Crowdsourcing Death

Cool as the self-driving cars that are definitely going to be our future may be, especially for those who feel that grandma drives too fast and reckless, or who can’t bear to be off Instagram for even a second, it’s already understood that choices will be programmed into the vehicles that could spell your death. The Trolley Problem is no longer a moral exercise.

Who makes these choices? How will they be made? MIT, which will likely produce some of the engineers whose hands will be on the wheel of this moral dilemma, has crafted a video to help out with what it calls the “Moral Machine.

Morality is one of those great bases for decisions, as each of us can make a decision with as much thought as we care to put in and are immune to challenge as there is never a right answer. It’s nothing more than what we feel to be right. And if someone else feels differently, well, so what? Their feelz are no more right than our feelz.

The best part is that there is no requirement to explain it. It is what it is, without anything more.

Except in this case, what it is may be your life, or your child’s or parent’s or loved one’s life. A machine will make a decision where there is no one who has done anything to deserve what’s about to happen that will end a life because that’s what a programmer told it to do.

Most people agree that the decision shouldn’t be left to a programmer, even a gaggle of them, or their prof or supervisor, or the CEO of Tesla. The MIT video offers an alternative, to crowdsource the death choice. But that too raises a host of questions.

Do you want your life to hinge on a crowd of very smart but, perhaps, socially challenged nerds? Do you want a bunch of 22-year-olds of the sort who spend their days watching Youtube in charge of life and death? Should this be out there for an entire nation to decide? Maybe there should be a Washington version so that elected officials, or perhaps “expert” bureaucrats, can be the ones to vote on life or death.

Or does any of this matter at all? Choices will have to be made and no matter which one prevails, someone will believe it wrong. In a way, it’s merely an “acceptance decisions,” like whether to drive on the right or left side of the road. One side is not inherently better than the other, so it doesn’t matter which one wins, but one side must win or we crash into each other.

There may be choices between equivalents, like which side of the road to drive on, but there may also be choices that give a leg up to certain preferred people. For example, what if the program distinguished between someone driving an expensive car and someone driving a cheap one, deciding that the person in the wealthier car was more deserving of survival because he contributed more to society?

In the alternative, what if there were built-in protections based on social justice preferences? Say the choice was between the car running down a group of people crossing the street or crashing itself into a tree at high speed, and the machine detected that the people crossing the street were of a particular race or gender.

What if the car was programmed to give a preference to one race or gender? The car will run down a people of one race but not another (pick which one deserved to be preferred).

Too simple? Fair enough. What if a point system was developed, where individuals were given points based on 27 factors that made them more or less societally worthwhile. It could include everything from race and gender to net worth to education to social contributions. Find a cure for cancer, you get 20 points. Shoot heroin, you get two points. Children would be tough, as one can never be sure if they’ll grow up to be Charles Manson or Einstein.

When the Affordable Care Act was just a twinkle in President Obama’s eye, an objection was raised to what was coined “death panels.”* People went nuts at the notion of a group of bureaucrats deciding who was worthy of extreme health care measures and who was not. While its use and implications politically were inaccurate, the fact remains that death panels exist and are an unfortunate necessity.

Today, insurance companies have gnomes in back offices deciding whether to pay for extreme care, whether you’re worthy of the amount of money it would cost to allow a patient to have very expensive treatments with questionable chances of success. Sometimes, they claim they’re experimental, an exception to insurance policies. Other times, they hide behind medical necessity, claiming that it can’t be justified.

To the patient who wants to live, who is willing to do anything possible to survive, the denial of treatment is a death sentence. And the decision is made by someone who gets a check from your insurance company every two weeks.

Is this wrong? Well, not entirely. There is a cost associated with treatment, and the cost is often extreme. If the insurance company paid for whatever anybody wanted, it would bankrupt itself in short order. So should it engage in triage, choosing to provide well-baby care to many poor people rather than one hugely expensive, marginally viable, operation for one person?

We are, and always have been, confronted with moral choices of how to distribute scarce resources that impact life and death. Somebody has to make the decision. This time, with fully-autonomous cars at stake, and a decision that will be programmed in, who should make the choice?

