"Once is happenstance. Twice is coincidence. Three times is enemy action." It sounds like the problem we have here is that each individual incident is analyzed

"Once is happenstance. Twice is coincidence. Three times is enemy action." It sounds like the problem we have here is that each individual incident is analyzed and catalogued as an incident, so we never get to 'enemy action'. The most urgent issue to resolve is whose insurance is going to pay for it, which requires assigning blame, and in each case it's usually clear where the blame lies. This is all the system working entirely as expected and we can't really skip this step. What ought to happen next is that the incidents get compiled, trends turn up, and we change our designs according to those trends. The article is pretty light on details about the extent to which this isn't happening among traffic engineers. Requiring smaller vehicles or banning touchscreens has to come from Congress, so it's outside the control of any city's engineers. Anecdotally, the engineers in my own city seem pretty determined to consider the possibility that any given accident had a bad design as the root cause, and regularly fix and improve intersections and road markings according to what is happening. They're not perfect, but I don't get the sense that they're resting easy on the assurance that it's all user error either.