Episode 266: Prof. Cass Sunstein: Practical Reason in Ordinary Life

28:54But it has a little more of a foundation than that. If people told me that dropped objects don't fall, which I don't believe to be true, I won't believe that. If they told me dropped objects do fall, I feel that confirms what I believe and confirmation bias would kick in. And it's rational, given your pre-existing beliefs, to update or not, depending on how it fits with what you think. So there's the motivated part, and then there's the rational updating part. So take that as confirmation bias. Then there's a kind of subset called motivated reasoning, which is the emotional part where people believe things that they want to believe.

29:38And that means that if you tell me something about politics or something about a politician that I really am saddened by, I might think you're biased and you don't study very hard. So you don't know that the politician I love is actually God's gift and couldn't possibly have done or said that bad thing. So motivated reasoning is a second thing. The third, the newer, is desirability bias, which suggests that we believe things that we find it desirable to believe. So if you tell me that actually the hair loss that I thought I had, I actually don't have, that it's an artifact of Zoom and some unflattered photographs, but I have a completely full head of hair, that's desirable to hear.

30:26Thank you for that. And I will find that particularly credible. If you tell me something like my male Labrador retriever is not so beautiful, he's okay looking, but he's not so beautiful, I'd find that highly undesirable and I'd not want to believe it. So there's desirability bias. Now, what's really fun, I think, is confirmation bias and desirability bias, if you're with me, often will go in exactly the same direction. that what confirms my beliefs I'll find agreeable and credible, and that will be desirable. But you can pull them apart. If you think, for example, that you aren't very good at sports, but you're given information suggesting you actually are good at sports, that is disconfirming information, but it's highly desirable information.

31:18And we have some data suggesting in a horse race between desirability bias and confirmation bias. Desirability bias is secretariat, is the better horse. That is, people will believe information that they want to believe, even if it is disconfirming of what they started out believing. So this is a very long-winded way of saying that what people believe fits with what they find it pleasing to believe, motivated reasoning, and with what they start out believing. That can mean, one last bit, that good news will be more credible than bad news with respect to almost everything, which can lead to unrealistic optimism with respect to, let's say, investments.

32:10And it can also lead to terrible mistakes in updating. I think I remember in the book you refer to those as sort of asymmetries in how we update our beliefs. How can people improve the symmetry in how they update their beliefs? This is maybe my favorite part of the book in the sense that it was the most fun to deal with. So thank you for this. If people go into a laboratory, let's say, and are asked to rate themselves in terms of intelligence and looks, and then are given objective information suggesting they're smarter and better looking than they thought, they update. Say, oh, great. I'm smarter and better looking than I thought.

32:52If they're given information suggesting that they're less smart and less good-looking than they thought, they think the person who made that assessment is mean and ignorant, and I am every bit as good-looking and every bit as smart as I thought. That's the asymmetry to which you point, that people either make a decision to update in a way that's asymmetrical in that way, or it just happens because the motivation is so strong. What we learned, my colleagues and I learned, is that this works for climate change too. That if people believe that climate change is really bad, information that suggests it's really, really bad will be more credible than information suggesting that it's not so bad.

33:37And if people believe that climate change is only a little bit bad, they will find that information suggesting it's very bad is incredible. But information suggesting it's not even a little bit bad, that is credible. So you have asymmetrical updating of very different kinds, depending on whether you think that climate change is a really bad problem or only a little problem. And that, I think, helps explain polarization on political issues, also economic issues, also things within the family even, where spouses might update in very different ways. the best way to respond to asymmetrical updating is just to be really cold-blooded in your predictive judgments, not like inhumanly cold-blooded, but to be very, very calm.

34:26And if you hear, for example, something suggesting that your performance isn't as good as you hoped it was, to think, okay, deep breath, I'm going to believe that, even though it's not very good news. Or if you hear something suggesting that the economic risks you face, given your current portfolio, are serious, and maybe you ought to change it, not to say, I don't like that, I don't believe it. But instead, to say that might be taken a little more seriously even than information that made me smile and jump for joy, and to try to be a maximizer with respect to information. I'm curious, what might deter someone from updating their beliefs when they are presented with new evidence?

