Modelling collective behaviour in computer simulations

See how our research, funded by the Engineering and Physical Sciences Research Council, is taking the first step towards placing group identity into computer models to better simulate the behaviour of crowds and improve public safety.


The use of computer models

Computer models are being increasingly used to plan the safety of large crowd events, for example in festivals, mass emergency evacuations, sporting events, and transport hubs.

Extensive research in to crowd psychology has shown that self-categorisation theory can explain the behaviour of numerous crowds who have one common attribute: they are ‘psychological’ crowds who share a group identity, a sense of ‘we’ amongst the crowd members.

However, a systematic review of the crowd modelling literature found that until now self-categorisation theory has been neglected in crowd modelling (Templeton, Drury, & Philippides, 2015). Modellers have either treated crowds as homogeneous masses of identical members where there are no individual differences, or as consisting entirely of individuals where there are no groups at all, or place small groups of two to five people within the crowd. While these computer models can simulate physical crowds, they cannot convey the behaviour of psychological crowds where an entire crowd can act as one group.


Our research

Our research, funded by the Engineering and Physical Sciences Research Council, takes the first step towards placing group identity into computer models of crowds to develop a model which can better simulate the behaviour of crowds who share a group identity.

By doing this we hope to contribute knowledge from social psychology to the areas of crowd modelling and crowd safety planning, with the intention increasing the safety of psychological crowds at mass events. 

Through collaboration with  team of computer modellers at the Munich University of Applied Sciences, we have created the first computer model which examines the effect of group identity on evacuation behaviour (von Sivers, Templeton, Köster, Drury, & Philippides, 2014). Based on the research of Drury, Cocking, and Reicher (2009) which found that in the immediate aftermath of 7 July 2005 London bombings the survivors remained in the tube carriages to support those who were injured, we replicated this behaviour by creating a model where people who share a group identity helped injured people to escape. We compared this to the results of a typical crowd model where individuals would evacuate alone and found that our results more accurately replicated the behaviour of a psychological crowd found in social psychological research.


The project

Our current project pinpoints key behavioural differences in walking patterns between physical and psychological crowds and uses these to create a computer model of collective behaviour.

To do this, we created and filmed a psychological crowd using identity manipulation techniques and compared the walking speed, walking distance, and density between the psychological crowd and a physical (control) crowd. We replicate these behaviours in a model which combines the Optimal Steps Model (a pedestrian movement model, see  and Köster [2012)), and an additional group identity level based on aspects of self-categorisation theory.

Through this research, we aim to create a working model of the collective behaviour of psychological crowds, which can be used by computer modellers and event safety professionals to better simulate and understand the behaviour of psychological crowds. 

Find out more:

  • Drury, J., Cocking, C., & Reicher, S. (2009).  The British Journal of Social Psychology, 48(3), 487–506.
  • Seitz, M. J., & Köster, G. (2012).  Physical Review E, 86, 046108. doi: http://dx.doi.org/10.1103/PhysRevE.86.046108
  • Templeton, A., Drury, J., & Philippides, A. (2015).  Review of General Psychology. http://dx.doi.org/10.1037/gpr0000032
  • Von Sivers, I., Templeton, A., Köster, G., Drury, J., & Philippides, A. (2014).  Transportation Research Procedia. 2,. 585-593.

You might also be interested in: