Counting crowds is a start – but we also need to understand them

A disaster waiting to happen?

There’s been a flurry of excitement about a new model for counting crowds using data generated by crowd members using Twitter and also their more general mobile phone usage. It was laid out in a recently published study, in which the authors highlight the importance of knowing crowd numbers in order to be able to prevent crowd disasters.

As someone with a close interest in the study of crowds, I applaud any study that contributes to improved crowd safety management. From my own experience of trying to estimate crowd numbers, I also appreciate the inherent problems involved in this area, and can see the benefits in finding more accurate forms of crowd counting, as traditional methods are often susceptible to human error and bias.

For instance, the numbers of those who attend political demonstrations in the UK are often hotly debated, depending on one’s perspective – but the basic fact is that protest organisers tend to overestimate crowd size, while the police and right-wing media usually underestimate them.

As with all scientific research, we need to be careful not to make claims that go beyond the actual findings of particular studies. So I do wonder whether creating crowd models based upon simply knowing the physical characteristics of crowds can really help us ensure their safe facilitation and management.

The limits of modelling

Other crowd modellers have sounded a note of caution about this research, highlighting the point that not everyone in crowds uses smart-phones and in any case such information depends upon mobile phone signals being available.

During large crowd events, mobile network coverage is often compromised by surges in demand. People affected by major incidents can flood it by trying to contact others, or the entire service can even be commandeered by the emergency services (as the City of London Police did briefly during the 7/7 bombings).

But there’s a broader issue with the use of crowd modelling in isolation. Being aware of a crowd’s mere physical size at any fixed point is generally of little use on its own. To prevent crowd disasters, it is vital to also consider the dynamic build-up of crowd size as well as what crowd members are actually doing.

It is far better to proactively implement efforts to ensure safe crowd management (maintaining steady crowd flow, monitoring pinch points for potentially dangerous levels of crowd density) than just responding to flare-ups once they occur.

In order to be able to take quick preventative action, it’s vital to have people in place who can communicate with crowd members and help facilitate their safe movement. This usually means placing trained stewards and crowd density spotters at selected vantage points to regularly report on crowd flow, density levels, visible signs of distress in crowd members and the like. and they are ideally situated to advise on crowd mood and/or behaviour as well as how to address any potential problems before they escalate to dangerous levels.

If these measures are not taken, events can easily escalate to a degree where it is too late to prevent tragedy.

A memorial to the victims of the 2010 Love Parade disaster. Beademung via German Wikipedia, CC BY-SA

That much was demonstrated in a study of the 2010 Love Parade tragedy in Duisburg, Germany, where 21 died and more than 500 were injured in a fatal crush that created a “crowd-quake” of pressure surging through the crowd which people were physically unable to prevent.

Back to psychology

I firmly believe that the fields of crowd psychology and crowd modelling need not be in opposition to each other, and that both disciplines can make valuable contributions to the study of safe crowd management.

But crowd modelling perspectives are only as good as the theoretical approaches that underpin them – and until recently too many of these models didn’t consider psychological theories of collective human behaviour in sufficient detail. When they did, they focused on outdated approaches that wrongly assume that people will inevitably “panic” and/or stampede in crowd emergencies. This is a fallacy, and it’s strongly rejected in my own and other crowd psychologists’ current work.

Nevertheless, I am happy that some very good recent examples of crowd modelling have taken social psychology into account when building their models, such as research into the 7/7 bombings and the work of Keith Still. So I would agree with the claims of the authors of this most recent study into crowd counting that it provides a good base on which to build further research.

But I would also suggest that activity on Twitter and mobile phone networks alone cannot tell us all we need to know about human behaviour in crowds and their safe management. We all have more work to do if we are to keep people safe in crowds and prevent often avoidable crowd disasters.

We produce knowledge-based, ethical journalism. Please donate and help us thrive. Tax deductible.