv-and-radio/tvandradioblog/2014/apr/14/game-of-thrones-king-joffrey-baratheon">despised sadist Joffrey Baratheon</a>, for one of <a href="https://en.wikipedia.org/wiki/Maisie_Williams">Maisie Williams</a>, the beloved <a href="https://www.factinate.com/things/38-fierce-facts-arya-stark/">Arya Stark</a>, is an egregious case of mistaken identity. </p>
<p>Admittedly, Gleeson is sporting dark hair rather than Joffrey’s customary blond, but even so. If nothing else, Gleeson is male and Williams female.</p>
<p>This is just one of the failures we have seen when we used characters from Game of Thrones to develop a new test of human face recognition. Thanks to the series, our <a href="https://osf.io/8g6tm/">research</a> shows that becoming familiar with a face and reliably recognising a person are complex processes.</p>
<h2>Facial recognition feels deceptively easy</h2>
<p>Humans are a social species and recognising people’s faces is a crucial skill. We seem to do it effortlessly, but there are immense individual differences. Some people are practically “<a href="http://doi.org/10.1016/j.tics.2005.02.011">face-blind</a>” and can’t identify people in their own family, while others, dubbed “super-recognisers”, <a href="http://doi.org/10.3758/PBR.16.2.252">claim they never forget a face</a>.</p>
<p>The latter seem so infallible that security and law enforcement agencies are now seeking to employ them. But it is critical to understand their limits, too.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/super-recognisers-accurately-pick-out-a-face-in-a-crowd-but-can-this-skill-be-taught-112003">Super-recognisers accurately pick out a face in a crowd – but can this skill be taught?</a>
</strong>
</em>
</p>
<hr>
<p>As researchers, we noticed significant limitations in existing tests of face recognition. Some tests rely on people’s recognition of famous faces – politicians, athletes, actors, musicians. But people vary widely in how often they might encounter these people so it is difficult to know if mistaken identities are the result of poor face recognition ability, limited exposure or a mere lack of interest in one of these areas. </p>
<p>Other tests use strictly controlled faces of strangers – stripped of hair, glasses and other adornments – that people study in the lab. But this seems very unlike the way we become familiar with faces in the real world, where we encounter people in many different situations and without any real intent to study their facial features.</p>
<p>Which brings us to Game of Thrones. Fans of the show have been exposed to hundreds of characters over a period of years, while they aged, changed hairstyles (think <a href="https://xtava.com/blogs/hairspiration/game-of-thrones-hairstyles-cersei">Cersei Lannister</a>) or became disfigured. One of us (Christel Devue) is supposedly good with faces, at least according to the standard lab-based tests, but noticed she kept mixing up similar-looking characters while watching the show. </p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=338&fit=crop&dpr=1 600w, https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=338&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=338&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=424&fit=crop&dpr=1 754w, https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=424&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/267277/original/file-20190403-177193-d5xb65.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=424&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Over the seasons of Game of Thrones, fans will have watched the same characters for several years.</span>
<span class="attribution"><span class="source">AAP</span>, <a class="license" href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA</a></span>
</figcaption>
</figure>
<h2>Game of Thrones as a research tool</h2>
<p>When we conducted our test, Game of Thrones had been running for six seasons and had featured many hundreds of actors. Because the show has such a dedicated following, the internet provided us with a <a href="https://gameofthrones.fandom.com/wiki/Category:Story">wealth of information</a> to develop a very well-controlled experiment. </p>
<p>For example, some fans had calculated how long each actor was visible on screen. As well as being able to gauge actors’ exposure levels, we also knew <a href="https://www.imdb.com/title/tt0944947/fullcredits?ref_=tt_ov_st_sm">how much time had elapsed</a> since characters were last seen (before facing a gruesome death).</p>
<p>We tested 32 participants who had watched all six seasons only once, as each one was released. This way, we were sure everyone had had the same exposure to all the actors and at about the same time. Participants had not read <a href="http://www.georgerrmartin.com/">George RR Martin</a>’s original novels, so would not know the characters from a different source. </p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/267278/original/file-20190403-177178-1mp7qi5.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">British actor Alfie Allen plays Theon Greyjoy, also known as Reek, in Game of Thrones. He sometimes has a beard and such simple changes can throw off recognition.</span>
<span class="attribution"><span class="source">AAP</span>, <a class="license" href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA</a></span>
</figcaption>
</figure>
<p>We showed participants 90 headshots of actors (not in character) who had four different levels of exposure in the show (as main heroes, lead characters, support characters and bit parts), mixed with 90 strangers. They judged whether each face was familiar, rated their confidence in that judgement, and tried to identify and name the character or the actor. </p>
<p>Half of the participants were shown pictures in which actors’ headshots were similar to their character in the show, while the others saw pictures in which the actors’ appearance differed (for example, different hairstyle, facial hair, make-up, glasses).</p>
<p>Surprisingly, we discovered that simple changes in hairstyle or the mere passage of time throw off recognition of actors in all exposure levels, even those that participants saw repeatedly for years. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/combining-the-facial-recognition-decisions-of-humans-and-computers-can-prevent-costly-mistakes-97365">Combining the facial recognition decisions of humans and computers can prevent costly mistakes</a>
</strong>
</em>
</p>
<hr>
<h2>Limitations in human face recognition</h2>
<p>No one, including the best recognisers, recognised all faces. The best hit rate was about 80% of the actors, but from someone who also falsely recognised about 50% of strangers. Good recognisers were distinguished not because they recognised more faces than others, but by their ability to reject novel faces as unfamiliar. </p>
<p>Some participants were extremely confident while they were in fact reporting very inaccurate information. Others were uncertain about their responses but were spot-on most of the time. Some could not remember many names and gave really convoluted descriptions of characters.</p>
<p>Joffrey and Arya were not the only cross-gender mix-up. A photograph of a short-haired <a href="https://variety.com/2018/tv/news/game-of-thrones-sibel-kekilli-shae-german-turkish-1203018180/">Sibel Kekilli</a> (<a href="https://gameofthrones.fandom.com/wiki/Tyrion_Lannister">Tyrion Lannister</a>’s mistress Shae) was identified by several participants as <a href="https://en.wikipedia.org/wiki/Isaac_Hempstead_Wright">Isaac Hempstead Wright</a> (<a href="https://gameofthrones.fandom.com/wiki/Bran_Stark">Bran Stark</a>). Such misidentifications and others were often based on superficial features like hair colour or style, facial hair or the shape of a head, showing how important these are.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=502&fit=crop&dpr=1 754w, https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=502&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/267279/original/file-20190403-177175-65p4dl.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=502&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Peter Dinklage as Tyrion Lannister. Some Game of Thrones fans have calculated how long each actor was visible on screen.</span>
<span class="attribution"><span class="source">AAP</span>, <a class="license" href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA</a></span>
</figcaption>
</figure>
