WEBVTT - At The Money: Algorithmic Harm with Professor Cass Sunstein 0:00:02.520 --> 0:00:17.280 Bloomberg Audio Studios, podcasts, radio news. 0:00:26.239 --> 0:00:30.760 Algorithms are everywhere. They determine the price you pay for 0:00:30.800 --> 0:00:35.360 your Uber, what gets fed to you on TikTok and Instagram, 0:00:35.800 --> 0:00:40.639 and even the prices you pay in the supermarket. Is 0:00:40.680 --> 0:00:45.680 all of this algorithmic impact helping or harming people? To 0:00:45.760 --> 0:00:49.360 answer that question, let's bring in Cass Sunstein. He is 0:00:49.400 --> 0:00:53.280 the author of a new book, Algorithmic Harm, Protecting People 0:00:53.360 --> 0:00:58.520 in the Age of Artificial Intelligence, co written with ornbar Gil. 0:00:58.920 --> 0:01:02.600 Cass is also a professor at Harvard Law School and 0:01:02.720 --> 0:01:06.520 is perhaps best known for his books on Star Wars 0:01:06.720 --> 0:01:12.080 and co authoring Nudge with Nobel Laureate Dick Taylor. So, pass, 0:01:12.160 --> 0:01:16.440 let's just jump right into this and start by defining 0:01:16.920 --> 0:01:18.960 what is algorithmic harm? 0:01:20.160 --> 0:01:23.760 Okay, So let's use Star Wars. So let's say the 0:01:23.880 --> 0:01:28.840 Jedi Knights use algorithms, and they give people things that 0:01:28.920 --> 0:01:32.639 fit with their tastes and interests and information, and people 0:01:32.720 --> 0:01:35.920 get If they're interested in books on behavioral economics, that's 0:01:35.959 --> 0:01:38.480 what they get at a price that suits them. If 0:01:38.480 --> 0:01:41.280 they're interested in a book on Star Wars, that's what 0:01:41.319 --> 0:01:44.240 they get at a price that suits them. The Sith 0:01:44.319 --> 0:01:48.760 by contrast, take advantage with algorithms of the fact that 0:01:48.840 --> 0:01:54.160 some consumers lack information and some consumers suffer from behavioral biases. 0:01:54.440 --> 0:01:57.320 So we're going to focus on consumers first. If people 0:01:57.560 --> 0:02:01.200 don't know much, let's say about healthcare product, an algorithm 0:02:01.280 --> 0:02:03.720 might know that that they're likely not to know much 0:02:04.480 --> 0:02:07.360 and might say, we have a fantastic baldness cure for you. 0:02:08.200 --> 0:02:11.760 Here it goes, and people will be duped and exploited. 0:02:12.040 --> 0:02:16.120 So that's exploitation of absence of information. That's algorithmic harm. 0:02:16.360 --> 0:02:19.520 If people are super optimistic and they think that some 0:02:19.680 --> 0:02:22.240 new product is going to last forever, when it tends 0:02:22.240 --> 0:02:26.200 to break on first usage, then the algorithm can know 0:02:26.240 --> 0:02:30.280 those are unrealistically optimistic people and exploit their behavioral bias. 0:02:31.440 --> 0:02:37.680 So I referenced a few obvious areas where algorithms are 0:02:37.760 --> 0:02:42.640 taking place. Uber pricing is one the books you see 0:02:43.040 --> 0:02:48.280 on Amazon is algorithmically driven. Clearly, a lot of social 0:02:48.360 --> 0:02:53.240 media for better or WORSEUS algorithmically driven, and even things 0:02:53.320 --> 0:02:56.600 like the sort of music you like on Pandora. What 0:02:56.680 --> 0:03:01.960 are some of the less obvious examples of how algorithms 0:03:02.080 --> 0:03:06.400 are affecting consumers and regular people? Every day. 0:03:07.320 --> 0:03:10.120 Okay, So let's start with us straightforward once and then 0:03:10.160 --> 0:03:13.560 we'll get a little subtle. So straightforwardly, it might be 0:03:13.680 --> 0:03:17.120 that people are being asked to pay a price that 0:03:18.000 --> 0:03:21.160 suits their economic situation. So if you have a lot 0:03:21.160 --> 