Crowd Dynamics: An Interview with Keith Still

Fifty-two-year-old mathematician Keith Still is the world’s leading expert on crowds. He is the G4S Professor of Crowd Dynamics & Crowd Management at the International Centre for Crowd Management and Security Studies (ICCMSS) at Buckinghamshire New University, and also lectures on critical infrastructure and emergency planning at the UK’s Cabinet Office Emergency Planning College. He has used his crowd modeling software and expertise to advise on crowd flow during the Hajj pilgrimage to Mecca and for the 2012 London Olympics. Christopher Turner spoke with him by phone.

Cabinet: In Crowds and Power (1960), Elias Canetti recalls that he was inspired to write his classic as a result of having been caught up in the stampeding masses during political violence in Austria in 1927. Was there a moment when you got interested in crowds?

Keith Still: It was at Wembley stadium in 1992, at a Freddie Mercury AIDS awareness concert. We were stuck in a queue for several hours, and people were getting very angry, but I thought it was fascinating. I went back to Wembley a couple of months later with permission to do research and perched myself above the players’ tunnel and spent many a happy weekend watching crowds at football matches.

People had a fluid model of crowd dynamics but I discovered that’s not how crowds move at all. If you imagine grains of sand in an egg timer, the fastest fluid moves down the middle. Crowds tend to move faster along the edges—you think about any queuing system, people come in from the left and right. That’s of course because, unlike grains of sand, people have choice and there are fewer interactions on the side. We don’t have the friction that grains of sand in an egg timer do.

How has the simulation tool you invented, Myriad 2, changed crowd modeling?

We amalgamated three very mathematical tools in one. We have network analysis, which looks at flow rate, the minimum or maximum efficiency of moving large quantities of people from one area to the next. We have a spatial analysis tool which looks at areas of dead space, best-utilized space, lines of sight, what’s visible, what’s not visible. And then, finally, an agent-based algorithm, which adds in various behavioral and psychological models. It’s the unique combination of having three very different mathematical disciplines, and the ability to put psychological and behavioral elements into the code at various levels, that makes the modeling that I do a little bit different than just network, spatial, or behavioral analysis.