35:12Okay, so there are two things. And our data on climate change doesn't enable us to see which one works. But there are two that are explanatory. Suppose you think that climate change is a really serious problem. You're given information suggesting it's an even more serious problem than you thought. That's very credible. So you're likely to update. Now, suppose you believe climate change is a really serious problem and you're giving information suggesting it's not that bad. You might think, who paid for that information? Given your previous belief, that information just is suspicious. I don't believe it, given my belief, my initial belief.

35:53And let's call that just rational evaluation, giving your prior judgments. If you read something suggesting the Holocaust didn't happen, that might not motivate you to doubt the historic fact of the Holocaust. You might just think, who wrote that? What's their agenda? That's the rational account. And that's consistent with the data we have. If someone thinks, I'm really good looking, that my life is consistent with that. And then you get information suggesting that someone thinks you're not good looking, you might think, I'm not going to believe that that person's just in a bad mood or trying to get me down.

36:32Okay, the other is that if you get information of certain kinds, it's just going to make you mad or sad. Your natural reaction is to think, I'm not going to believe it, because I don't want to be mad or sad. An example of this improbably comes from the writing of the Star Wars movies, that is the George Lucas movies, where George Lucas was in a pitched battle with one of the great screenwriters of all time, Lawrence Kasdan, about what would happen in the third of the original series. And if you don't know Star Wars, hang with me for a moment. I think you'll get it. Lawrence Kasdan said that Luke should die.

37:10Luke Skywalker, our hero, should die. And George Lucas said, Luke Skywalker isn't going to die. And then Lawrence Kasdan said, well, Princess Leia should die. And then George Lucas said, Princess Leia isn't going to die. And then Lawrence Kasdan gave a kind of poetic speech about how in art, if you lose someone you love, the depth of your connection with the work grows. In great art, someone dying is actually a pivotal moment that cements the audience and the work of and George Lucas responded really quick. This dialogue is all captured in real time. He said, I don't like that, and I don't believe that.

37:51That's amazing for our purposes. I don't like that, and I don't believe that. Not liking preceded not believing. Lucas thought, I think, the idea that someone you're going to die makes the audience like the movie better. These are my characters. I don't like that. And then he says, I don't believe it. So that's the basic non-rational account, though I think Lucas called it exactly right on Star Wars, that if you don't like something, you won't believe it. And that's not about rational updating. That's about denial of unpleasant truths. Is it safe to call the items that we just talked about, the asymmetries and updating information and beliefs, is it safe to call those biases?

38:36I think so, if we emphasize the motivated nature of updating. I think not if we emphasize the rationality of updating. And you see this, you're putting your finger on something that the data doesn't sort out. So in the climate change work, what we don't know is whether strong climate change believers discounted evidence of it's not so bad because they just thought it was incredible and probably someone bought that material or what just upset them because they're invested in thinking climate change is a big problem. That's part of the way they live in the world. We don't know which is more powerful.

39:25I have a hunch, which is I think the motivations are more powerful, but both operate in different contexts. Suppose you have really good health, and then you go to a new doctor, and the doctor says, I think you have a pretty bad problem. You might think, I don't know who this doctor is. I'm not sure why I went to see that doctor. And that need not be motivated. It might be based on 15 years of great health. I want a second opinion. Not because I don't want to believe what this new doctor said, but because I think it's not a real doctor. It's a witch doctor. What causes people to make inconsistent decisions?

40:08That's a fantastic question. And there are a million reasonable answers. Let's give two. One is mood. I was in a group of friends soon after I married my wife. And I said, a group of maybe six male friends, I said, I knew within two weeks of meeting my wife that I wanted to marry her, which is true. And one of my male friends said, if I married every woman whom I knew I wanted to marry within two weeks of marrying her, I would have 17 ex-wives. I'm happily married. I hasten to add 15 years in. But my friend was saying that you can be inconsistent because of mood. So if you are all excited and charged up for one or another reason, you might make a decision on Monday that would be radically different from what you do on Tuesday and Wednesday.

41:04So and the factors that influence mood are innumerable. and sometimes they're not even visible to us in real time. If your city has a very sunny day, you might take a lot of risks that you wouldn't take if it was raining and bleak, and you might have no idea that the weather caused your risk-taking. One thing that I'm particularly interested in is whether you're choosing something in isolation or whether you're choosing something in the context of other options. So let's say you're deciding which laptop to buy. If you see a laptop, I'm seeing a laptop right now, it's a good laptop, you might think that's a great laptop, I want that laptop.

