Driverless Dilemma

Hey.

You're listening to Radiolab from WNYC Three One.

I'm jad abumrad.

I'm Robert Krulwich and you know what this is.

This is Radiolab. Okay, so we're going to play you a little bit of tape first just to set up what we're going to do today. About a month ago, we were doing the thing about the fake news.

Yeah, we were worried about a lot of fake news a lot of people are, but in the middle of doing that reporting, we were talking with a fellow from Vanity Fair.

My name is Nick Bilton, I'm a special correspondent for Vanity Fair.

And in the course of our conversation, Nick, and this had nothing to do with what we were talking about. By the way, Nick just got into a sort of a well, he went into a kind of a nervous reverie, I'd say.

Yeah, he was like, you guys want to talk about fake news, but that's not actually what's eating at me.

The thing that I've been pretty obsessed with lately is actually not fake news, but it's automation and artificial intelligence, and driverless cars. Because it's going to have a larger effect on society than any technology that I think has ever been created in the history of mankind. I know that's kind of a bold statement, but you've got to imagine that. There will be in the next 10 years, 20 to 50 million jobs that will just vanish to automation. You've got a million truckers that will lose their jobs.

My blood pressure has gone up. So that's a fairly compelling description of a very dangerous future.

Yes, but you know what, it's funny, one of the things that I mean. We couldn't use that tape initially at least, but we kept thinking about it because it actually weirdly points us back to a story we did about a decade ago.

It may be that what Nick is worried about, and what we were worried about 10 years ago have now come dangerously close together.

So what we thought we'd do is we're going to play you the story as we did it. Then, sort of the full segment, and then we're going to amend it on the back end. And by way of just disclaiming, this was at a moment in our development where there's just like, way too many sound effects, just gratuitous.

Well, you don't have to apologize.

No, I'm going to apologize because it was just too much, just too much. And also like, we talk about the FMRI machine, like, it's this amazing thing when it was, it's sort of commonplace now. Anyhow. It doesn't matter, we're going to play it for you and then talk about it.

Okay.

There are five workers on the tracks working, they've got their backs turned to the trolley, which is coming in the distance.

You mean, they're repairing the track?

They are repairing the track.

This is unbeknownst to them. The trolley is approaching.

They don't see it, you can't shout to them, okay, and if you do nothing, here's what'll happen.

Oh my God, that was a horrible experience. I don't want that to happen. No, you don't.

But you have a choice you can do a, nothing or B. It so happens. Next to you is a lever, pull the lever, and the trolley will jump onto some side tracks where there is only one person.

Working.

Yeah, so there's your choice do you kill one man by pulling a lever? or do you kill five men by doing nothing?

Well, I'm going to pull the lever.

Naturally, all right, here's part two. You're standing near some train tracks. Five guys are in the tracks, just as before.

Same five guys working on the track.

Same five guys.

Back to the train, they can't see anything.

Yeah, yeah, exactly. However, I'm going to make a couple of changes now. You're standing on a footbridge that passes over the tracks, you're looking down onto the tracks.

What do you mean? There's a guy?

A large guy, a large guy, a large individual, standing next to you on the bridge, looking down with you over the tracks. And you realize, Wait, I can save those five workers. If I push this man, give him a little tap.

No, no.

No, no, no, no, no.

No, no, no, no.

No, right. Oh yeah, I'm not gonna do that, I'm not gonna do that. But surely you realize the math is the same?

All right, here's the thing if you ask people these questions, and we did, starting with the first. Is it okay to kill one man to save five using a lever? Nine out of ten people will say Yes, yes, yeah, but if you ask them, is it okay to kill one man to save five by pushing the guy?

And if having a moral sense is a unique and special human quality, then maybe we, we, us two humans anyway. You and me should at least inquire as to why this happens. And I happen to have met somebody who has a hunch.

How do people make this judgment, forget whether or not these judgments are right or wrong? Just what's going on in the brain that makes people distinguish so naturally and intuitively between these two cases, which, from an actuarial point of view, are very, very, very similar. If not I.