<p>Although recognition rates increased with greater levels of exposure, performance was far from what you’d expect of fans of a show. Identification errors occurred at all exposure levels. Only the most prominent actors were correctly identified and named more often than they were just recognised. As for bit parts, they were never correctly identified and only rarely recognised, even if their brief appearance was sometimes shattering (one of them was the cause of the <a href="https://inews.co.uk/culture/television/red-wedding-game-of-thrones/">Red Wedding</a>). Familiarity judgements for these actors were barely higher than for strangers.</p>
<p>This research has important practical implications. Criminals know simple disguises make it possible to escape prosecution. Many innocent people are convicted based on errors in eyewitness testimony. </p>
<p>Our research confirms recognition based on brief encounters is likely to be wrong. Confidence of a witness in these conditions is not a good indicator of whom they have actually seen. Moreover, while super-recognisers have made useful contributions to law enforcement agencies, they are not infallible.</p>
<p>As for people planning to watch the final season of Game of Thrones, perhaps a refresher viewing is in order to avoid any confusion and ensure you really do know who’s who.</p><img src="https://counter.theconversation.com/content/114587/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Facial recognition is a crucial skill, but difficult to test accurately. Researchers are taking advantage of the popularity of Game of Thrones to test the limits of our ability to identify a familiar face.Christel Devue, Lecturer in Cognitive Psychology, Te Herenga Waka — Victoria University of WellingtonGina Grimshaw, Associate Professor, Te Herenga Waka — Victoria University of WellingtonLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/1120032019-02-20T18:46:41Z2019-02-20T18:46:41ZSuper-recognisers accurately pick out a face in a crowd – but can this skill be taught?<figure><img src="https://images.theconversation.com/files/259874/original/file-20190220-148533-6r77ft.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">One of these people is on a wanted list for theft. A super-recogniser may pick them at a glance. </span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/benicasim-spain-july-19-people-fans-188040776?src=necrspCtWSs3EeN0FrxxZg-1-12">from www.shutterstock.com</a></span></figcaption></figure><p>Yenny is 26 years old, lives in Melbourne, and has a very specific talent. </p>
<p>One day, she was driving her car when she recognised a man who had been several years below her at high school and whom she hadn’t seen for more than ten years. What makes this particularly impressive is that she recognised him from the briefest glimpse in her rear-view mirror while he was driving the car behind hers. </p>
<p>Yenny recounts many such amazing feats of recognition and is one of a very small proportion of the population known as “super-recognisers”. She was the top performer on a <a href="https://facetest.psy.unsw.edu.au">national test of face recognition abilities</a> in Australia, coming first out of 20,000 participants.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/combining-the-facial-recognition-decisions-of-humans-and-computers-can-prevent-costly-mistakes-97365">Combining the facial recognition decisions of humans and computers can prevent costly mistakes</a>
</strong>
</em>
</p>
<hr>
<p>Could you learn to spot a face as well as Yenny? Well … maybe. Our <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0211037">new research</a> shows that many training courses offered in this field of expertise are ineffective in improving people’s accuracy in face identification. </p>
<p>But other ways of learning how to identify faces may work; we’re just not yet sure exactly how. </p>
<h2>In-demand expertise</h2>
<p><a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3904192/">Super-recognisers</a> are used by police and security agencies to spot targets in crowded train stations, monitor surveillance footage, and track people of interest. </p>
<p>During the 2011 London riots, for example, super-recognisers from the Metropolitan Police <a href="http://www.bbc.com/future/story/20150611-the-superpower-police-now-use-to-tackle-crime">identified more than 600 people</a> from very poor-quality surveillance footage – a task that not even the best facial recognition software can perform reliably.</p>
<p>So can anyone become a super-recogniser? Can you make up for a lack of superpowers through training? In our <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0211037">paper</a> we assessed the effectiveness of training courses given to practitioners who make facial identification decisions for a living.</p>
<p>We reviewed 11 training courses that comply with international training standards from Australia, UK, US and Finland. </p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=293&fit=crop&dpr=1 600w, https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=293&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=293&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=368&fit=crop&dpr=1 754w, https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=368&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/259884/original/file-20190220-148506-1mbtagr.PNG?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=368&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Sample test of face recognition: are the side-by-side images of the same, or different people? Answers can be found in the paper acknowledgements (click on journal link).</span>
<span class="attribution"><a class="source" href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0211037">Towler and colleagues, PLOS ONE</a>, <a class="license" href="http://creativecommons.org/licenses/by/4.0/">CC BY</a></span>
</figcaption>
</figure>
<p>We found that training courses typically teach facial anatomy – focusing on the muscles, bones and shape of the face – and instruct trainees to inspect faces feature by feature. Novices and genuine trainees completed one of four training courses and we tracked their identification accuracy from before to after training.</p>
<p>Surprisingly, we found the training courses had almost no effect on people’s accuracy. This was especially surprising to the people who took the training – an astonishing 93% of trainees thought the training had improved their ability to identify faces. </p>
<p>Our research shows that even the world’s best available training – used to train police, border control agents, forensic scientists and other security personnel – does not compensate for talent in face recognition. </p>
<p>This is consistent with recent research suggesting that our face identification abilities are largely predetermined by <a href="https://www.pnas.org/content/107/11/5238.long">genetics</a>.</p>
<h2>Forensic facial examiners</h2>
<p>This may come as disappointing news to people who hope to become a super-recogniser. But all is not lost.</p>
<p>Scientists have recently discovered that some specialist groups of practitioners show very high levels of accuracy. Forensic facial examiners routinely compare images of faces to turn CCTV images into informative face identification evidence in criminal trials. Recent work shows that they too outperform novices in very <a href="https://www.pnas.org/content/115/24/6171">challenging tests</a>.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/259875/original/file-20190220-148536-14jt4tv.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">CCTV footage doesn’t always produce clear images, so identifying people can be difficult.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/cctv-system-security-inside-restaurantsurveillance-camera-1064540882?src=fGFdv2-lM7CohNJQnywhhQ-1-25">Shutterstock.com</a></span>
</figcaption>
</figure>
<p>Forensic facial examiners present a paradox for scientists. They perform face identification tasks with a high degree of accuracy, and this ability appears to be acquired through professional experience and training.</p>
<p><a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0211037">Our study</a> suggests there is no benefit of face identification training courses when tested immediately before and after. </p>
