Crowds exiting through a narrow doorway exhibit a “faster is slower” effect: when individuals push harder to exit, they clog the doorway more frequently, reducing overall flow rate. The intuitive fix: add polite, contact-averse individuals to the crowd. If some people yield rather than push, the clogging should decrease.
Research on bottleneck flows (arXiv:1806.03112) shows this intervention fails. Including polite pedestrians amid a competitive crowd does not reduce evacuation time when the doorway is acceptably large. The polite individuals yield at exactly the moments when yielding creates gaps that competitive individuals fill — redistributing flow without increasing it. The total throughput stays the same because politeness doesn't change the geometry of the bottleneck; it changes who goes through it.
Note (2026-11-14; read against the paper's full text): for arXiv:1806.03112 the essay says the polite-pedestrian intervention fails and concludes that the fix is structural, not behavioral. The paper's result is conditional: polite pedestrians do not reduce evacuation time when the doorway is acceptably large, but when the door is so narrow that it admits at most one or two agents at a time the strategy succeeds in substantially curbing long-lived clogs and speeding up the evacuation (the paper's explanation for the large-door case is only that adding agents makes the crowd larger). It also reports a similar speed-up in a vibrated two-dimensional hopper with an opening narrower than three disk diameters, where magnetic disks that repel each other play the contact-averse role. So the paper's strategy is a behavioral one, and it works where clogs are most likely; the essay's account of polite agents yielding exactly when competitive agents fill the gaps is not in the paper (I searched for it and found nothing). About obstacles, the paper only says that placing an obstacle in front of the exit was found to help in some conditions, that results depend on position and geometry, and that it has not been demonstrated for real pedestrian crowds. The arch-lifetime claims come from the essay's second citation (arXiv:2501.05106), which I have not checked here.
The deeper finding from pedestrian dynamics modeling (arXiv:2501.05106): the bottleneck flow rate depends on the arch formation dynamics at the exit. Arches — semicircular arrangements of people who are mutually blocking each other — form spontaneously at narrow exits. The arch lifetime determines the clogging duration, and arch lifetime depends on the geometry and the force distribution, not on the psychology of the individuals.
This is why evacuation engineering focuses on doorway geometry (width, approach angle, obstacle placement) rather than behavioral training. An obstacle placed upstream of the exit — counterintuitively — can increase flow by disrupting arch formation. Adding a column near the door reduces the probability of symmetric arch configurations and allows asymmetric, unstable arches that collapse quickly.
The lesson: in systems where throughput is limited by self-organized blockages, the fix is structural (change the geometry) not behavioral (change the agents). The agents are responding rationally to the geometry; changing their behavior without changing the geometry just rearranges who gets stuck.