(May 28, 2014 at 8:55 pm)MindForgedManacle Wrote: I certainly wouldn't want to be part of the dev team that has to program the cars' ethical guidelines, being the minefield that it will necessarily be. After all, when the child/adult(/animal?) darts in front of the car, what will be the general protocols the cars abide by? Will they be fundamentally deontological, where the car will essentially try and avoid the death or injury of its occupants, regardless of the potential consequences to others? Or will they fundamentally be consequentialistic, where the cars' A.I. will weigh the overall harm done (perhaps it'd, say, swerve out of the kids' way even though it must hit an elderly person in order to avoid the kid? To go further, if the cars' A.I. is forced into a situation where the most optimal consequentialistic route is to crash itself to save more lives overall, will the car opt to instead save its occupants who presumably bought the car under the expectation that it would take their lives most seriously, or will it be a true consequentialist?
This is the sort of technology where I really hope Google is taking on numerous ethicists as consultants in order to help them take such considerations to heart. They are going to be able to parse out the relevant ethical space and the hazards/benefits than any non-specialist is likely to even be able to attempt.
All that said, I think self-driving cars will be an overall net boon for us on multiple fronts once it's fully implemented (how to deal with legacy technology is another concern that comes to my mind). Aside from essentially wiping out the scourge of drunk driving accidents (assuming these don't get released with optional manual control), we are almost certain to see traffic congestion go waaaayyy down, given that these will presumably be able to coordinate so as to provide a very efficient system, and be less prone to car accidents in general. And while the ethics of a car choosing how to act when lives are at stake is a minefield, it's not like it isn't a problem with regular ol' people. Heck, when we swerve out of the way, we aren't even capable of as much deliberate consideration as these will be capable of in such little time.
This is a good and insightful post. I never thought about driverless cars from this perspective. Thanks.
I remember one time driving down a highway. A deer pops out onto the side of the road slightly in the right of way. I swerved just a little because I decided that losing control of the car would be worse than hitting the deer. I clipped the deer. If it was a child, I hope I would have swerved more even if it meant loosing control of the car. I don't think Google-Car can make decisions like this......but then I am not entirely sure that I always can.