<p>In addition, <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0103510">previous work</a> has suggested that merely performing face-matching tasks in daily work is not sufficient to improve accuracy. Some passport officers have been working for 20 years and perform no better than others who have been working for just a few months.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/passport-staff-miss-one-in-seven-fake-id-checks-30606">Passport staff miss one in seven fake ID checks</a>
</strong>
</em>
</p>
<hr>
<p>This paradox suggests there is something particular about the type of training and professional experience that forensic facial examiners receive that enables them to develop visual expertise in identifying faces, and which isn’t provided by standard training courses.</p>
<h2>How do they do it?</h2>
<p>In our current research we are working closely with government agencies to uncover the basis of forensic facial examiners’ expertise. For example, we now know that part of their expertise comes from using a very particular <a href="https://www.ncbi.nlm.nih.gov/pubmed/28045276">comparison strategy</a>, where they break the face down into individual facial features and then slowly and systematically assess the similarity of each feature in turn. </p>
<p>Interestingly, the nature of this expertise appears to be qualitatively different to that of super-recognisers – Yenny recognised her old classmate using a quick, intuitive process as she glanced in the rear-view mirror.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=338&fit=crop&dpr=1 600w, https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=338&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=338&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=424&fit=crop&dpr=1 754w, https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=424&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/259871/original/file-20190220-148523-r80rmc.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=424&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Super-recognisers can pick out a face they may have seen only once before.</span>
<span class="attribution"><a class="source" href="https://www.shutterstock.com/image-photo/men-women-walking-through-counter-multiethnic-1045777096?src=bkJ1_powZ3RFFDBQzZVMFQ-1-56">Shutterstock.com</a></span>
</figcaption>
</figure>
<p>However, these snap judgements made by super-recognisers may not be suitable for the type of identification evidence that forensic facial examiners give in court, where a careful analysis of facial images is necessary to support identification decisions. Importantly, forensic facial examiners provide detailed reports of the observations used to support their decisions, which can then be cross-examined in court.</p>
<h2>Trainable vs hardwired</h2>
<p>Super-recognisers and forensic facial examiners use distinct routes to high performance in face identification. </p>
<p>Effective training appears to target the slower, deliberate and analytical visual processing that characterises forensic facial examiners. </p>
<p>The faster and more intuitive skill that enabled Yenny to recognise faces of relative strangers in her rear-view mirror is likely to be untrainable, and hard-wired.</p>
<p>This raises the question of how to balance these different sources of expertise. It may be that super-recognisers are best suited to surveillance-type roles, such as monitoring CCTV or searching for targets in large crowds. </p>
<p>Forensic facial examiners may be better suited to providing identification evidence to the court, which requires thorough explanations of how and why the expert came to their decision.</p>
<p>Alternatively, it may be possible to train super-recognisers in the expert skills characterising forensic facial examination, or to form teams that include both types of expert.</p>
<p>The aim of our work is to integrate these sources of human expertise with the latest face recognition software to improve the accuracy of face identification evidence. Such a system can make society safer, but also fairer, by reducing the likelihood of wrongful convictions. </p>
<p>Can you beat Yenny’s high score of 88% on the super-recogniser test? Find out <a href="https://facetest.psy.unsw.edu.au">here</a>.</p><img src="https://counter.theconversation.com/content/112003/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>David White receives funding from The Australian Research Council (DP190102916, LP160101523).</span></em></p><p class="fine-print"><em><span>Alice Towler does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Even the world’s best available training – used to train police, border control agents and other security personnel – does not compensate for natural talent in face recognition.Alice Towler, Post-doctoral Research Fellow, UNSW SydneyDavid White, Scientia Fellow, UNSW SydneyLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/973652018-06-04T20:28:22Z2018-06-04T20:28:22ZCombining the facial recognition decisions of humans and computers can prevent costly mistakes<figure><img src="https://images.theconversation.com/files/221482/original/file-20180604-177134-149ocx3.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Students tested on their ability to tell whether two images were of the same person were wrong 30% of the time.</span> <span class="attribution"><a class="source" href="https://www.shutterstock.com/download/confirm/246484747?src=Jt8i_XloE2Ty7iVPIDiVgA-1-0&size=huge_jpg">Shutterstock</a></span></figcaption></figure><p>After a series of bank robberies that took place in the US in 2014, police <a href="https://theintercept.com/2016/10/13/how-a-facial-recognition-mismatch-can-ruin-your-life/">arrested Steve Talley</a>. He was beaten during the arrest and held in maximum security detention for almost two months. His estranged ex-wife identified him as the robber in CCTV footage and an FBI facial examiner later backed up her claims.</p>
<p>It turned out Talley was not the perpetrator. Unfortunately, his arrest left him with extensive injuries, and led to him losing his job and a period of homelessness. Talley has now become an example of what can go wrong with facial identification. </p>
<p>These critical decisions rest on the ability of humans and computers to decide whether two images are of the same person or different people. Talley’s case shows how errors can have profound consequences. </p>
<p>My research focuses on how to improve the accuracy of these decisions. This can make society safer by protecting against terrorism, organised crime and identity fraud. And make them fairer by ensuring that errors in these decisions do not lead to people being wrongly accused of crimes. </p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/dna-facial-prediction-could-make-protecting-your-privacy-more-difficult-94740">DNA facial prediction could make protecting your privacy more difficult</a>
</strong>
</em>
</p>
<hr>
<h2>Identifying unfamiliar faces</h2>
<p>So just how accurate are humans and computers at identifying faces?</p>
<p>Most people are extremely good at recognising faces of people they know well. However, in all of the critical decisions outlined above, the task is not to identify a familiar face, but rather to verify the identity of an unfamiliar face. </p>
<p>To understand just how challenging this task can be, try it for your self: are the images below of the same person or different people?</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=445&fit=crop&dpr=1 600w, https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=445&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=445&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=559&fit=crop&dpr=1 754w, https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=559&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/221322/original/file-20180601-142075-feuuwq.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=559&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Same or different person? The correct answer is provided at the end of this article.</span>
</figcaption>
</figure>
<h2>Humans versus machines</h2>
<p>The above image pair is one of the test items my colleagues and I used to evaluate the accuracy of humans and computers in identifying faces, in a paper published last week in <a href="http://www.pnas.org/content/early/2018/05/22/1721355115">Proceedings of the National Academy of Science</a>. </p>
