I suspect I've been rattling on a bit about why we should be using models of various systems lately. Certainly I think the use of models that are based on abstractions derived from observations of the real world are better than those based on abstractions derived from whatever the modeller thinks the world should be like. Primarily because people the former is likely to be a lot more useful. We know that already though. So today, I'd like to draw your attention to a couple of articles about people modelling pedestrian dynamics. The Economist gives us the wisdom of crowds and Slate gives us sidewalk science, both worth reading. especially the Economist article.
It nicely demonstrates the perils of not paying attention to your assumptions. A lot of the early attempts at modelling crowds of pedestrians assumed that each pedestrian moved as a individual particle. Except, if you think about it, that's not always the case - people moving in groups (like tourists, groups of friends or even couples out for a stroll) don't act independently, they maintain a group cohesion which behaves significantly different from a number of individual particles.
Sidewalk science on the other hand takes a more story orientated approach, but illustrates again why it's good to pay attention to the system you're trying to model - different countries, even different cities have different behaviour. In individual cities people tend to break either left or right when two pedestrians approach each other head on as well as having pedestrians that move at different speeds in different environments (Germans move slower than Indians when in a crowd apparently) - models aren't necessarily immediately transferable to all situations.
So? Well, town planning for one can use this sort of thing. Architects trying to design escape routes from buildings. Stadium planners. There's all sorts of uses for this sort of modelling. The actual models are getting better but will always be dependent on the validity of the assumptions used to make them.