Josh is, by the way, a philosopher and a neuroscientist, so this gives him special powers. He doesn't sort of sit back in a chair, smoke a pipe and think, now why do you have these differences? He said. No, I would like to look inside people's heads. Because in our heads we may find clues as to where these feelings of revulsion or acceptance come from in our brains. So we're here in the control room.

Brain scanner. I'll tell you a funny story. You can't have any metal in there because the magnet. So we have this long list of questions that we ask people to make sure they can go in. Do you have a pacemaker?

Anyhow, what Josh does is he invites people into this room, has them lie down and what is essentially a cot on rollers. And he rolls them into the machine. Their heads are braced, so they're sort of stuck in there. You ever done this?

Oh, yeah, yeah, several times, and then he tells them stories.

He tells them the same two, you know, Charlie, tales that you told before. And then, at the very instant, that they're deciding whether I should push the lever or whether I should push the man at that instant. The scanner snaps pictures of their brains. And what he found in those pictures was, frankly, a little startling. He showed us some.

I'll show you some, some, some stuff, okay?

Think the picture that I'm looking at is the sort of a, it's, it's a brain looked, I guess, from the top down.

Yeah, it's top down and sort of sliced, you know?

Like, like a like, a deli slicer. And the first slide that he showed me was a human brain. Being asked the question, Would you pull the lever? and the answer in most cases, yes, yeah, I pull the level when the brain saying yes. You'd see little kind of peanut shaped spots of yellow. This little guy right here and these two guys right there. The brain was being active in these places, and, oddly enough, whenever people said yes to the lever question, the very same pattern lit up.

So when people answer yes to the Lever question, there are there are places in their brain which glow, but when they answer no.

I will not push the man. Then you get a completely different part of the brain lighting up.

Even though the questions are basically the same, mm-hmm. What does that mean? and what does Josh make of this?

Well, he has a theory about this, a theory not proven, but I think that this is what I think the evidence suggests. He suggests that the human brain doesn't hum along like one big unified system. And so, he says, maybe in your brain, in every brain, you'll find little warring tribes, little subgroups. One that is sort of doing a logical, sort of counting kind of thing.

You've got one part of the brain that says, huh, five lives versus one life?

Wouldn't it be better to say five versus one? And that's the part that would glow when you answer yes, I'd pull the lever, Yeah, pull the lever. But there's this other part of the brain which really, really doesn't like personally killing another human being. And gets very upset at the fat man case and shouts in effect.

You know, it understands it on that level and says, No, no bad, don't do.

This is not a trivial discovery that you struggle to find right and wrong, depending upon what part of your brain is shouting. The loudest. This is. It's like bleachers morality.

Do you buy this? hmm.

You know, I just don't know, yeah, I always kind of suspected that. A sense of right and wrong is mostly stuff that you get from your mom and your dad, and from experience that it's culturally learned. For the most part, Josh is kind of a radical in this respect. He thinks it's biological, I mean, deeply biological, that somehow we inherit from the deep past a sense of right and wrong.

Primate ancestors, before we were full-blown humans, had intensely social lives. They have social mechanisms that prevent them from doing all the nasty things that they might otherwise be interested in doing. And so deep in our brain, we have what you might call basic primate morality, and basic primate morality. Doesn't understand things like tax evasion, but it does understand things like pushing your buddy off of a cliff.

Oh, so, you're thinking them at the man on the bridge, that I'm on the bridge next to the large man. And then I have hundreds of, thousands of years of training in my brain. That says, don't murder the large man, right? Whereas even if I'm thinking, if I murder the large man, I'm gonna save five lives and only kill the woman. But there's something deeper down that says, don't murder the large man, right?

Now that case, I think it's a pretty easy case, even though it's five versus one. In that case, people just go with what we might call the inner chimp.

But there are other, but there are. Chimp is your unfortunate way of describing an act of good, of deep goodness. Well, that's, that's what's interesting, inner chimp, right?

Well, what's interesting is that we think of basic human morality as being handed down from on high, and it's probably better to say that it was handed up from below. That our most basic core moral values are not the things that we humans have invented, but the things that we've actually inherited from other people. The stuff that we humans have invented are the things that seem more.