<p>We recruited two groups of professional facial identification experts. One group were international experts that produce forensic analysis reports for court (Examiners). Another group were face identification specialists that made quicker decisions, for example when reviewing the validity of visa applications or in forensic investigation (Reviewers). We also recruited a group of “<a href="https://facetest.psy.unsw.edu.au/UNSWfacetest.html">super-recognisers</a>” who have a natural ability to identify faces, similar to <a href="http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150036">groups</a> that have been deployed as face identification specialists in the <a href="https://www.newyorker.com/magazine/2016/08/22/londons-super-recognizer-police-force">London Metropolitan Police</a>. </p>
<p>Performance of these groups compared to undergraduate students and to the algorithms is shown in the graph below. </p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=566&fit=crop&dpr=1 600w, https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=566&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=566&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=712&fit=crop&dpr=1 754w, https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=712&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/221325/original/file-20180601-142083-1gvsk9d.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=712&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Accuracy of participant groups and face recognition algorithms in Phillips et al (2018).</span>
<span class="attribution"><span class="source">PNAS</span></span>
</figcaption>
</figure>
<p>Black dots on this graph show the accuracy of individual participants, and the red dots show the average performance of the group. </p>
<p>The first thing to notice is that there is a clear ordering of performance across the groups of humans. Students perform relatively poorly as a group – with over 30% errors on average – showing just how challenging the task is. </p>
<p>The professional groups fare far better on the task, making less than 10% errors on average and nine out of 87 attaining the maximum possible score on the test. </p>
<p>Interestingly, the super-recognisers also performed extremely well, with three out of 12 attaining the maximum possible score. These people had no specialist training or experience in performing face identification decisions, suggesting that selecting people based on natural ability is also a promising <a href="http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0148148">solution</a>.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/class-action-against-facebook-over-facial-recognition-could-pave-the-way-for-further-lawsuits-95215">Class action against Facebook over facial recognition could pave the way for further lawsuits</a>
</strong>
</em>
</p>
<hr>
<p>Performance of the algorithms is shown by the red dots on the right of the graph. We tested three iterations of the same algorithm as the algorithm was improved over the last two years. There is a clear improvement of this algorithm with each iteration, demonstrating the major advances that <a href="https://arxiv.org/abs/1804.06655">Deep Convolutional Neural Network</a> technology have made over the past few years.</p>
<p>The most recent version of the algorithm attained accuracy that was in the range of the very best humans. </p>
<h2>The wisdom of crowds</h2>
<p>We also observed large variability in all groups. No matter which group we look at, performance of individuals spans the entire measurement scale – from random guessing (50%) to perfect accuracy (100%). </p>
<p>This variation is problematic, because it is individuals that provide face identification evidence in court. If performance varies so wildly from one individual to the next, how can we know that their decisions are accurate?</p>
<p>Our study provides a solution to this problem. By averaging the responses of groups of humans, using what is known as a “<a href="https://en.wikipedia.org/wiki/Wisdom_of_the_crowd">wisdom of crowds</a>” approach, we were able to attain near-perfect levels of accuracy. Group performance was also more predictable than individual accuracy.</p>
<p>Perhaps the most interesting finding was when we combined the decisions of humans and machines. </p>
<p>By combining the responses of just one examiner and the leading algorithm, we were able to attain perfect accuracy on this test – better than either a single examiner or the best algorithm working alone.</p>
<h2>Face recognition in Australia</h2>
<p>This is a timely result as Australia rolls out the <a href="https://theconversation.com/lets-face-it-well-be-no-safer-with-a-national-facial-recognition-database-85179">National Face Identification scheme</a>, which <a href="https://www.theguardian.com/technology/2018/may/30/facial-matching-system-is-racist-human-rights-law-centre-warns">will enable</a> police agencies to search large databases of images using face recognition software.</p>
<hr>
<p>
<em>
<strong>
Read more:
<a href="https://theconversation.com/close-up-the-governments-facial-recognition-plan-could-reveal-more-than-just-your-identity-92261">Close up: the government's facial recognition plan could reveal more than just your identity</a>
</strong>
</em>
</p>
<hr>
<p>Importantly, this application of face recognition technology is not automatic – like automated border control systems are. Rather, the technology generates “candidate lists” like the one shown below. For the systems to be of any use, humans must review these candidate lists to decide if the target identity is present. </p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=579&fit=crop&dpr=1 600w, https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=579&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=579&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=728&fit=crop&dpr=1 754w, https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=728&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/221328/original/file-20180601-142093-su3iox.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=728&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">A ‘candidate list’ returned by face recognition software performing a database search. Humans must adjudicate the output of these systems by deciding whether the person in the ‘probe’ image – the image at the top – is pictured in the array below, and if so to select the matching face. The correct answer is provided at the end of this article.</span>
</figcaption>
</figure>
<p>In a <a href="http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0139827">2015 study</a> my colleagues and I found that the average person makes errors on one in every two decisions when reviewing candidate lists, and chooses the wrong person 40% of the time! </p>
<p>False positives like these can waste precious police time, and have potentially <a href="https://theintercept.com/2016/10/13/how-a-facial-recognition-mismatch-can-ruin-your-life/">devastating effect on people’s lives</a>. </p>
<p>The study we published this week suggests that protecting against these costly errors requires careful consideration of both human and machine components of face recognition systems. </p>
<hr>
<p><em>Correct answers: The pair of images are different people. The matching image in the candidate list is top row, second from left.</em></p><img src="https://counter.theconversation.com/content/97365/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>David White receives funding from an Australian Research Council Linkage Project in partnership with the Department of Foreign Affairs and Trade (LP160101523 ). He is a UNSW Scientia Fellow and a member of the Academic Research and Innovation Group of the Biometric Institute. </span></em></p>Same person or different person? Most people are extremely good at recognising faces of people they know well, but not so much strangers. See how well you perform on the tests in this story.David White, Scientia Fellow, UNSW SydneyLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/974182018-06-04T13:49:48Z2018-06-04T13:49:48ZHow CCTV surveillance poses a threat to privacy in South Africa<figure><img src="https://images.theconversation.com/files/220930/original/file-20180530-120505-j9xs87.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">CCTV cameras are becoming a “normal” feature of public life, tracking peoples’ movements as a matter of course.</span> <span class="attribution"><span class="source">Shutterstock</span></span></figcaption></figure><p>Locational privacy is a fairly new and novel aspect of privacy rights. It refers to the right of people to move about freely, without having their movements tracked. </p>
