Traditional traffic engineering is grounded in the discipline of civil engineering and uses mathematical models to predict the flow of vehicles and, to a lesser extent, pedestrians. When the results of a traffic study are presented at a public meeting or in a report, there’s a precision to the analysis: People do this and that under existing conditions. People will do these other things under future conditions.
It’s all very crisp and scientific, describing transportation behavior in a neat and orderly way, as if we follow particular lines and make turns at designated areas like a real life game of PacMan.
But you know better, because you've been inside a shopping mall. Some people move slowly, and others are in a terrible rush. Most are polite, but some use their elbows or bags to get to the escalator first. The mobility habits of people inside a mall are similar to what you’d see on a busy street corner.
The ballet of the good city sidewalk never repeats itself from place to place, and in any one place is always replete with new improvisations. —Jane Jacobs
The good city sidewalk is a place where children play, neighbors converse, activists gather, and cultures mix. It’s where community bonds are strengthened.
The Jane Jacobs ballet isn’t choreographed with precise steps and predictable rhythms. It’s composed of spontaneous, improvised movements that mix in with the routine action. Unpredictability is part of a neighborhood’s charm and its social function.
Sophisticated traffic models focus primarily on efficiency and order, which inadvertently engineers out the disorderly but vital interactions that make public spaces lively and engaging. These interactions are the essence of what makes a place feel vibrant, fun, and safe.
From a satellite view, people appear to follow strict paths. Up close, it looks more like a middle school dance.
Traffic models can be helpful. Just because the profession favors high-speed motoring doesn’t mean the models are always junk. I can’t tell you how valuable it is for an engineer to be able to coordinate a system of traffic signals—hundreds of lights—to keep people moving through a network. Commuters, deliveries, trash pickup, emergency vehicles…they’re all stuck if a traffic engineer does a poor job.
I’ve been thinking about this recently after hearing urbanism advocates (i.e. my people) talk about traffic engineering as if the models are always wrong and the math is always junk. I get it. I started my career as a traffic engineer, and I write a lot about the hocus pocus approach to mobility analysis and the blatant fortune telling when analyzing traffic conditions 20 years from now.
I don’t know if you’ve seen The Naked Gun, but maybe it’ll help you understand traffic models.
It's like eating a spoonful of Drano. Sure, it'll clean you out, but it'll leave you hollow inside. —Lt. Frank Drebin, The Naked Gun
Models are useful when they’re applied correctly to the real life context. The humans who create the sidewalk ballet on neighborhood or downtown streets are the same humans who operate motor vehicles.
This should give some relief to engineers who are worried about AI taking their jobs. Transportation departments need human engineers who can interpret one street type from another. Human engineers who make different assumptions about a highway, a transit corridor, a plaza, and so on. Very few corridors operate like PacMan.
We need experts who can evaluate scientific treatments like left-turn calming, roundabouts, chicanes, high-visibility crosswalks, floating bus stops, flex posts, sign placement, safety camera placement, curb ramps, etc. And we need those experts to evaluate based on how humans move through space, using any and all modes of travel.
It can be very helpful to have models in the traffic engineering cabinet, just like it’s helpful to have a bottle of Drano under the sink.
What you have touch here is exactly my point in my paper which make an argument why we need to think about pedestrians like a quantum particle rather than deterministic classical traffic engineering. (https://dl.acm.org/doi/10.1145/3600100.3626348)