41:51If you see the laptop I'm seeing right now in the context of some other laptop, you might think it doesn't look so good, the screen's kind of small and the Beezles are kind of big. And the other one, the screen's bigger and hardly any Beezle at all. And I want that one. Now, it might be that in isolation, you would be thrilled with your laptop, your house, your investments, your doctor, any product you have at home, but that if you were comparing it to something else, you wouldn't be so thrilled with it. And here's the upshot. In separate evaluation of things, we may prefer A over B. But if you see A and B together, you might choose B over A.

42:38And smart marketers are very well aware of that. And they might tell you when you enter the store, here are your two options. And they know they can get you to get, let's say, the more expensive option because they showed you the other one, Even if you saw the two in isolation, you would choose the less expensive. So the three of us are in relationships. I think we're all in happy relationships. Can you talk more about how evaluating a big purchase like a home jointly versus individually affects the decision? Let's say that you have a house which has a big yard and beautiful views that's very far from your work.

43:25You might think in isolation, that's not a good choice. Because even though it's an amazing house, the commute will be terrible. And if that's the only house you see, let's say, in August, you might think I'm not going to get that house. Now, let's suppose in September, you're shown another house, which doesn't have amazing views and isn't that big and beautiful, but that is really convenient for getting to your work. You're now just seeing that house in September. The other house isn't on your view screen anymore. you think that house is the one for me because it's a good house and because my life is going to be completely fine given the ease of commute.

44:13Then you get, let's say, the less good house with an easy commute. But if in August you had been shown the two houses, there's a very good chance the data suggests that they go for the better house with the bad commute. Because you might well think, look, if they're priced at the same, a beautiful big house is just what I want. A house is to live in. It's not to commute to and from, which means that a characteristic of the house that in separate evaluation, that is, it's not that big and not that beautiful, might seem like a mild negative, will in joint evaluation, where you're comparing it with a huge and gorgeous house, look pretty disappointing.

44:57And this can be done for home purchases, for vacations. Do you want a vacation that's far away and in the most beautiful place imaginable? Or do you want a vacation in a place that's pretty close and very good? In joint evaluation, there's a good chance people will go to the far away vacation in the magnificent place because a vacation is for magnificent places. In separate valuation, they might be willing to pay more for the place that is close on the ground that proximity of vacation to many people really matters because they don't like a terrible commute. These are a little bit of exotic examples, but it works every day in stores.

45:40If you're shown a product, let's say a cell phone, you might think that's great. I'm going to get that one. But if you're shown that cell phone in the context of another cell phone, it might be some characteristic that you wouldn't give much weight to in separate evaluation. starts to loom large because that other cell phone does better along that dimension that you wouldn't even notice or care about unless you saw the two together. It's mind-blowing stuff to think about because we make decisions like this all the time. Is one of separate evaluation or joint evaluation objectively better than the other for most decisions?

46:16So the cool kids say that separate evaluation is better. If there are any academic the cool kids. This is what they think. And the reason is they think that once you get, let's say, a laptop or a house or something that involves music, a guitar, something that plays music, you're not going to be comparing it to the other thing all the time, unless you're a weird person, you're going to be enjoying it in separate evaluation. So they say, get the thing that in separate evaluation is fine. So if it's, let's say, something that has good sound quality, but not a sounding sound quality that's really beautiful and fits, don't reject that in favor of the thing with the even more amazing sound quality because you just won't notice.

47:00So the cool kids say separate evaluation is better. And they're usually right. The reason they might be wrong is if in joint evaluation, your attention is concentrated on something in the product that really does matter, that you wouldn't evaluate or pay attention to unless you saw the two. That's too abstract. Let's say you have two cell phones, and one of them has an incredible camera, and one has a very good camera. Now, for most of us, separate evaluation is what matters, and a very good camera is plenty good enough. But let's say you're a photographer, or you live for amazing photographs.