<p>But as CCTV cameras become more widespread in public spaces for use in a range of functions such as crime-fighting, it’s becoming more difficult for people to protect this kind of privacy in public spaces.</p>
<p>The cameras, linked to a display monitors, can be used to monitor human movements in particular spaces, including streets and shopping centres. A video recorder can also be added to record activities. But, the problem with CCTV is always the human capacity to process the information gleaned from the cameras. The cameras can only film fixed areas. Unless they are ubiquitous, they cannot be used to track movements.</p>
<p>The need for human monitoring places a natural limit on the analysis of camera footage. But, with digital tools of analysis, this is changing. When linked to a computer loaded with software capable of algorithmic analysis, huge amounts of footage can be analysed. These camera based surveillance systems can capture information about a person’s physical location. Some may only provide real time information, while others may record information for further analysis.</p>
<p>But governments of a more authoritarian bent can misuse this information to establish people’s movements, political activities and associations. People may not participate as robustly in democratic life as they would if they feel that they are being watched, and their movements tracked.</p>
<p>Invasive forms of data analysis such as number plate and facial recognition are being introduced in South African cities without any public debate about the implications for privacy in public spaces. Likewise, there’s no debate about about their implications for the ability of citizens to practice a range of rights in these spaces, such as the right to assemble.</p>
<h2>Ubiquity</h2>
<p>Increasingly, CCTV cameras are becoming a “normal” feature of public life, tracking peoples’ movements as a matter of course. Video analysis tools also allow for more sophisticated analyses of footage. </p>
<p>Computer analysis enables CCTV to be turned into “smart dataveillance” devices (that conduct surveillance through the collection and computerised analysis of data), which make individuals and their movements more visible to the state. These are meant to assist in “smart” policing, whereby police use data tools to enhance the effectiveness of policing.</p>
<p>Another example is facial recognition technologies. These can be used to identify a particular person from a facial database. Potentially, these technologies can, and are, being used to identify people engaging in politically activities, such as protests. This triggers concerns that governments may be tempted to use them for anti-democratic purposes. </p>
<p>South Africa has followed international trends in street-level surveillance and embraced technologies whose affect on crime fighting and intelligence work are, at best, unclear and contested. International academic <a href="https://www.routledgehandbooks.com/doi/10.4324/9780203814949.ch3_2_c">research</a> points to CCTV systems being most effective in specific contexts, such as parking lots, and least effective in open spaces.</p>
<p>Other kinds of crime such as white collar crime and domestic crime, are not recorded by street cameras, which perpetuates an ideology of crime being street crime perpetrated by strangers. </p>
<p><a href="https://www.routledge.com/Eyes-Everywhere-The-Global-Growth-of-Camera-Surveillance/Doyle-Lippert-Lyon/p/book/9780415696555">Critics</a> have also blamed the use of CCTV systems for displacing crime, rather than deterring it. Where reductions in crime levels have taken place because of CCTV, they were localised and often not statistically significant.</p>
<figure class="align-right ">
<img alt="" src="https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=237&fit=clip" srcset="https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=425&fit=crop&dpr=1 600w, https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=425&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=425&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=534&fit=crop&dpr=1 754w, https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=534&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/220939/original/file-20180530-120487-1yxlckl.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=534&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">A security officers monitoring activity captured on CCTV.</span>
<span class="attribution"><span class="source">shutterstock</span></span>
</figcaption>
</figure>
<p>The difficulties of assessing the impacts of CCTV on crime is made harder by the fact that local authorities have not been undertaking independent impact assessments (including on privacy). This means that the public is forced to rely on the state’s version of events, which for public relations purposes, emphasises the positive impacts. Yet, in Cape Town in 2015 for instance, the police were criticised for making only 107 arrests following 2640 criminal incidents <a href="https://businesstech.co.za/news/technology/111529/sa-police-not-using-cctv-footage-to-catch-criminals-da/">caught on camera</a>.</p>
<p>In 2016, the City of Johannesburg <a href="http://ewn.co.za/2016/06/26/City-of-JHB-installs-smart-cameras-around-city">announced</a> that it was rolling out smart CCTV cameras complete, with automatic number plate and facial recognition technologies, as part of its <a href="http://www.gautengfilm.org.za/news/news-archive/2008/december-2008/360-making-the-inner-city-safer-a-city-of-joburg-initiative">‘safe cities’ initiative</a>. </p>
<p>Yet at the time of writing, the City had enacted no requirement for signage at the entrance to an area under CCTV surveillance – a key privacy protection requirement. The City was in the process of finalising a policy on the roll-out of CCTVs, coupled with a master plan, but these were still at draft stage, pointing to the fact that the technology had run ahead of the policy.</p>
<p>CCTV rollouts tend to <a href="http://www.saflii.org/khayelitshacommissionreport.pdf">“follow the money”</a>. In other words, they tend to follow patterns of wealth in the major metropolitan cities in South Africa. This contributes to the <a href="https://businesstech.co.za/news/lifestyle/209473/joburg-is-looking-at-making-big-changes-to-boomed-suburbs/">enclosure of city spaces</a> by private capital, and consequently to the privatisation of public spaces and the reproduction of <a href="http://journals.sagepub.com/doi/abs/10.1177/0269094215618595">spacial inequalities</a>.</p>
<p>It’s not at all clear if the growing capacity of local governments to collect street-level data on peoples’ movements is making a substantial contribution to policing, as the police do not use this data routinely.</p>
<h2>The risk of dumbing down policing</h2>
<p>Technology is being used as a silver bullet for policing of public spaces, when more basic interventions may be more appropriate (such as improving investigative techniques), risks dumbing down policing. Yet, at the same time, the regulation of CCTV for its impacts on privacy is lagging behind the actual rollout of the technology.</p>
<p>Data-driven surveillance tools, such as smart CCTV, consistently over promise but under deliver in fighting crime. Yet, governments are adept at creating panic about crime to obscure these failings. People’s fear of crime, and their need to feel protected from it, should not stop them from asking the critical questions that need to be asked. </p>
<p><em>This is an edited excerpt from the author’s latest book, <a href="https://witspress.co.za/catalogue/stopping-the-spies/">Stopping the Spies: Constructing and Resisting the Surveillance State</a>, published by Wits University Press.</em></p><img src="https://counter.theconversation.com/content/97418/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Jane Duncan receives funding from the Open Society Foundation for South Africa. She is affiliated with the Media Policy and Democracy Project and the Right 2 Know Campaign. </span></em></p>As CCTV cameras become more widespread, it’s becoming more difficult for people to protect their locational privacy in public.Jane Duncan, Professor and head of the Department of Journalism, Film and Television, University of JohannesburgLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/672112016-11-09T10:19:15Z2016-11-09T10:19:15ZCan you teach recognition to the face-blind?<figure><img src="https://images.theconversation.com/files/145028/original/image-20161108-16685-1pzupa0.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Sorry, but you are...?</span> <span class="attribution"><a class="source" href="http://www.shutterstock.com/pic-213490135/stock-photo-businessman-with-paper-bag-on-the-head-in-the-street.html?src=e1UoWybw8LJb-Z53OcAMig-6-70">Tomas Urbelionis/www.shutterstock.com</a></span></figcaption></figure><p>Every day, we often scan hundreds of faces, instantly recognising those we know. It’s an innate ability that we rarely think about, and in some ways take for granted. But not everyone has this skill: those with a condition called prosopagnosia – also known as “face blindness” – are simply unable to recognise others by their faces.</p>