0:03:23.639 of money, the algorithm knows that maybe the price will 0:03:23.680 --> 0:03:26.080 be twice as much as it would be if you 0:03:26.160 --> 0:03:31.080 were less wealthy. That I think is basically okay. It 0:03:31.160 --> 0:03:34.239 leads to greater efficiency in the system. It's like rich 0:03:34.320 --> 0:03:38.000 people will pay more for the same product than poor people, 0:03:38.040 --> 0:03:41.160 and the algorithm is aware of that. So that's not 0:03:41.280 --> 0:03:44.600 that subtle. But it's important. Also, not that subtle is 0:03:45.080 --> 0:03:49.640 targeting people based on what's known about their particular tastes 0:03:49.680 --> 0:03:52.600 and preferences. Let's put wealth to one side, and so 0:03:52.680 --> 0:03:55.720 it's known that certain people are super interested in dogs, 0:03:55.760 --> 0:03:58.440 other people are interested in cats. And there we go, 0:03:59.000 --> 0:04:02.640 and all that is very straightforward's happening. If consumers are 0:04:02.680 --> 0:04:07.240 sophisticated and knowledgeable, that can be a great thing to 0:04:07.320 --> 0:04:10.240 make markets work better. If they aren't, it can be 0:04:10.280 --> 0:04:14.200 a terrible thing to make consumers get manipulated and hurt. 0:04:14.680 --> 0:04:18.159 Here's something a little more subtle. If an algorithm knows 0:04:18.200 --> 0:04:22.400 for example, that you like Olivia Rodrigo, and I hope 0:04:22.440 --> 0:04:24.840 you do, because she's really good, then there are going 0:04:24.920 --> 0:04:27.400 to be a lot of Olivia or Rodrigo songs that 0:04:27.440 --> 0:04:30.040 are going to be put into your system. And let's 0:04:30.040 --> 0:04:33.159 say no one's really like Olivia Rodrigo, but let's suppose 0:04:33.160 --> 0:04:36.000 there are others who are vaguely like her, and you're 0:04:36.000 --> 0:04:38.320 going to hear a lot of that. Now, that might 0:04:38.440 --> 0:04:42.000 seem not like algorithmic harm, that might seem like a 0:04:42.040 --> 0:04:46.120 triumph of freedom and markets, but it might mean that 0:04:46.240 --> 0:04:49.760 piece of people's tastes will calcify, and we're going to 0:04:49.800 --> 0:04:54.080 get very bulkanized culturally with respect to what people see 0:04:54.080 --> 0:04:56.679 in here. So they're going to be Olivia or Arigo people, 0:04:56.720 --> 0:04:58.440 and then they're going to be led Zeppelin people, and 0:04:58.480 --> 0:05:00.840 they're going to be Frank Sinatra people. And there was 0:05:00.880 --> 0:05:03.360 another singer called Bach. I guess I don't know much 0:05:03.360 --> 0:05:06.560 about him, but there's Bach, and there would be Bach people. 0:05:06.960 --> 0:05:11.359 And that's culturally damaging, and it's also damaging for the 0:05:11.440 --> 0:05:14.360 development of individual tastes and preferences. 0:05:14.960 --> 0:05:19.400 So let's put this into a little broader context than 0:05:19.680 --> 0:05:23.280 simply musical tastes and I like all of those, so 0:05:23.560 --> 0:05:28.760 I haven't become bulkanized yet. But when we look at 0:05:29.040 --> 0:05:32.920 consumption of news media, when we look at consumption of information, 0:05:33.800 --> 0:05:37.960 it really seems like the country has self divided itself 0:05:38.360 --> 0:05:42.560 into these happy little media bubbles that are either far 0:05:42.680 --> 0:05:45.520 left leaning or far right leaning, which are is kind 0:05:45.520 --> 0:05:48.440 of weird because I always learned the bulk of the 0:05:48.480 --> 0:05:53.360 country and the traditional bell curve, most people are somewhere 0:05:53.680 --> 0:05:56.279 in the middle. Hey, maybe they're center right or center left, 0:05:56.600 --> 0:05:59.839 but they're not out on the tails. How does the 0:06:00.000 --> 0:06:05.719 these algorithms affect our consumption of news and information? 