Peripheral and variable, but something as basic as Thou shalt not kill, which many people think was handed down in tablet form. From a mountaintop, from God directly to humans. No chimps involved, right? You're suggesting that hundreds of thousands of years of on-the-ground training have gotten our brains to think, don't kill your kin, right?

You know, that should be your default response. I mean, certainly, chimpanzees are extremely violent and they do kill each other, but they don't do it as a matter. Of course, they, so to speak, have to have some context sensitive reason for doing so. So now we're getting to the rub of it.

You think that profound moral positions may be somehow embedded in brain chemistry?

Yeah, and Josh thinks. There are times when these different moral positions that we have embedded inside of us, in our brains, when they can come into conflict. And in the original episode, we went into one more story, this one you might call the Crying Baby dilemma.

The situation is is somewhat similar to the last episode of Mash, for people who are familiar with that. But the way we told the story, it goes like this It's wartime. There's an enemy patrol coming down the road. You are hiding in the basement with some of your fellow villagers.

Nobody make a sound until they've passed us.

So there you are, you're huddled in the basement all around your enemy troops, and you're holding your baby in your arms.

They hear your baby, they're gonna find you and the baby and everyone else, and they're gonna kill everybody. And the only way you can stop this from happening is cover the baby's mouth. But if you do that, the baby's going to smother and die. If you don't cover the baby's mouth, soldiers are gonna find everybody and everybody's gonna be killed, including you, including your baby.

And you have the choice would you smother your own baby to save the village, or would you let your baby cough, knowing the consequences?

And this is a very tough question, people take a long time to think about it and.

Some people say yes, and some people say no. Children are a blessing and a gift from God, and we do not do that children. Yes, I think I would kill my baby to save everyone else and myself. No, I would not kill a baby I feel, because it's my baby.

I'd like to say that I would kill the baby, but I don't know if I'd have the inner strength.

No, if it comes down to killing my own child, my own daughter and my own son, then I choose that. Yeah, if you have to, of course. It was done in World War Two, when the Germans were coming around. There was a mother that had a baby that was crying, and rather to be found, she actually suffocated the baby, but the other people lived.

In the final mash episode, the Korean woman who's a character in this piece?

My god. oh my God.

Just wanted it to be quiet.

It was a baby, she smothered around, baby.

Well, who breaks the time? I mean, they had to answer something right, that's a good question, and.

Now is there a do? What happens, is that just two cries that fight each other out, or is there a judge?

Well, that's an interesting question.

And that's one of the things that we're looking at. When you are in this moment, with parts of your brain contesting, there are two brain regions, these two areas here towards the front, right behind your eyebrows, left and right, that light up.

It's those sort of areas that are very highly developed in humans as compared to other species.

So when we have a problem that we need to deliberate over the light, the front of the brain. This is above my eyebrow. Sort of, yeah, right about there. And there's two of them, one on the left and one on the right, bilateral, and they are the things that monkeys don't have as much of that.

We have certainly these parts of the brain are more highly developed in humans.

Looking at these two flashes of light at the front of a human brain, you could say we are looking at what makes us special. That's a fair statement, a human being wrestling with a problem, that's what that is, yeah, where it's both emotional.

But there's also a sort of a rational attempt to sort of sort through those emotions. Those are the cases that are showing more activity in that area.

So in those cases, when these dots above our eyebrows become active, what are they doing well?

He doesn't know for sure, but what he found is in these close contests. Whenever those nodes are very, very active, it appears that the calculating section of the brain gets a bit of a boost. And the visceral inner chimp section of the brain is kind of muffled. Oh.

You know, that's the hypothesis. So it's going to take a lot of more research to sort of tease apart what these different parts of the brain are doing. Or if some of these are just sort of activated in an incidental kind of way, I mean, we really don't know, this is all all very new.

You?

Okay, so that was the story we put together many, many, many years ago, about a decade ago. And at that point, the whole idea of thinking of morality is kind of purely a brain thing. It's relatively new, and certainly the idea of philosophers working with FMRI machines. It was super new, but now here we are, ten years later, and some updates. First of all, Josh Green.