<p>Throughout years of research into the condition, I have met many people from all walks of life who suffer from face-blindness. I met <a href="http://www.springer.com/gp/book/9783642407833">Tiffany, for example,</a> in Sydney, Australia. She told me about the earliest experience of prosopagnosia she could remember:</p>
<blockquote>
<p>I was about seven years old. There was a low footbridge which my siblings and I crossed to get from our home to school. It had been washed away during heavy rain, so we were forced to walk about 2km to another bridge. </p>
<p>A man driving past stopped and offered us a lift. My siblings accepted, but I refused as I thought he was a stranger. It turned out he was the priest at the church we all attended each week. I had known him for two years – I did not recognise him because he was in normal clothes.</p>
</blockquote>
<p>As you might expect, living in a world where “all faces look the same” can lead to a multitude of daily struggles, forcing sufferers to avoid certain situations out of fear they won’t be able to recognise familiar people.</p>
<figure>
<iframe width="440" height="260" src="https://www.youtube.com/embed/79NVrMQlxvo?wmode=transparent&start=0" frameborder="0" allowfullscreen=""></iframe>
</figure>
<p>Facial recognition is not something that humans learn, rather it is something we are born with. But face-blind sufferers can develop their own coping mechanisms to overcome any problems the condition may cause. They might focus on extra-facial cues such as hairstyle, gait, clothing and jewellery; or on some distinctive facial features such as bushy eyebrows, a big nose, or a face mole instead. What they cannot apparently do is perceive a face as a whole, as a “holistic” stimulus. </p>
<h2>Learning to recognise</h2>
<p>Though prosopagnosia, and <a href="https://theconversation.com/how-an-army-of-super-recognisers-could-help-spot-criminals-and-missing-persons-67571">super recognition at the other end of the spectrum</a>, are now the subject of much research, the disorder is documented as far back as the time of the ancient Greeks. Early descriptions of serious face recognition difficulties come from soldiers <a href="https://books.google.co.uk/books?id=2AfABAAAQBAJ&pg=PA42&lpg=PA42&dq=Peloponnesian+War+facial+recognition&source=bl&ots=VbK3sBmEux&sig=91MEBWszsJLehZ2PBM9rmcqLj9s&hl=en&sa=X&ved=0ahUKEwi8_ePK25jQAhUkCsAKHe3_DDgQ6AEIKTAC#v=onepage&q=Peloponnesian%20War%20facial%20recognition&f=false">injured in the Peloponnesian War</a>. </p>
<p>These historical cases, along with the many others that have been reported much more recently, are cases of acquired prosopagnosia. They describe a form of prosopagnosia that follows a brain injury such as a stroke or head trauma. We now know, however, that prosopagnosia can also exist in people, like Tiffany, who have not been affected by any brain lesion. This second form of prosopagnosia is known as congenital, or developmental, prosopagnosia (CP). CP is not as rare as you might think either: it affects around <a href="https://www.ncbi.nlm.nih.gov/pubmed/19921582">2-2.9% of the general adult population</a> – around 1.5m people in the UK alone.</p>
<p>Our ability to recognise people by their faces is mediated by clusters of neurons that form “face processing areas” in the posterior part of the brain. Our own research has demonstrated that these areas in the brains of those with CP <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/pmc4230164/">work in an atypical way</a>, and do not properly discriminate face from non-face stimuli. </p>
<p>Despite many researchers working to find the causes of CP, we still don’t have a single concrete reason for explaining why these anomalies are present – but we are working on it. Evidence has shown that that, overall, our ability to recognise faces has a <a href="http://www.pnas.org/content/107/11/5238.abstract?ijkey=812fd73874ff75292120f79e1362814a5d53dd44&keytype2=tf_ipsecsha">strong genetic component</a>, and that CP <a href="https://www.ncbi.nlm.nih.gov/pubmed/19334307">runs in families</a>. This suggests that CP may be a genetic condition.</p>
<p>So what can we do to help prosopagnosia sufferers? It is not as simple as sitting someone down to “learn” facial recognition. Though research has found that face-blindness can <a href="http://journal.frontiersin.org/article/10.3389/fnhum.2014.00491/full">be improved with training</a>, there is no simple “cure”. </p>
<p>Our team recently adopted “neuromodulation” to test whether human face recognition can be improved. This is a neuroscientific technique that involves delivering a very small (undetectable) current through the scalp via two small sponge electrodes. We have demonstrated that neuromodulation actually <a href="http://www.sciencedirect.com/science/article/pii/S0028393215302700">leads to improved face-processing skills</a>, and though it is no a “cure”, it does suggest the potential of this approach in future studies aimed at treating CP. </p>
<p>One of our team’s main goals is to find ways to enhance the face processing skills of those suffering with congenital prospagnosia. However, it is currently very difficult to study CP – although there are a large number of sufferers, a relatively small number of people have been <a href="https://theconversation.com/always-forget-a-face-so-does-brad-pitt-dont-just-blame-your-memory-50334">properly diagnosed</a>. This is why we are now <a href="https://www.uel.ac.uk/Schools/Psychology/Research/Prosopagnosia">launching a new website</a>, working with schools and increasing awareness to encourage those who may suffer from the condition to learn more, and to take part in further research.</p><img src="https://counter.theconversation.com/content/67211/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Davide Rivolta does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>We’re working on a ‘cure’ for congenital prosopagnosia.Davide Rivolta, Senior Lecturer in Psychology, University of East LondonLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/675712016-11-07T08:41:10Z2016-11-07T08:41:10ZHow an army of ‘super recognisers’ could help spot criminals and missing persons<figure><img src="https://images.theconversation.com/files/144676/original/image-20161105-27925-1m9vwbk.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">How keen is your eye?</span> <span class="attribution"><a class="source" href="http://www.shutterstock.com/pic-427926511/stock-photo-closeup-of-womans-face-with-digital-pattern.html?src=Juv9wH-f9_oVBB2TWwXleA-1-63">Shutterstock</a></span></figcaption></figure><p>While most of us find it easy to recognise highly familiar faces such as those of family and friends, identifying faces that we have only briefly encountered is much more difficult. In fact, some research suggests that even experienced passport control officers <a href="http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0103510">make a large number of errors</a> when matching faces to identity documents. Yet, recent work reveals that a small number of people may have extraordinary face recognition skills, outperforming typical people on a range of face recognition tasks.</p>
<p>These so-called “super recognisers” have an uncanny ability to recognise faces, remembering people they have not seen for decades, who have substantially changed in appearance, and who <a href="http://www.prosopagnosiaresearch.org/about/super-recognition">they have only fleetingly encountered</a>. Some super recognisers have even been accused of stalking because the person that they recognised did not reciprocate the familiarity.</p>