0:06:07.560 --> 0:06:10.320 About fifteen twenty years ago, there is a lot of 0:06:10.400 --> 0:06:16.159 concern that through individual choices, people would create echo chambers 0:06:16.240 --> 0:06:19.440 in which they would live. And that's a fair concern 0:06:19.600 --> 0:06:23.520 and it has created a number of, let's say, challenges 0:06:23.760 --> 0:06:29.120 for self government and learning. What you're pointing to is 0:06:29.160 --> 0:06:33.839 also emphasized in the book, which is that algorithms can 0:06:34.040 --> 0:06:38.040 echo chamber you. An algorithm might say, you know, you're 0:06:38.279 --> 0:06:42.000 keenly interested in immigration, and you have this point of view, 0:06:42.480 --> 0:06:44.679 so boy, are we going to funnel to you lots 0:06:44.680 --> 0:06:48.039 of information because clicks are money, and you're going to 0:06:48.040 --> 0:06:50.720 be clicking, clicking, clickling, click kicking, And that might be 0:06:50.839 --> 0:06:53.960 a very good thing from the standpoint of the seller, 0:06:54.080 --> 0:06:56.480 so to speak, or the user of the algorithm, but 0:06:56.600 --> 0:07:00.200 from the standpoint of view, it's not so fantastic. And 0:07:00.320 --> 0:07:03.920 from the standpoint of our society it's less than not 0:07:04.080 --> 0:07:08.400 so fantastic, because people will be living in algorithm driven 0:07:09.080 --> 0:07:13.600 universes that are very separate from one another, and they 0:07:13.680 --> 0:07:15.680 can end up not liking each other very much. 0:07:16.200 --> 0:07:19.240 Even worse than not liking each other. Their view of 0:07:19.320 --> 0:07:23.240 the world aren't based on the same facts or the 0:07:23.320 --> 0:07:28.520 same reality. Everybody knows about Facebook and to a lesser degree, 0:07:28.680 --> 0:07:33.360 TikTok and Instagram and how it very much bulkanized people 0:07:33.480 --> 0:07:37.480 into things. And we've seen that in the world of media. 0:07:37.600 --> 0:07:40.200 You have Fox News over here, in MSNBC over there. 0:07:41.640 --> 0:07:48.400 How significant of a threat does algorithmic news feeds present 0:07:48.640 --> 0:07:57.119 to the country as a democracy, self regulating, self determined democracy. 0:07:56.880 --> 0:08:01.320 Really significant? And there's algorithms and then there's large language models, 0:08:01.800 --> 0:08:05.440 and they can both be used to create situations in which, 0:08:05.560 --> 0:08:09.280 let's say, the people in some city let's call Los Angeles, 0:08:09.400 --> 0:08:13.000 are seeing stuff that creates a reality that's very different 0:08:13.400 --> 0:08:15.960 from the reality that people are saying, and let's say, 0:08:16.280 --> 0:08:20.200 boise Idaho. And that can be a real problem for 0:08:20.920 --> 0:08:25.480 understanding one another and also for mutual problem solving. 0:08:26.360 --> 0:08:29.880 So let's apply this a little bit more to consumers 0:08:29.960 --> 0:08:35.720 and markets. You describe two specific types of algorithmic discrimination. 0:08:36.520 --> 0:08:40.360 One is price discrimination and the other is quality discrimination. 0:08:41.520 --> 0:08:44.240 Why should we be aware of this distinction? Do they 0:08:44.400 --> 0:08:46.440 both deserve regulatory attention? 