Autonomous cars are going to bring to a head some interesting ethical issues — including the idea that people agree broadly on ethical concerns, but tend to differ — a lot — when it comes to specific circumstances.
With human drivers, we sort of give this a pass. In the heat of the moment, if a person makes the wrong decision then we tend to forgive them. We know that people make mistakes, make erroneous snap judgments. If I swerve to avoid the dog in the road that I see, but don't realize I'm turning into a couple of children, or off of a cliff, people will tut-tut, feel sorry for me, but not ascribe any particular malice or ethical flaw.
But if a computer driving a car does that, or decides that going over the cliff (killing the driver) is better than hitting a kid in the road, or that swerving to avoid a kid is okay even if it kills an elderly person — or any other such permutation — then people are going to get into some very sharp arguments (and law suits) over those rules.
It's going to be fascinating to watch — but not much fun to be involved in.
Of Cats And Cliffs: The Ethical Dilemmas Of The Driverless Car
We make decisions every day based on risk – perhaps running across a road to catch a bus if the road is quiet, but not if it’s busy. Sometimes these decisions must be made in an instant, in the face of dire circumstances: a child runs out in front of your car, but there are other dangers to either side, say a cat and a cliff. How do you decide? Do you risk your own safety to protect that of others?
I made the indisputable argument that officially wins this entire quandary months ago. I suppose I need to retype it and post it later.
+Yonatan Zunger had some good analysis of this a little while ago, too. https://plus.google.com/+DaveHill47/posts/Lvnqthb9i8G, arguing that the issue is that it actually forces us (programmers and car owners alike) to understand and own the ethical decisions to be made.
Maybe, if it can be distilled down enough, we need to have autonomous car owners sign a release that they understand the ethical settings. Maybe we need to have settings that drivers will choose (X cats = Y dogs = Z children = N grannies = M people in the car being driven), and thus be held accountable for. Or maybe we just wait for the legislators to do it (ugh).
I think the article makes the right arguments but doesn't put them together. Any acceptable system is going to want to minimize accidents. And since automation is going to have less accidents than manual, you want a high opt in rate for automation. That means the preferences of the operator must come first. This is the utilitarian argument. There are other arguments that reach this conclusion as well (ie: the operator has done his utmost to minimize risks, it's thereby wrong to have his car terminate him rather than the kids who didn't look both ways).
Ranking people is right out: violates everything our culture is founded upon. And where do they get off ranking cats higher than dogs, anyway?
I don't think that the ethical discussions that autonomous vehicles provoke mean they shouldn't be used (if they are, in fact, better at avoiding accidents in general). But they do force those discussions.
+Gary Roth BC our cat overlords want it that way.
+Dave Hill My point is that automation is so much safer that maximizing adoption of automating become paramount (ie: it outweighs almost everything else). As a consequence, almost any stance which would hinder adoption is itself unethical. This includes directing a vehicle to sacrifice its operator for anyone other than, perhaps, the closest family and friends.
+Gary Roth Ah. I see. It makes a certain measure of sense, but I doubt the public will be that rational.
+Dave Hill Just because the public favors something doesn't mean it's either just or intelligent. And all sensible variations on that.
+Gary Roth I would never suggest it did. 🙂 But ultimately the public is going to want to confront this issue, rationally or not.
+Dave Hill Then let's get out in front of the problem with the solution that doesn't lend itself to headlines like, "Your automated car is programmed to murder you to protect jaywalking idiots."
+Gary Roth Or "Your automated car will save your life — if you're willing to run over your kid."
Touche. With face recognition, we should totally let the user decide.