The sensors themselves are imperfect, the thing you actually measure is probably the thing easiest to measure rather than actually pushing the tire grip toward its limits, even if you actually measure the right thing (hard acceleration/deceleration) you miss the fact that even an ordinary trip around town is often safer with more acceleration than a naive model would predict (e.g., short on ramps, or just that your speed when close to merging should probably be a bit higher than your intended merge speed in most cars), and actions like hard braking can rightly indicate appropriately slowing down for unexpected construction.
Misaligned incentives aren't something to be trifled with (something something Goodhart's Law). Maybe cutting all the hard-braking at every stop sign would be worth it, but if even a few people chose to merge more slowly onto the freeway that would greatly increase accidents and have ripple effects in rush hours and traffic patterns.
The sensors themselves are imperfect, the thing you actually measure is probably the thing easiest to measure rather than actually pushing the tire grip toward its limits, even if you actually measure the right thing (hard acceleration/deceleration) you miss the fact that even an ordinary trip around town is often safer with more acceleration than a naive model would predict (e.g., short on ramps, or just that your speed when close to merging should probably be a bit higher than your intended merge speed in most cars), and actions like hard braking can rightly indicate appropriately slowing down for unexpected construction.
Misaligned incentives aren't something to be trifled with (something something Goodhart's Law). Maybe cutting all the hard-braking at every stop sign would be worth it, but if even a few people chose to merge more slowly onto the freeway that would greatly increase accidents and have ripple effects in rush hours and traffic patterns.