So in the in the long, long stream of times, I assume now you have three giraffes.

To Bob's rats and children, Yeah, so two kids and we're close to adding a cat. We talked to him again.

He has started a family, he switched labs from Princeton to Harvard. But that whole time, that interim decade, he has still been thinking and working on the trolley problem. Did you ever write the story differently? Absolutely. For years, he's been trying out different permutations of the scenario on people like, Okay.

We asked people about hitting a switch that opens a trap door on the footbridge and drops the person. In one version of that, the switch is right next to the person, in another version, the switch is far away. And in yet another version, you're right next to the person. And you don't push them off with your hands.

But you push them with a pole who? And to cut to the chase. What Josh has found is that the basic results that we talked about, that's roughly held up. Still, the case that people would like to save the most number of lives. But not if it means pushing somebody with their own hands or with a pole, for that matter. Now here's something kind of interesting, uh? He and others have found that there are two groups that are more willing to push the guy off the bridge.

Psychopaths, I mean, some people just don't care very much about hurting other people, they don't have that kind of an emotional response.

That would be the psychopaths, whereas the Buddhist monks, presumably, are really good at shushing their inner chimp, as he called it, and.

Just saying to themselves, You know? I'm aware that this is that killing somebody is a terrible thing to do, and I feel that. But I recognize that this is done for a noble reason, and therefore it's, it's, it's okay.

So there's all kinds of interesting things you can say about the trolley problem. As a thought experiment, at the end of the day, it's just that it's a thought experiment. What got us interested in revisiting it is that it seems like the thought experiment is about to get real.

You?

Jet Robert Radiolab Okay, so where we left, it is that the trolley problems about to get real.

Here's how Josh Green put it. You know, now, as we're entering the age of self-driving cars, oh, this is, this is like, the trolley problem now finally come to life.

The future of the automobile is here.

This legislation is the first of its kind focused on the car of the future that is more of a supercomputer on wheel.

Okay, so self-driving cars, unless you've been living under a muffler, they are coming. It's gonna be a little bit of an adjustment for some of us.

It hit the brakes.

But what Josh meant when he said, it's the trolley problem, come to life, he's basically this.

Imagine this scenario the self-driving car now is headed towards a bunch of pedestrians in the road. The only way to save them is to swerve out of the way. But that will run the car into a concrete wall and it will kill the passenger in the car. What should the car do? should the car go straight and run over, say, those five people, or should it swerve in and and and kill?

Suddenly is a real-world question.

If you ask people in the abstract, like, what?

Theoretically, should a car in this situation do? They're much more likely to say, I think you should sacrifice.

One for the good of the many. They should just try to do the most good or avoid the most harm.

So if it's between one driver and five pedestrians, logically would be the driver kill that I'm driver, be selfless. I think I should kill the driver. But when you ask people.

Forget the theory would you want to drive in a car? That would potentially sacrifice you? To save the lives of more people? In order to minimize the total amount of harm, they say, No, I wouldn't buy it, no.

Absolutely not. That would kill me in it. No, so I'm not gonna. I'm not gonna buy a car that's gonna purposely kill me.

For sure, no, I.

Sell it, but I won't buy it.

So there's your problem. People would sell a car and an idea of moral reasoning that they themselves wouldn't buy. And last fall, an exec at Mercedes-Benz face planted.

Show home to some of the most beautiful cars in the world. Okay, October.

2016 The Paris Motor Show. You had something like a million people coming in over the course of a few days. All the major car makers were there. Here is Ferrari, you can see the.

LaFerrari, Alberta and, of course, the new GT 4C.

Lusso, T.

Everybody was debuting their new cars, and one of the big presenters in this whole affair was this guy. In the future, you'll have cars.

But you don't even have to have your hands on the steering wheel anymore. But. But maybe you watch a movie on the head-up display, or maybe you want to do your emails?

That's really what we are striving for. This is Christophe On Hugo, a senior safety manager at Mercedes-Benz. He was at the show sort of demonstrating a prototype of a car that could sort of self-drive its way through traffic in this e-class today, for example.

