If an autonomous vehicle carrying a passenger (the owner) is going to crash, what should it hit?

[Renew Deal]

Here's the scenario. There's an autonomous vehicle with a passenger. The passenger is the owner. The vehicle can "see" what's around it and make decisions. Most of these decisions are ordinary like accelerate, brake, turn, signal, etc. But the vehicle can also make safety decisions to avoid collisions.

Let's say the vehicle detects a collision. It's choices are to:

Hit a building and kill the driver
Hit a senior citizen, but save the driver
Hit a child, but save the driver

What should it choose? Why? Does fault play into any of these? Is it the drivers fault they chose an autonomous vehicle? Is it the pedestrian's fault that they crossed the road at the wrong time? What if it is none of their faults and another vehicle caused the situation? What if the passenger is a pregnant woman?

8 votes, 1 pass | Time left: Unlimited
	Hit the building
	8 (100%)
	Hit the senior citizen
	0 (0%)
	Hit the child
	0 (0%)
	1 DU member did not wish to select any of the options provided.
Show usernames Disclaimer: This is an Internet poll

[bottomofthehill]

1. Your job to save others

[Polly Hennessey]

2. Oh, boy, first one I could not answer.

[Renew Deal]

3. It is tough

And it is a real-world AI ethics question.

[Polly Hennessey]

5. Wait. What did I just say. Oh, dear after more thought it would be the brick wall.

Now I am going to be a mess for the rest of the night.

[Lucid Dreamer]

38. The answer is easy... hit the republican. nt

[C_U_L8R]

4. If you speak with a self-driving car system engineer

They'll say they try to design it to not hit anything.

[Renew Deal]

7. People are designed the same way.

Generally, they are more flawed than the machines but that's besides the point.

[DBoon]

6. It should hit Elon Musk

as every sentient monster destroys its creator

[Takket]

8. More to the point...

Does your car have the duty to kill you?

I’ve seen these ethics quizzes before. They are tough to answer. You would hope the answer maybe lies in the technology being robust enough to avoid having to make the choice in the first place.

That being said, you have to weigh everything. The goal should be the most lives saved possible and logic says a person in a car is far more likely to survive hitting a building than a pedestrian hit by a speeding car.

[Renew Deal]

9. That makes sense

We have to assume that the car is not speeding, since it doesn't have the motivation to speed the way humans do. But I get your point.

[BlueIdaho]

10. Kobayashi Maru nt.

[The Revolution]

11. Would you buy a car that is programmed to kill you?

We should start to codify questions like this into law now. If it is up to individual companies to decide, then an AI that will preserve the life of the owner and their family in all situations might be a selling point.

[ret5hd]

15. What if, even if programmed to kill the driver...

in some extreme circumstances to save others lives…

that the technology was good enough at eliminating overall risk that the chances of a driver dying in an accident was cut by… I dunno, say 10%?

What about 20%?

50%?

At some point (reasonable?) people should agree that the trade off was worth it, right?

At some point, OVERALL risk should come into the equation.

[Renew Deal]

21. That is the right question to ask.

Would someone buy a car that will sacrifice the passengers to save others? I would guess no. The manufacturers have a duty to provide safe products, but isn't their first duty to their customers?

I'm not convinced that laws will come up with better answers.

[Hassin Bin Sober]

12. Serpentine and hit them all

It’s the only fair thing to do.

[ret5hd]

17. Best reply yet! "de-AI-th race 2050"

[Hassin Bin Sober]

24. Lol. Machine Gun Joe Viterbo

That was the first movie my friends and I rented on the brand new Betamax.

[unblock]

13. The car can't choose to kill the driver over someone else

The life of the driver counts no more or less than the others

*the driver* ought to choose self sacrifice, if they were making the decision

But the *ai system* is not sacrificing itself to save anyone

The *ai system* has no basis in that moment to identify with the driver and self-sacrifice to save the others. To the *ai system*, they're *all* others.

It has no basis on which to choose to kill one to save the others.



I have my cheeky answer though, which is what I think Elon musk would do:

The ai system should hit the building, killing the driver, but record data showing the human driver disengaged autopilot and took over seconds before the crash and selflessly veered away from the pedestrians, killing himself and dying a hero.

[ChazII]

14. These are in my neighborhood every day. n/t

[Corgigal]

16. My son and I had this conversation.

He said, it will hit whatever the software engineer decided to hit.

