Comment A realy cost/benefit analysis would be nice... (Score 1) 304
Would it be too much to ask for them to explicitly discuss cost/benefit of something like this?
Example: Our car has some "smart" routine for detecting glare ice on the road. I don't know if it has ever been right - but there have been literally hundreds of false posltives over the years. Thankfully, it doesn't do anything but beep annoyingly.
Imagine if your car foes into full emergency braking, whenever it thinks an accident is imminent. What level of false positives is acceptable? What level of false negatives? How many accidents are statistlcally likely to be prevented? How many will be caused. Assuming a positive balance, what are the financial costs of building this system into all vehicles - and what is the resultant cost in dollars/life? These are the kinds of information that the Traffic Safety Administration ought to be publishing with their proposal.
If you look at the detailed report, they break the system into three parts. All together, they expect the system to prevent about 100 deaths per year (plus a larger number of injuries. There is a very brief discussion of false positives that arose in their test scenarios (e.g., in section 4.8.1.4), but absolutely no attempt to estimate the number of accidents caused by the system.
Consider how many rear-end accidents the average person has, over how many years and miles. Then figure the reliability - the number of false positives - that can be tolerated - the number is essentially zero. Achieving this will require extraordinarily reliable sensors and software, which willwill be technically difficult and financially costly. None of this is addresses in the report.