0:08:47.840 --> 0:08:52.480 So if there is price discrimination through algorithms in which 0:08:52.520 --> 0:08:56.120 different people get different offers depending on what the algorithm 0:08:56.280 --> 0:09:00.120 knows about their wealth and tastes, that's one thing, and 0:09:00.240 --> 0:09:03.160 it might be okay. People don't stand up and cheer 0:09:03.280 --> 0:09:06.079 and say hooray. But if people who have a lot 0:09:06.160 --> 0:09:09.800 of resources are given an offer that's not as let's 0:09:09.800 --> 0:09:14.160 say seductive, as one that is given to people who 0:09:14.160 --> 0:09:17.040 don't have a lot of resources, just because the price 0:09:17.160 --> 0:09:19.400 is higher for the roots than the poor, that's okay. 0:09:19.559 --> 0:09:24.720 There's something efficient and market friendly about that. If it's 0:09:24.800 --> 0:09:29.920 the case that people who are let's say, not caring 0:09:30.080 --> 0:09:33.280 much about whether a tennis racket is going to break 0:09:33.480 --> 0:09:36.840 after multiple uses, and other people who think that tennis 0:09:36.920 --> 0:09:40.640 racket really has to be solid because I play every day, 0:09:40.880 --> 0:09:42.640 and I'm going to play for the next five years. 0:09:43.080 --> 0:09:47.320 Then some people are given the let's say, immortal tennis racket, 0:09:47.440 --> 0:09:50.640 and other people are given the one that's more fragile. 0:09:51.040 --> 0:09:55.240 That's also okay, so long as we're dealing with people 0:09:55.280 --> 0:09:57.720 who have a level of sophistication. They know what they're 0:09:57.760 --> 0:10:00.600 getting and they know what they need. It's the case 0:10:00.720 --> 0:10:05.400 that for either pricing or for quality, the algorithm is 0:10:05.520 --> 0:10:09.960 aware of the fact that certain consumers are particularly likely 0:10:10.080 --> 0:10:14.480 not to have relevant information, then everything goes haywire. And 0:10:14.600 --> 0:10:18.640 if this isn't frightening enough, note that algorithms are an 0:10:18.760 --> 0:10:24.400 increasingly or an increasingly excellent position to know. This person 0:10:24.480 --> 0:10:28.040 with whom I'm dealing doesn't know a lot about whether 0:10:28.160 --> 0:10:31.160 products are going to last, and I can exploit that. 0:10:31.840 --> 0:10:34.880 Or this person is very focused on today and tomorrow 0:10:35.360 --> 0:10:38.400 and next year doesn't really matter. The person's present biased 0:10:38.559 --> 0:10:41.679 and I can exploit that. And that's something that can 0:10:42.040 --> 0:10:47.160 damage vulnerable consumers a lot, either with respect to quality 0:10:47.320 --> 0:10:48.480 or with respect to pricing. 0:10:48.960 --> 0:10:52.160 So let's flesh that out a little more. I'm very 0:10:52.280 --> 0:10:56.800 much aware that when Facebook sells ads, because I've been 0:10:56.920 --> 0:11:02.640 pitched these from Facebook. They could target an audience based 0:11:02.720 --> 0:11:06.800 on not just their likes and dislikes, but their geography, 0:11:07.000 --> 0:11:11.599 their search history, their credit score, their purchase history. Like, 0:11:11.840 --> 0:11:15.240 they know more about you than you know about yourself. 0:11:15.920 --> 0:11:21.280 It seems like we've created an opportunity for some potentially 0:11:21.679 --> 0:11:27.679 abusive behavior. Where is the line crossed from Hey, we 0:11:27.920 --> 0:11:30.640 know that you like dogs and so we're going to 0:11:30.760 --> 0:11:34.360 market dog food to you to we know everything there 0:11:34.440 --> 0:11:37.360 is about you, and we're going to exploit your behavior 0:11:37.480 --> 0:11:41.520 biases and some of your emotional weaknesses. 0:11:42.120 --> 0:11:46.520 Okay, So suppose there's a population of Facebook users who 0:11:46.640 --> 0:11:53.480 are super well informed about food and really rational about food. 0:11:54.080 --> 0:11:58.920 So they particularly happen to be fond of sushi, and 0:11:59.480 --> 0:12:03.240 Facebook is going hard at them with respect to offers 0:12:03.360 --> 0:12:08.520 for sushi and so forth. Now let's suppose there's another population, 0:12:09.080 --> 0:12:12.199 which is they know what they like about food, but 0:12:12.320 --> 0:12:18.320 they have kind of hopes and false beliefs both about 0:12:19.040 --> 0:12:22.040 the effect of food on health. Then you can really 0:12:22.160 --> 0:12:25.360 market to them things that will lead to poor choices. 