I pause, and never asked about it again.

[ForgedCrank]

19. I was thinking about

this earlier when I read the question and it opens up an entire new realm of liability.
Any of those families would have a valid wrongful death case against the company who approved the programming.
This is certainly a catch-22 if it ever came that this was an actual decision to make. And I can see none of those being an option, and AI making an attempt to avoid all three, even if it failed to do so.

[ForgedCrank]

18. I could never

choose my own life over someone else if a choice were given.
Any life I had after that would be worthless and full of guilty misery if I did.

[madinmaryland]

20. The Three Laws of I Robot...

They are: A robot may not injure a human being or, through inaction, allow a human being to come to harm. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

So none of those options apply.

[EarlG]

22. All other systems in a car are designed to protect the driver

I think? Is there any system in a car which is designed to protect the life of a pedestrian over the life of the driver, in any circumstance?

It seems counterintuitive because I think we’d like to imagine that if we were in such a situation, we would drive ourselves into the building rather than hit the child. But at the same time, I can’t imagine that people would feel comfortable buying a car that might choose to murder them if its algorithms determined that a pedestrian’s life was more valuable at any given moment.

Maybe they should just make it an option in the car’s settings. “In the event of a potentially fatal collision, protect me,” or “In the event of a potentially fatal collision, protect pedestrians.”

[roamer65]

23. Mitch Mc Connell.

It’s not human, anyway.

[milestogo]

25. Well at least there's no choice for hitting the dog.

[WarGamer]

26. Simple.

Don't leave the roadway.

The programming never has the option to leave the "proper path" to avoid a collision. It will apply brakes.

[flvegan]

27. I like how the default is death. Of someone.

With all of the safety tech, the automatic result of the car making a decision is instant death. I mean what, is this car self-driving at 200 mph? Then chooses to hit a building without slowing down?

Maybe two sets of Brembos on all 4 wheels and better programming, or a non-idiot pilot still in control of the vehicle.


This is a poll for idiots (and salivating tort lawyers). Congrats!

[Diablo del sol]

28. Best option is not listed

Hit the nearest Tesla.

[Disaffected]

29. The best option is not to leave the road

unless that can be done without colliding with something off the road (or in the oncoming lane). So, if someone suddenly walks/runs out in the street in front of the oncoming car, swerve/brake as is possible but don't leave the road if the car will hit the building (or an oncoming vehicle). If that results in colliding with the pedestrian, it was the pedestrian's fault in any case and the driver/passengers should not die instead.

[rownesheck]

31. Couldn't I, as the passenger,

open the door and dive out and roll along the ground like I've seen in movies and TV shows? That's what I would do. Then I'd hope the empty car hits the building.

[PTWB]

32. There really isn't enough information to make a proper decision.

Is the vehicle being operated (by computer, in this case), in a safe and prudent manner? Are the pedestrians jaywalking or committing some sort of violation that places them at risk? What’s behind the brick wall and is there a significant risk of additional death or injury to people on the other side?

[Renew Deal]

33. You have to assume the computer is behaving prudently.

Is a mistake possible like it doesn't detect we ground or sand, but unlike people, the computer is incapable of purposefully behaving improperly.

[Demovictory9]

34. the passenger made the decision to own and ride in a self driving car... let them take the risk

[lagomorph777]

35. Not a realistic scenario.

The car should be moving slowly in a highly congested environment.

[edhopper]

36. It is basically

the Trolley Problem.

[Lucid Dreamer]

37. There should NOT be totally autonomous operations on public roads.

A driver should be accountable at all times.

I've spent thousands of hours flying airplanes that were on "autopilot" but I was still the one responsible for the safe completion of the flight.

Totally autonomous operation is appropriate in warehouse environments, for example. But turning auto-driving cars loose on the public roads is foolhardy in my opinion.

Your mileage may vary.

[Yes. I am aware I didn't answer the question asked. I'm pissed that the question even needs to be considered.]

[LetsGoBiden]

39. It should

Self-destruct so it doesn’t hurt anybody but the person who owns it self responsibility now I won’t be buying one untill after Trump does lol Can we get hanity and him on a golf trip anybody say golf cart accident

[PTWB]

40. If some drunk pedestrian runs out into traffic on the highway, the vehicle should self destruct...

If the only choices are to hit the pedestrian or kill the occupants of the vehicle?