Thanks for the post! As I’ve mentioned, I assign your work to my students. You have inspired me to move my own content onto Substack, and I’ll circle back in some future post along the following lines…

I see a need to look at AVs from a different perspective. It’s the basic premise of scientific research that if you understand a system, you can predict its behavior. (Think of residuals and regression analysis.) Famously, any research performed on swans prior to the 19th century would come to the certain conclusion that “Swans are always white”, until the Dutch (?) explored Australia and found black swans. Systems are easily predictable if we simply ignore relevant facts in the surrounding environment.

Except for the “edge cases”.

That these keep appearing implies that the AV mfrs cannot conceive of Black Swans. Notably, the mfrs are unable to emulate human drivers’ cognitive ability in situational awareness (your dog leash example). Waymo is quick to discuss reaction times (girl falling from a scooter, braking before hitting a child in a school zone), but fail to discuss how to avoid the situation altogether. That AV mfrs attempt to address an infinite supply of “edge cases” implies that they cannot predict (and thus do not comprehend) the more systemic behavior of our auto-based transportation system.

Which brings us to Daniel Kahneman and the human brain’s reliance on easily available information, to the exclusion of objectively valid counterpoints. When asked a hard question, we instinctively choose to provide an answer to an easier, unstated, question.

I see mega-billion dollar business models based on solving the driving problems (reaction times) that technologist believe that they can solve, rather than the transportation solutions required by society.

Crashes lead to $1B judgements, eventually the money dries up, and we can attempt a more systemic solutions.