0:12:25.840 --> 0:12:29.440 And I've made a stark distinction between fully rational which 0:12:29.520 --> 0:12:34.280 is kind of economic speak, and you know, imperfectly informed 0:12:34.320 --> 0:12:38.520 and behaviorally biased people. Also economic speak, but it's really intuitive. 0:12:38.920 --> 0:12:41.040 There's a radio show maybe This Will Bring It Home, 0:12:41.080 --> 0:12:43.719 that I listen to when I drive into work, and 0:12:43.880 --> 0:12:48.079 there's a lot of marketing about a product that is 0:12:48.120 --> 0:12:52.280 supposed to relieve pain. And I don't want to criticize 0:12:52.320 --> 0:12:55.599 any producer of any product, but I have reason to 0:12:55.679 --> 0:12:59.319 believe that the relevant product doesn't help much. But the 0:13:00.480 --> 0:13:04.400 the station that is marketing this product to people, this 0:13:04.600 --> 0:13:09.359 pain relief product, must know that the audience is vulnerable 0:13:09.679 --> 0:13:12.719 to it, and they must know exactly how to get 0:13:12.760 --> 0:13:15.840 at them. And that's not in the interest of that's 0:13:15.880 --> 0:13:17.640 not going to make America great again. 0:13:18.240 --> 0:13:23.000 To say the very least. So we've been talking about algorithms, 0:13:23.760 --> 0:13:29.000 but obviously the subtext is artificial intelligence, which seems to 0:13:29.040 --> 0:13:33.839 be the natural extension and for the development of algos 0:13:34.480 --> 0:13:39.239 tell us how as AI becomes more sophisticated and pervasive, 0:13:39.760 --> 0:13:43.520 how is this going to impact our lives as employees, 0:13:43.600 --> 0:13:46.079 as consumers, as citizens. 0:13:48.440 --> 0:13:54.439 Chat GPT, chances are, knows a lot about everyone who 0:13:54.640 --> 0:13:58.719 uses set. So I actually asked chat GPT recently, I 0:13:58.840 --> 0:14:01.559 use it some not huge. I asked it to say 0:14:01.640 --> 0:14:05.360 some things about myself, and it said a few things 0:14:05.480 --> 0:14:10.160 that were kind of scarily precise about me based on 0:14:10.640 --> 0:14:13.719 some number dozens, not one hundreds. I don't think of 0:14:13.880 --> 0:14:18.240 engagements with chatch ept. So large language models that track 0:14:18.360 --> 0:14:22.360 your prompts can know a lot about you, and if 0:14:22.560 --> 0:14:26.440 they're able also to know your name, they can, you know, instantly, 0:14:26.560 --> 0:14:30.080 basically learn a ton about you online, and we need 0:14:30.160 --> 0:14:33.880 to have privacy protections that are working there. Still, it's 0:14:33.960 --> 0:14:38.400 the case that AI broadly is able to use algorithms, 0:14:38.720 --> 0:14:42.560 and generative AI can go well beyond the algorithms we've 0:14:42.600 --> 0:14:48.640 gotten familiar with both to make the beauty of algorithmic engagement, 0:14:48.800 --> 0:14:52.200 that is, here's what you like, here's what you want, 0:14:52.360 --> 0:14:56.560 we're going to help you, and the ugliness of algorithms, 0:14:57.280 --> 0:14:59.840 here's how we can exploit you to get you to buy. 0:15:00.640 --> 0:15:03.200 And of course I'm thinking of investments too, so in 0:15:03.280 --> 0:15:05.200 your neck of the woods, it would be a child's 0:15:05.200 --> 0:15:09.600 play to get people super excited about investments, which AI 0:15:09.840 --> 0:15:14.240 knows the people with whom it's engaging are particularly susceptible to, 0:15:14.720 --> 0:15:17.320 even though they're really dumb, engagements. 