You've got a maximum of comfort and support systems you'll actually look forward to being stuck in traffic jams.

Won't you? Oh, of course, of course. He was doing dozens and dozens of interviews through the show, and in one of those interviews, unfortunately, this one we don't have on tape. He was asked What would your driverless car do in a trolley problem type dilemma? Or maybe you have to choose between one or many?

Quote If you know, you can save.

One person, at least, say at one, if you know you can save one person, save that one person, save the one in the car. This is Michael Taylor, correspondent for Car and Driver magazine. He was the one that Christophe on Hugo said that to.

One death can be prevented, and that's your first priority.

Now, when he said this to you, this producer Amanda Rancic, did it seem controversial at all? In the moment? In the moment, it seemed incredibly logical. I mean, all he's really doing is saying, what's on people's minds, which is that, no, I wouldn't buy it. Who's gonna buy a car that chooses somebody else over them anyhow? He makes that comment, Michael prints it, and.

That's Christophe von Hugo from Mercedes.

But then when you lay out the questions, you sound like a bit of a heel because you want to save yourself.

As opposed to the pedestrian, doesn't it ring, though, of, like, just privilege? It does, yeah.

What would you do? It's you or a pedestrian, and it's just, you know, I don't know anything about this pedestrian. It's just you were a pedestrian, just a regular guy walking down the street.

Screw everyone who's not in a Mercedes. And there was this kind of uproar about that. How dare you drive these selfish, you know, make these selfish cars? And then he walked it back and he said, No, no, what I mean, is that just that we that we have a better chance of protecting the people in the car. So we're going to protect them because they're easier to protect. But of course, you know, there's always going to be trade-offs.

Yeah, and those trade-offs could get really, really tricky.

I'm the co-director of the GM CMU Connected and Autonomous Driving Collaborative Research Lab.

He is one of the guys that is writing the code that will go inside GM's driverless car, he says. Yeah, the sensors at the moment on these cars still evolving, pretty basic.

We are very happy. If today you can actually detect pedestrian can detect a bicyclist, a motorcyclist, different vehicles of different shapes, sizes and colors. But, he says, it won't be long before you can actually know a lot more about.

Who these people are? Eventually, they will be able to detect people of different sizes, shapes and colors. Like, Oh, that's a skinny person, that's a small person, tall person, black person, white person, that's a little boy, that's a little girl. So forget the basic moral math, like, what does a car do if it has to decide?

We've also had that term car-to-car communication, well, that's also one of the enabling technologies and highly automated driving.

Mercedes guy basically said, In a couple years, the cars will be networked, they'll be talking to each other. So just imagine a scenario where, like cars are about to get into accidents. And right at the decision point, they're like, conferring, Well, who do you have in your car, me?

So you can basically tie yourself up and not wrap yourself around an axle. We do not think that any programmer should be given this major burden of deciding who survives and who gets good. I think these are very fundamental, deep issues that society has to decide at large. I don't think programmer eating pizza and sipping coke should be making the call.

But how does society decide? I mean, help me imagine that. I think it really has to be an evolutionary process, I believe. Raj told us that two things basically need to happen first, we need to get these robo cars on the road, get more experience with how they interact with us human drivers, and how we interact with them. And to there need to be, like, industry wide.

Summits. No one company is going to solve that. This is Bill Ford Jr.

Of the Ford company, giving a speech in October of 2016 at the Economic Club of D.C.

And we have to have, because he. Could you imagine if we had one algorithm and Toyota had another, and General Motors had another? I mean, it would be, I mean.

Well, obviously couldn't do that, because, like, what if the Tibetan cars make one decision in the American cars make another? So we need to.

Have a national discussion on ethics, I think, because we've never had to think of these things before. But the cars will have the time and the ability to do.

Leave a colleague.

Who is this pasta? So far, Germany is the only country that we know of that has tackled this head-on.

This biggest, I know. One of the most significant points the ethics commission made is that autonomous and connected driving.

Is an ethical imperative? They? The government has released a code of ethics that says, among other things, that self-driving cars are forbidden. To discriminate between humans in almost any way. Not on race, not on gender, not on age, nothing.