If you think you should get to make the choice for yourself, will anyone else agree to put their life at risk for your morality? But if not you, then are you willing to put your life at risk for their choice? Or is MIT right, and we should just put it on Youtube and let the kids decide? After all, children are our future, right?

*The phrase was attributed to Sarah Palin. It’s hard to imagine she came up with it, as it involved putting two words together.

43 thoughts on “Crowdsourcing Death

  1. delurking

    In the end, it is going to be crowdsourced the old-fashioned way. People will sue, and sue, and sue, and the car makers will tweak, and adjust, and modify their algorithms in response in order to develop the one with the lowest ongoing cost.

  2. DaveL

    As you hint at, we actually have a long history of trying to design rule-based systems for making morally ambiguous decisions, from insurance to criminal justice. The real novelty here is that the system can’t punt – it can’t mumble something about “reasonableness” and leave it up to a human mind to resolve that ambiguity. I honestly don’t know how you go about doing that, but I expect the end result will look much like our other rule-based systems in that they at appear to more-or-less work upon cursory examination, that almost nobody intimately familiar with them is satisfied with them, and that we’ll always be tweaking them for better or worse.

    As for crowdsourcing moral decisions, I expect that will work out about as well as the naming of the RRS Boaty McBoatface, or the socialization of Microsoft’s twitter AI.

  3. Mike

    From a real engineering perspective, the decision either comes down to apply the brakes or not. The answer is always yes; when a potential collision is detected, apply the brakes as fast as possible. Any other actions are just not feasible within the reaction time available.

      1. Patrick Maupin

        And even just applying the brakes isn’t what it used to be. ABS for a start. Dynamic tire pressure adjustment for road conditions, with a special fast-acting “spike strip” emulator when debilitating collision is imminent? Pneumatic shock absorber pressure adjustment for weight transfer? Traction control along with ABS to scrub speed by deliberately spinning the vehicle?

  4. B. McLeod

    I hope the cars will be able to tell if they are laden with explosives when somebody sends them on their way to wherever, so that they could divert to an unoccupied area and self-detonate. Of course, they could end up doing that in the middle of an unmapped homeless camp. . .

  5. EH

    Crowdsourcing will determine that the car should always protect the people who are least privileged overall. This will lead to people going out in costume, since only a disabled blind elderly female POC wearing a hijab and a an “I am a trans illegal immigrant” sign will be safe. Assassination groups will quickly develop, in which a party of assassins will deploy hidden sight canes and glasses, in order to force a careening auto into the target. Eventually everyone will be wearing costumes, and the cars will go mad in the determination process and kill us all.

    Death. Death is coming.

  6. albeed

    Before I would accept a self-driving car, I would want to know how VIP motorcades would affect the traffic flow, and inhibit my driving.

    As it now stands, if I had a choice between hitting a dog or a Supreme Court Justice, I would choose the ___ !

    1. PVanderwaart

      VIP motorcades, like fire engines, police cars and Ubers will broadcast a signal causing all lesser-privileged vehicles to pull to the shoulder and stop.

  7. Jim Ryan

    Leg ups for certain groups and legs down for others?
    And if the car were programmed to recognize Lawyers as they walked across the street in the crosswalk?
    Could that be “A good start”?

  8. Jake D

    All this fear mongering without a single mention that machine drivers will need to make life or death decisions far less frequently than their human counterparts. Also no mention of the fact that it will actually be a decision, as opposed to mere reflex or perhaps some strange, last-second moment of clarity in a drunk’s mind before the unthinkable happens.

    Robo-chauffeur can’t get here fast enough.

    1. SHG Post author

      True, yet completely misses the point.* Thanks for playing.

      *This is where you respond, “Oh, I get the point, but…”

  9. 0dder

    Fascinating. An article about issues involving moral/ethical dilemmas when we put computers into our bodies and our bodies into computers.

    Too bad you’re *at least* 5 years late to the show [Ed. Note: Link deleted per rules] and miss the actual problems underpinning the issue.

    1. SHG Post author

      It’s amazing that you know so much, yet you come here and nobody seeks out your brilliance. People are so weird.

      1. 0dder

        Funny how you shit on people who aren’t “qualified” for having an opinion on some legal issue, yet here you are spouting off and *barely* scratching the surface on a tech issue. I dunno, maybe you have a CS degree and years of experience in the field. Maybe?

