self-driving cars are also much worse at handling the ambiguity of mixed vehicle-human spaces. They work best when the infrastructure is specifically dedicated to them - a place designed for cars.
This is definitely a huge challenge, but the progress that Waymo is making (stark contrast to Tesla or Cruise) is very encouraging. An attentive driver can definitely handle the most confusing situations better than the current self-driving vehicles currently, but attentive and courteous drivers are rare in the US.
You can actually design roads to encourage attentiveness - by using road trees, narrower lanes, roundabouts, non-straight roads, or obstacles like bollards or similar. These aren't popular though because people want generally want to zone out and move fast.
Courteousness is a bit more difficult to address with just road / public space design, but designing to encourage more driver / non-driver interaction can help.
edit: a very non-american way to force people to pay attention on country roads is to make them BARELY wide enough to accommodate two cars. This forces attentiveness, because you don't want a head-on collision. Sounds problematic, but actually kinda nice in practice.
Agreed. In their quest for efficiency, American roads have had some huge unexpected externalities. I love how the central areas of Europe chaotically blend pedestrian, car, and bicycle traffic which causes everyone to be more attentive and careful. In the US, our cities were built in the era of cars and not cows so conversion is cost prohibitive. For the US, I think "smart endpoints and dump pipes" (where self-driving cars are the endpoints, and existing roads the pipes) is the more promising way to a better future. https://docs.aws.amazon.com/whitepapers/latest/running-conta...