<p>It is currently unknown how many people truly have superior face recognition skills. Popular tests assess participants’ ability to recognise photographs of celebrities that were taken a long time before they became famous. While these “before they were famous” tests are certainly fun to complete, it is very difficult to control for participants’ previous exposure to each celebrity. </p>
<p>A more reliable option is to assess performance on computerised tests that require participants to memorise faces and to later recall them. The number of correct responses can then be compared to the average score achieved by people with typical face recognition skills. This statistical procedure simply identifies the top 2% of the population – meaning that <a href="http://journal.frontiersin.org/article/10.3389/fpsyg.2016.01378/full">one in 50 people are currently classed as super recognisers</a>.</p>
<p>Recent research carried out at Bournemouth University investigated whether these people are actually processing faces in a different manner to the rest of the population. It has long been known that the optimal way to process faces involves the use of a “configural” or “holistic” processing strategy. </p>
<p>This involves seeing faces as a whole, taking account of all of the facial features and the spacing between them. Interestingly, all of the super recogniser participants displayed heightened configural processing <a href="http://www.sciencedirect.com/science/article/pii/S0010945216301186">on at least one task</a>. We also monitored their eye movements as they looked at faces. While control participants mostly looked at the eyes, super recognisers spent <a href="http://www.tandfonline.com/doi/abs/10.1080/17470218.2016.1161059">more time looking at the nose</a>. It is possible that this more central viewing position promotes the optimal configural processing strategy.</p>
<h2>What makes a super recogniser?</h2>
<p>We also examined the potential causes of super recognition, finding no evidence that these people have higher intelligence levels or excel at all visual or memory tasks. In fact, their superior ability is restricted only to the recognition of faces. It currently seems that some people are simply predisposed to developing this skill, and there is <a href="http://www.pnas.org/content/107/11/5238.short">increasing evidence that face recognition skills are heritable</a>. Twin studies report a higher correlation in face recognition ability for identical compared to non-identical twins, and disorders of face recognition – <a href="http://www.prosopagnosiaresearch.org/index/information">prosopagnosia or face blindness</a> are known to run in families, too.</p>
<figure class="align-center ">
<img alt="" src="https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=400&fit=crop&dpr=1 600w, https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=400&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=400&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=503&fit=crop&dpr=1 754w, https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=503&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/144675/original/image-20161105-27911-fadmw.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=503&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px">
<figcaption>
<span class="caption">Spotting a face in the crowd.</span>
<span class="attribution"><a class="source" href="http://www.shutterstock.com/pic-229320925/stock-photo-city-commuters-high-key-blurred-image-of-people-walking-in-the-street-unrecognizable-faces.html?src=Juv9wH-f9_oVBB2TWwXleA-1-17">Shutterstock</a></span>
</figcaption>
</figure>
<p>Another important finding is that some people seem to be superior at specific face recognition tasks. For instance, while some of our super recognisers were excellent at remembering faces, others had typical face memory skills yet were extremely good at deciding whether pairs of simultaneously presented faces were of the same person or two different people. A further skill that has not yet been tested is the “spotting” of faces in a crowd. Many super recognisers claim to be particularly proficient at this task, and it is possible that some people may be “super spotters” yet not excel at other tasks.</p>
<p>The possibility that there are different subtypes of super recognition is particularly important when considering the applied value of this research. Passport control is one clear candidate for the use of super recognisers, and many policing scenarios would also benefit. Super recognisers might assist with the matching of faces captured on CCTV footage, the comparison of faces to identification documents, or the scanning of crowds for known troublemakers, wanted perpetrators or even missing persons. </p>
<p>They may also help with victim identification, or deciding whether a person moving between borders is using a fraudulent identity or is even a missing child. There may not be enough super recognisers to fill all these tasks in all locations, particularly given different individuals may be needed for different jobs, but an elite team of super recognisers could be deployed in times of need. </p>
<p>There has already been some movement towards this ideal, and certain police forces are now working with academics to <a href="http://whatworks.college.police.uk/About/News/Pages/Facial_recognition.aspx">screen their officers for super recognition</a>. Like many sectors, security and policing agencies are increasingly being assisted by technological advances. However, the discovery of super recognisers presents a refreshing example of how a previously untapped human aptitude can make a real difference to local, national and global security.</p>
<p><em>If you think you are a super recogniser, you can register to participate in research <a href="http://www.prosopagnosiaresearch.org/index/register">here</a>.</em></p><img src="https://counter.theconversation.com/content/67571/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>Sarah Bate does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.</span></em></p>Some people have a remarkable aptitude for recognising a face – are you one of them?Sarah Bate, Senior Lecturer in Psychology, Bournemouth UniversityLicensed as Creative Commons – attribution, no derivatives.tag:theconversation.com,2011:article/578532016-04-15T15:39:29Z2016-04-15T15:39:29ZTelling people apart: new test reveals wide variation in how well we recognise faces<figure><img src="https://images.theconversation.com/files/118928/original/image-20160415-11155-uustgq.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=496&fit=clip" /><figcaption><span class="caption">Face blind or super-recogniser?</span> <span class="attribution"><span class="source">Gunter Loffler</span></span></figcaption></figure><p>It’s three o’clock and you are picking up your son from school. The bell rings. A class of six-year-olds charges out of the building. And you have absolutely no idea which child is your son. They all look exactly the same. </p>
<p>This is not a scene from a low-budget horror movie, but how it feels to have <a href="http://www.nhs.uk/conditions/prosopagnosia/Pages/Introduction.aspx">face blindness</a>. It is actually remarkable that most of us are so good at discriminating between faces, since they are all approximately the same – two eyes above a nose above a mouth. Our visual system has had to become highly sensitive to very subtle differences between them. It’s not foolproof, which is why we often see faces in inanimate objects. </p>
<figure class="align-right zoomable">
<a href="https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=237&fit=clip" srcset="https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=600&fit=crop&dpr=1 600w, https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=600&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=600&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=754&fit=crop&dpr=1 754w, https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=754&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/118776/original/image-20160414-2614-1ifb50a.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=754&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Just another pepper?</span>