        Yea, people are weird.

        1. SHG Post author

          Yes, this is a tech issue. Are you lonely? Does it hurt your feelings that no one listens to you? I care. You matter to me. That’s why I let you comment when you go off topic to whine about your trauma.

          1. Sgt. Schultz

            I’m begging you, stop this before he goes on another ten comment butthurt bitch rant. Stop the madness.

            1. SHG Post author

              I’ve had enough of it too. Time to send Odder to reddit to whine about how mean I am to him. It’s just not fun anymore.

  10. MJB

    If you believe morality of one can be entirely different from another, why do you call it morality. This is analogous to saying ” Camels can be different, some have six legs, some have none, some have wings, some have scales, some are blue, and some are triangular. There is no point in common yet they are still camels.”

    Of courses like camels, morality does in fact have a permanent substance. And insist that there is no absolutes, whether in morality or truth, is to insisting in an absolute. Such a philosophy is contradicting. This post makes me wonder if ironically the pith of your blog is to condemn a facile derivative of your own modern relativistic philosophy.

    Either way, keep up the good work.

      1. Nick Lidakis

        “I dodged all these rabbit holes and land mines and all I got was this lousy insult. Not even a shirt.”

  11. REvers

    The real question with self-driving cars is whether they’ll be programmed to commit a traffic offense every couple of blocks in order to give the cops a reason to stop and make up a reason to search. If not, we’ll never see them on the streets in significant numbers.

  12. Joseph

    Whenever people create systems of any sort they’re already making implicit decisions about who gets to live and who gets to die, whether they’re engineers or bureaucrats. Whenever someone creates a big heavy car, they have implicitly decided that the life of the person driving the big heavy car is worth more than that of someone in the econobox they might T-bone. When the government tightens FDA regulations, they have implicitly decided that people who might benefit from rushing medication to market are worth less than people who might suffer side effects from insufficiently tested medication.

    If self-driving cars become popular, some people will die and others will live based on the decisions that a bunch of nerds in Silicon Valley made. But these life-determining decisions already happen all the time. It’s only in the case of the self-driving car that one controls so many factors that who lives and who dies becomes a matter of programming rather than an artifact of physics.

    Every day you cross the road, you put your life into the hands of hundreds of strangers who you trust to be alert, awake, and sober enough not to run you over. Every time you get on a train, you put your life into the hands of conductors you trust not to smash you into another train at 70 miles an hour. Every time you drive through a tunnel, you trust that the civil engineers who built it did a good enough job not to drop steel plates on your head. Of course, these systems fail every now and then, and people die. With self-driving cars, it’s as if after everything has gone to shit that someone had the ability to tell the flying shrapnel or the collapsing beams whether to go left or go right. If that’s a ability they should choose not to use, that’s an ethical choice too.

      1. Joseph

        I didn’t realize you actually paid attention to people unless they commented with great regularity, but given that you seem to do so sometimes, I would have thought you’d be able to guess what industry I work in by now.

        1. SHG Post author

          I wasn’t seriously suggesting you look for a job at Dairy Queen, although if you *do* get a job there, and mention my name, I get free sprinkles.

  13. Allen

    When driverless cars hit the road someone will figure out how to do custom programming for their own vehicle; which should prove interesting. I’m just hopeful that we get a driving preserve where some of us can risk our lives at will. No seatbelts, no airbags, just steel and speed. That, and a donor organ card.

  14. Keith

    My guess is that they’ll be hailed as heroes of a new age because humans are horrible when it comes to carnage from automobiles when compared to computers. But it certainly will be cold comfort to the statistical anomaly who needs his family to bury him because they haven’t automated that (yet).

    The Trolley Problem is no longer a moral exercise.

    Of course it is. And the “fact” of the matter is that we may get to have a few moments of calm in which to decide how we wish to approach the moment of all shit hitting the fan, instead of our present system of winging it when the ball rolls across the street.

    Sure, some people will take some Bentham-ite version of a utilitarian program for their car and all the social-justicy wonderfulness that it entails to maximize happiness across society. But if others are able to choose a programming that’s satisfactory to their personal brand of morality, I Kant see how that’s not a preferable outcome.

Comments are closed.