These shouldn't be programmed into the cars. One can imagine a few classes being added in the Geneva convention, if you will, of what these automatically should, a globally accepted standard, if you, how we get there.

To that globally accepted standard is anyone's guess. And what it will look like, whether it'll be like a coherent set of rules or like. Rife with the kind of contradictions we see in our own brain, that also remains to be seen. But one thing is clear.

Oh.

Okay, we do need to caveat all this by saying that the moral dilemma we're talking about. In the case of these driverless cars, it's gonna be super rare. Mostly, what'll probably happen is that, like the plane loads full of people that die every day from car accidents, well, that's just gonna hit the floor. And so you have to balance the few cases where a car might make a decision you don't like against the massive number of lives saved.

I was thinking actually of a different thing, I was thinking. Even though you dramatically bring down the the the number of bad things that happen on road, you dramatically bring down the collisions, you dramatically bring down the mortality, you dramatically lower the number of people who are drunk. Coming home from a party and just ram someone sideways and killing three of them and injuring two of them for the rest of their lives, those kinds of things go way down, but they're the ones that remain, are engineered like they are calculated, almost with foresight. Mm-hmm. So here's the difference, and this is just an interesting difference. Like, Oh damn, that's so sad that happened, that that guy got drunk and out of that.

It's a little bit like when you know, should you kill your own baby to save the village, like in a particular instance of that one child? It's dark, but against the backdrop of the life saved, it's just a tiny pinprick of darkness.

That's all, but you know how humans are if you argue back that, yes, a bunch of smarty-pants is concocted. A mathematical formula which meant that some people had to die, and here they are there, many fewer than before. A human being, just like Josh would tell you, would have a roar of feeling and of anger and saying, How dare you engineer this? In no, no, no, no.

And that human being needs to meditate, like the monks, to silence that feeling, because the feeling in that case is just getting in the.

Way, yes and no. And that may be impossible unless you're a monk for God's sake.

See, we're right back where we started now.

Yeah, we should go Chad, you have to thank some people. No, yes.

This piece was produced by Amanda Aron Check with help from Bethel Hobb Tate. Special thanks to Iyad Rowan, Edmund Awad and Sidney Levine from the Moral Machine Group at MIT. Also thanks to Sir Tack, Karaman, Shin Chang and Robo Race for all their help. And I guess we should go now, yeah.

I'll, um, I'm chad, I boom rod not getting into your car.

You don't mind, just take my own. I'm gonna rig up an autonomous vehicle to the bottom of your bed, so you're gonna go to bed and suddenly find yourself on the highway, driving you wherever I want.

No, you won't.

Anyhow, okay, we should go. I'm Jed, I'm Robert Coleridge, thanks for listening.

You?

Radiolab was created by Chad a Boom Rod, and it's edited by Soren Wheeler, Lulu Miller and Professor. Our co-hosts. Dylan Keith is our director of sound design. Our staff includes Simon Adler, Jeremy Bloom, Becca Bresler, Rachel Cusick, But Katie Foster. Keys, W. Harry Fortuna, David Gable, Maria Pause, Gutierrez, Sindhu, Nyana, Samba Them, Matt Guilty, Annie McEwan, Alex Nissen Sarkari on a rescue at pause, Alyssa John Perry, Sarah Sound back area and whack, Pat Walters and Molly Webster with help from Timmy Brother. Our fact-checkers are Diane Kelly, Emily Krieger and Natalie Middleton.

Leadership support for Radiolab science programming is provided by the Gordon and Betty Moore Foundation, Science Sandbox, A Simon Foundation initiative, and the John Templeton Foundation. Foundational support for Radiolab was provided by the Alfred P. Sloan Foundation.

Since WNY Sees first broadcast in 1924, we've been dedicated to creating the kind of content we know the world needs. Since then, New York public radio's rigorous journalism has gone on to win a Peabody Award and a Dupont Columbia Award, among others. In addition to this award-winning reporting, your sponsorship also supports inspiring storytelling and extraordinary music that is free and accessible to all. To get in touch and find out more, visit sponsorship wnyc. org.