<span class="attribution"><a class="source" href="https://www.flickr.com/photos/blakespot/9234512598/in/photolist-f52hSs-3UVrAQ-cCts85-m1NfMs-aK5NU-8QFSiU-8QCLjR-hereTa-4WTKBh-riWetH-4WPp58-8QFN9q-8WtAk3-dpodsB-dNZ8yB-2EKKV9-daoQY3-o7qbF4-akE7zb-fddzUt-oNrap5-ditkH2-3EymNn-8QCL54-pm6tf9-aCReoA-8Pqijw-domNA1-8QCHAX-afrwGG-ayNRvF-8QFRwq-d5ZfMN-aA1YTS-aATYA8-gbDrp6-eZAL2F-92nY8n-bqEZVx-oQt1oT-2GCX7x-aXNNV6-pRmBoU-7c4e4u-pTsbJP-pSY1LP-7fzVDB-pTAjMy-5xufaN-dpHoJL">Blake Patterson</a>, <a class="license" href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA</a></span>
</figcaption>
</figure>
<p>Face blindness, also known as prosopagnosia, affects <a href="http://onlinelibrary.wiley.com/doi/10.1002/ajmg.a.31343/full">as many as</a> one in 50 people – upwards of one million in the UK alone. Some are born with the condition, while others develop it after a stroke or serious head injury. Either way, sufferers can have otherwise normal vision. </p>
<p>Since faces and their expressions allow us to read someone’s emotions and judge their state of mind, face-blind people are at <a href="http://www.sciencedirect.com/science/article/pii/S0022399908001578">high risk</a> of social embarrassment, isolation, anxiety and depression. They have to develop strategies to cope, such as recognising people by their hair, height, clothes, posture, gait or voice. And still they can be vulnerable – if a friend gets a new hairstyle, for example, they can become unrecognisable. </p>
<p>Though the condition is not curable, it can be very reassuring to be diagnosed and have the nature of the condition explained. Once the person has told those around them, it often makes them less anxious about being labelled arrogant or rude. It means healthcare workers can show patients new strategies to compensate, such as trying to deal with individuals rather than groups of people, and interacting with the same person each time where possible – sticking to the same GP, for instance. </p>
<p><a href="http://brain.oxfordjournals.org/content/early/2014/03/29/brain.awu062.full">Some reports also suggest</a> that intensive training may help improve their facial recognition, though this has not been widely available to date. Then there are legal implications, since someone with face blindness will not be much help at an identity parade. </p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=424&fit=crop&dpr=1 600w, https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=424&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=424&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=533&fit=crop&dpr=1 754w, https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=533&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/118931/original/image-20160415-26305-o2ukqj.jpg?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=533&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">‘It was …erm’</span>
<span class="attribution"><a class="source" href="http://www.shutterstock.com/pic-259572743/stock-vector-line-up-police.html?src=csl_recent_image-1">Arkela</a></span>
</figcaption>
</figure>
<h2>A testing problem</h2>
<p>It’s actually not easy to test face perception well. Some tests use photographs, but the viewer can get clues from things like clothing or facial expressions. Other tests rely on memorising faces, but again that could be a result of memory rather than face blindness. These tests tend to be much better at dividing the world into people with good and bad face recognition than grading the wide spectrum that likely exists. They also take between ten and 15 minutes, which is not ideal for health professionals. </p>
<p>We <a href="http://www.sciencedirect.com/science/article/pii/S0042698915003624">recently developed</a> a new test that avoids these shortfalls. It uses simplified face images that <a href="http://www.sciencedirect.com/science/article/pii/S0042698902003620">have been synthesised</a> from ordinary photographs. Unlike photographs, however, they can be very precisely modified to make the faces more or less different from one another. It takes only four minutes from start to finish. </p>
<p>Participants are tested on their ability to tell the odd one out among four faces on a computer screen, cycling through 30 different sets over the duration.
The computer monitors show how well the participant is scoring on each set of faces and adapts the difficulty level using an algorithm. Those scoring well get increasingly more similar faces, while strugglers get less similar faces until the test converges on precisely how sensitive someone is. </p>
<p>If you look at the two sets of faces below, for instance, the left set is much easier to differentiate. About 95% of typical adults would be able to spot that the face on the bottom is the odd one out. In the set on the right, only about 25% of typical adults are able to choose the left-hand face correctly. </p>
<figure class="align-center zoomable">
<a href="https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=45&auto=format&w=1000&fit=clip"><img alt="" src="https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&fit=clip" srcset="https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=45&auto=format&w=600&h=304&fit=crop&dpr=1 600w, https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=30&auto=format&w=600&h=304&fit=crop&dpr=2 1200w, https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=15&auto=format&w=600&h=304&fit=crop&dpr=3 1800w, https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=45&auto=format&w=754&h=382&fit=crop&dpr=1 754w, https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=30&auto=format&w=754&h=382&fit=crop&dpr=2 1508w, https://images.theconversation.com/files/118799/original/image-20160414-2617-l0ovbn.png?ixlib=rb-1.1.0&q=15&auto=format&w=754&h=382&fit=crop&dpr=3 2262w" sizes="(min-width: 1466px) 754px, (max-width: 599px) 100vw, (min-width: 600px) 600px, 237px"></a>
<figcaption>
<span class="caption">Can you spot the odd one out from each group?</span>
<span class="attribution"><span class="source">Gunter Loffler</span></span>
</figcaption>
</figure>
<h2>What we found</h2>
<p>We tested 52 young adults with no known face-perception difficulties or eyesight or neurological problems. Some turned out to be highly sensitive to small face differences, while others much less so. This confirms that face perception can vary substantially between individuals who are not face-blind. </p>
<p>We then tested a female patient who did have lifelong difficulties with face recognition. When she took some of the established face tests, her score was statistically just outside normal. But the Caledonian face test scored her several times poorer than the norm – even the poorest person in our main group was approximately twice as good at telling faces apart. </p>
<p>In short, our test potentially offers two kinds of benefits. First, it offers a much more sensitive way of detecting face blindness than the current alternatives. Not only could this help healthcare workers decide what level of intervention is necessary, it can be a tool for the intensive training we mentioned earlier: because the test faces are constantly altered by the algorithm, participants can use the test a number of times without becoming familiar with them. </p>
<p>Being able to differentiate between the face-recognition abilities of “normal” people is useful as well. It may help identify people who are particularly good with faces. The police are beginning to use these so-called “<a href="http://www.bbc.co.uk/news/uk-england-34544199">super-recognisers</a>” to review CCTV footage to connect different crimes, while they could also be helpful at border control, for example. Differentiating abilities could also lend more or less weight to different identification witnesses in court cases, of course. </p>
<p>The test is also quick to administer and suitable for use by optometrists, GPs, psychologists and neurologists. It is ready to be used with adults up to the age of 50, while we’re currently doing more work into face recognition in older people. We are optimistic that a major step forward in this field could now be within reach.</p><img src="https://counter.theconversation.com/content/57853/count.gif" alt="The Conversation" width="1" height="1" />
<p class="fine-print"><em><span>The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.</span></em></p>As many as one in 50 people are face blind, meaning they can’t tell one face from another.Gunter Loffler, Professor of Optometry, Glasgow Caledonian UniversityAndrew J Logan, Lecturer in Optometry, University of BradfordGael Gordon, Senior Lecturer, Vision Sciences, Glasgow Caledonian UniversityLicensed as Creative Commons – attribution, no derivatives.