47:42It really matters to you. In separate evaluation, you won't realize that the camera is merely very good. In joint evaluation, you'll see there's another one that's incredible. And for that person and that choice, joint evaluation is better because it puts a spotlight on something that really should matter to you. This next question is the one I had in my mind as I read your book. How should people decide what to value making decisions? Like for example, is maximizing wellbeing the right objective for all decisions? We're going to do some philosophy here. Maybe the greatest political philosopher of the 20th century had a footnote and a manuscript which he never published, which says we post a signpost, no deep thinking here, things are bad enough already, but we're going to do a little deep-ish thinking at least.

48:30I say that across a very wide terrain, our decision should be based on our well-being. I'll qualify that in two ways in a moment. But if you're deciding what to do with respect to the summer or what to do with respect to fall, an essential question and possibly the only question, qualifications to come, is what makes your summer or fall better? And better can include various things. Let's note three that people seem to care about. What makes you happy? What gives you a good week? Are you going to be smiling or scowling? Second is what makes your life meaningful? Are you going to be thinking, I've binged on seven TV shows and I really enjoyed it, but what the heck was I doing with my time then?

49:21It's kind of meaningless. And the third thing is, is there a variety? So is it psychologically rich? People will sacrifice happiness for the sake of variety. They want, I went to a tennis match last night, by the way. I didn't really want to, a professional tennis match. My wife said, give you psychological richness. You haven't done this for a long time. And it wasn't that meaningful. Did I enjoy it? Some. Am I glad I did it? Absolutely. Because it was something different, something I don't normally do. So people care about happiness, meaning, and psychological richness, and they balance them in different ways.

49:59That's about well-being. And that's, I think, very fundamental to our decisions. And we can get very intuitive about it, or we can get very kind of elaborate about it, but that's a lot of the terrain. It's also good to do the right thing. So if, let's say, a successful act of theft would make you happy because what you stole was really great, still shouldn't do it. It's immoral. And that is a qualification to promoting welfare. So we can all discuss what's right and what's wrong, but there's a certain number of things that are clearly wrong. Then there's the welfare of others. So it might be that you'd spend some hours or maybe some weeks working to help people who are struggling or trying to help maybe your kid.

50:56And it might be that that's not a very joyful thing to do. It might be, but it might not be. If it is, that's not why you're doing it. It might be meaningful, but if you're doing it, you're probably not doing it. if you're helping others, it's not because you're thinking, I want to have a meaningful life. It's because you're thinking, I want to help others. And it might be that something you're doing that you think is good for people you care about or people generally adds richness to your life. And that's great, but that's not why. So well-being, I would put very much in the forefront, but not one's own.

51:37And maybe let's talk about John Stuart Mill with your indulgences for a second. John Stuart Mill was a great philosopher. He was utilitarian. He cared about well-being, but he wasn't a very emotional writer most of the time. At one point on the question you asked, he got full of fire. And he said, when you think about utility or well-being, what's maximized is not the agent's own, not the decision makers. It's everybody's. And then he says, the only time I think Mill ever referred to Christianity. He said, and this is chills down the spine stuff, in the golden rule of Jesus of Nazareth, do unto others, we find the complete spirit of the ethics of utopia.

52:18Wow. You did say that we were going to get into some philosophical stuff, and we did. A warning. Warning label. Yeah, warning. Proceed with compassion. I don't know if this next question takes us in the completely opposite direction, but when do you think that algorithms are useful for making good decisions? We have data suggesting that if doctors are deciding whether to test people for a cardiac incident, including a heart attack, they do well, but they don't do as well as an algorithm. The reason the algorithm appears to do better is doctors overweight the current symptoms, chest pain, say, and demographics.

53:02So if it's a young woman, the algorithm will give that, let's say, roughly its appropriate weight, and the doctors will give it a little more weight. So in deciding whether to test someone for having had a heart attack, rely on the algorithm, it looks like, or at least pay very careful attention to the algorithm. If you're trying to decide whether, let's say, a song is something you like, it may well be that a good algorithm will nail your tastes. And even someone who knows you very well, like your best friend or your spouse, won't do that. So rely on the algorithm. If there's a ton of data that the algorithm has that isn't biased in any relevant respect, chances are it's going to do really well.

