Traffic congestion is a major problem all across the US (and many other parts of the world), and some of these problems are caused by problematic driving patterns. Many of these patterns could likely be corrected if they are identified in detail. As an example, zipper merging can reduce jamming when a lane ends but many people in my area prefer to cut people off, and slam the accelerator instead. The way I see things going is that in the future, BeamNG will be able to realistically model traffic flow once the AI system is up to par (even if it is only mods, or BeamNG.Research that do this). There is an opportunity for learning how to reduce congestion using flow patterns which will be able to be simulated in BeamNG. The only question is... will it be taken advantage of or just ignored?
I dont see the connection between this and BeamNG as a game. Models to predict and simulate traffic flow have existed for years, on an incredibly complex scale that is way beyond what is needed for a video game. If anything, Beam could use data from those models as a guideline to make its own AI traffic more realistic, with Beam's whole concept of being as lifelike as possible.