53:52But you might not want the algorithm even so if you like deciding for yourself or if you want to take responsibility for the choice that affects you. So the algorithm might be a little bit of a science fiction world in which you get the right outcome, but you didn't get it in the right way. Some people will think with respect to, let's say, vacation planning. I can find an algorithm that will tell me what vacation I'll like. But the point of a vacation includes preparation for it, figuring it out, seeing the options. And even if the algorithm says, go to Key West, young man, I'd rather figure out for myself whether Key West is where I should go.

54:36You said something earlier that stuck with me as well, with kids, that you probably want to decide yourself when it's related to your kids. completely. That with your kids, the feeling of fobbing off a decision to an algorithm feels like you're not really being a parent. There's also learning. So this is something I have some data, unpublished data on, that people often want to make choices and they decide, they make a second order decision to decide because they think that the process of doing that will give them an asset that they can use. They might like it, but they also will learn, let's say, about vacation spots or about cars or about something that matters to them.

55:20The algorithm could make the right decision, but then they wouldn't know anything about vacation spots and cars. Are there other cases where algorithms are not useful? You have exactly the right time for that question, meaning this is maybe the question for our era. algorithms seem pretty bad at figuring out who you're going to fall in love with. They really struggle to know that. They don't do well in figuring out whether a song is going to be a hit, and they don't do well in predicting whether there's going to be a revolution. Now, what is it about those three things? And I give the three examples because they have something in common.

56:02We have really good data suggesting they don't do well in predicting who's going to fall in love, that if there's social interactions that can go in any number of directions that are hard to anticipate in advance, then the algorithms aren't going to be helpful. So whether there's going to be a revolution or a rebellion depends on lots of moving parts. And maybe you can say something about the probability that the parts will move in a certain way. But maybe you can't say a lot even about the probability. Depends on who says what, who is noticed by whom. To have a rebellion, a ton of things have to happen kind of on the spot.

56:46And a romantic spark is a little bit like that. It might be that two people meet, and one person happens to say something, and the woman, let's say, says, God, that is really a creepy thing to say, or pretty clueless. Game over. That guy has no chance. But if the guy had said something else or just silenced himself at that moment, she would have given him a chance, and maybe sparks would have flown. Or maybe she was feeling really great because something could to happen at work that day. And so she's up for fun and she's hilarious rather than grim. And he's smitten in the first 10 minutes. And it's just because things were good at work that day.

57:34And how can the algorithm predict that she's going to have a good day Thursday at work? So if there are multiple factors that feed into social interactions, it may be that algorithms will really struggle. A more mundane thing, I'll tell you about judges, if I may, that algorithms outperform judges in deciding who should get bail. The reason they do is that algorithms do a bit better than judges in predicting who's going to flee. They outperform the judges. It's kind of spectacular data because it suggests in a big city, you could jail the same number of people and have a lot fewer crimes because the algorithm would figure it out better.

58:20Or you could have the same number of crimes and jail many fewer people. You can just make your choice. The algorithm will get you there. But here's the but. 10 % of judges, the best 10 % outperform the algorithm. They do better. This isn't like about falling in love or about whether some song by a new group is going to be a big hit. It's that people sometimes will know things that are either impossible or really hard to program an algorithm to notice. Now, it might be that the 10, we don't know why the 10 % of judges crush the algorithm. It might be that they know something from experience that the algorithm hasn't been programmed to take account of.

59:03It might be really hard to do, or it might be that something happens on the spot where the judge sees that defendant and says that defendant is a good person, even though their last 10 years have been pretty rough. The algorithm can't see that. Okay. So we covered when algorithms are maybe not so useful. Can you talk about cases where they're harmful? So you could have an algorithm that is going to predict, let's say, whether a plane is in trouble or not, and that tells a pilot what to do that doesn't capture something on the spot that the pilot will be alert to. So it might be by training or experience, the pilot knows this plane is actually not in trouble, or that that solution to the problem is going to make things worse.

59:52And the algorithm is too mechanical and rigid to capture that. We actually have data in some domains that are quite like that that suggests algorithms will not take account of the full set of things that we should consider. And that's a big problem. I'll give you an example from personal experience, which is grading, where an experienced teacher will pay attention to maybe 30 things. And to program an algorithm to do that is very challenging to program an algorithm to take account of 20 things might not be so challenging. And if you're thinking of doctors or pilots or engineers or real estate agents, it might be similar.

1:00:36It's also true that the garbage in, garbage out principle might ruin algorithms or impair their performance. So if you ask an algorithm to say whether employees are going to leave the workplace, it might end up being a discriminator against women because see that women, let's say in the relevant place are more likely to leave because they're going to be full-time mothers for a period. And then you hire mostly men. And that's very bad. And a human being might think, I know the women are more likely to leave than the men are, but I'm going to give everybody a chance. We've talked about shrouded attributes in complex financial products on this podcast before, that there are also things like dark patterns in software.

1:01:25How should people deal with the possibility that their decisions are being manipulated? The word manipulation is easier to use than to define. Let's say roughly it's where a situation overrides your capacity to make a rational choice. That might be because some characteristic of the product, let's say, isn't visible to you. It might be because your emotions are triggered such that you think, I've got to get that like now, and you're being manipulated through hope or fear to get the relevant thing. I think from the standpoint of private and public conduct, we want a right not to be manipulated.

1:02:11Just the very idea of a right not to be manipulated is probably a step forward. And we want to specify it by forbidding things that clearly violate that right, as when people are automatically enrolled in terms that they can't easily see and that aren't in their interest, like you're going to make referring payments that aren't very visible. I think for all of us, before we make something that's costly to do maybe 20 % more due diligence than we'd be inclined to do. That's a very good idea, especially if it's online. Your work on nudges is widely known, widely applied, and certainly impactful.

1:02:55I have to ask you, how are nudges different from manipulation? Manipulation, we're kind of defining as bypassing or not enlisting people's rational capacities. Some nudges are educative, like a reminder. Speaking of rational reminders, so if you get a reminder on this phone that says you have a doctor's appointment tomorrow, I actually do. So this is a salient example. Thank you for helping me to remember that. That is a non-manipulative act. It is educating me about a thing. If you go to the grocery store and there's a warning that if you're allergic to peanuts, don't buy this. That's not manipulative.

1:03:39That's educative. If you're given disclosure about the side effects of medicines that says something about if you have X or Y or Z, you probably shouldn't take this medicine, that's not manipulative. So educative nudges wouldn't be manipulative. There are others that are architectural, and then we have to identify them to figure out whether they might be manipulative. One kind of architectural nudge asks you whether you want to do this or whether you're sure you don't want to do that. So if you're about to delete all your files, your computer might say, are you sure you want to? Or if you're about to make a very big purchase, the same question might be asked.

1:04:20And that's not manipulative. That's an effort at enlisting your deliberative capacities to make sure you're not being impulsive or reckless. So those nudges wouldn't be manipulative. Graphic health warnings may or may not be manipulative, depending on what they are and what they do. So if you're given a graphic health warning about smoking, we can have a discussion of whether that's manipulative. I think the data suggests it isn't. That a graphic health warning for cigarettes, which is a nudge, actually informs people of smoking. So it doesn't bypass or exploit their intuitions. It actually improves their judgments.

1:05:01But that's one to discuss. if you're automatically enrolled in something without knowing it, that's a pretty bad nudge. And it wouldn't be a nudge that Thaler or I would approve of, and it would be manipulative. If you are automatically enrolled in something and you're told you're automatically enrolled in this, you want to opt out, then it wouldn't be manipulative. So I think we'd want to go through the universe of nudges, the ones that are used by governments, at least the governments I like and familiar with, it's hard to find any of them as manipulative, but graphic warnings and automatic enrollment at least would be worth discussing.

1:05:45Our final question for you, Cass, how do you define success in your life? I'd be a little focused on the three variables we've discussed. Happiness is a big part of it, whether your days are great or not, whether you feel there's meaning in your life. I had a friend in the White House I saw actually in the White House one day, and I asked him how things are going, and he gave me a surprising answer. He said, my day-to-day happiness is terrible, but my sense of meaning is fantastic. And I thought that was a little more information than I had asked for. I just said, how are you doing? But it was instructive.

1:06:25So happiness, a sense of meaning or purpose is really important to that, as is a sense of joy or suffering, and they're different, and a sense of whether you have variety in your life. So you could have a thousand happy days in a row, but they can be the same. That's not successful, or as successful as if you had the same days with variety. So variety, meaning, and let's say happiness in the sense of ordinary language. Great answer. And you made today a great day. Thank you so much, Cass. Great pleasure. Thanks to you.