0:15:17.480 --> 0:15:23.440 Really really interesting. So since we're talking about investing, I 0:15:23.600 --> 0:15:28.800 can't help but bring up both AI and algorithms trying 0:15:28.920 --> 0:15:33.920 to increase so called market efficiency, and I always go 0:15:34.200 --> 0:15:38.080 back to Uber's surge pricing. As soon as it starts 0:15:38.160 --> 0:15:42.160 to rain, the prices go up in the city. It's 0:15:42.360 --> 0:15:47.360 obviously not an emergency. It's just an annoyance. However, we 0:15:47.520 --> 0:15:52.160 do see situations of price gouging after a storm, after 0:15:52.240 --> 0:15:54.760 a hurricane, people only have so many batteries and so 0:15:54.960 --> 0:15:58.680 much plywood, and they kind of crank up prices. How 0:15:58.760 --> 0:16:02.400 do we determine what is the line between something like 0:16:02.680 --> 0:16:07.600 surge pricing and something like, you know, abusive price gouging. 0:16:08.160 --> 0:16:12.560 Okay, so you're in a terrific area of behavioral economics. 0:16:12.880 --> 0:16:16.800 So we know that in circumstances in which let's say 0:16:16.920 --> 0:16:22.360 demand goes up high because everyone needs a shovel and 0:16:22.400 --> 0:16:26.600 as a snowstorm, people are really mad if the prices 0:16:26.720 --> 0:16:30.280 go up, though it might be just a sensible market adjustment. 0:16:31.000 --> 0:16:35.600 So as a first approximation, if there's a spectacular need 0:16:35.760 --> 0:16:41.360 for something, let's say, shovels or umbrellas, the market inflation 0:16:41.560 --> 0:16:44.800 of the cost, while it's morally abhorrent to many, and 0:16:44.920 --> 0:16:49.520 maybe in principle morally abhorrent from the standpoint of standard economics, 0:16:49.760 --> 0:16:53.480 it's okay. Now, if it's the case that people under 0:16:53.920 --> 0:16:56.640 short term pressure from the fact that there's a lot 0:16:56.720 --> 0:17:02.320 of rain are especially vulnerable and some kind of emotionally 0:17:02.440 --> 0:17:06.680 intense state, so they'll pay kind of anything for an umbrella, 0:17:07.480 --> 0:17:11.880 then there's a behavioral bias which is motivating people's willingness 0:17:11.920 --> 0:17:14.040 to pay a lot more than the product is worth. 0:17:14.840 --> 0:17:18.119 So let's talk a little bit about disclosures and the 0:17:18.240 --> 0:17:21.880 sort of mandates that are required when we look across 0:17:21.960 --> 0:17:24.840 the pond, when we look at Europe, they're much more 0:17:24.880 --> 0:17:29.480 aggressive about protecting privacy and making sure big tech companies 0:17:29.520 --> 0:17:32.640 are disclosing all the things they have to disclose. How 0:17:32.760 --> 0:17:36.840 far behind is the US and that generally and are 0:17:36.960 --> 0:17:40.399 we behind when it comes to disclosures about algorithms or 0:17:40.480 --> 0:17:42.000 AI I. 0:17:42.040 --> 0:17:45.560 Think we're behind them in the sense that we're less 0:17:45.640 --> 0:17:50.200 privacy focused. But it's not clear that that's bad, And 0:17:50.359 --> 0:17:53.920 even if it isn't good, it's not clear that it's terrible. 0:17:54.560 --> 0:17:57.960 I think neither Europe nor the US has put their 0:17:58.119 --> 0:18:03.280 regulatory finger on the actual problem. So let's take the 0:18:03.400 --> 0:18:08.600 problem of algorithms not figuring out what people want, but 0:18:08.760 --> 0:18:13.359 algorithms exploiting a lack of information or a behavioral bias 0:18:13.800 --> 0:18:16.600 to get people to buy things at prices that aren't 0:18:16.600 --> 0:18:19.880 good for them. That's a problem. It's in the same 0:18:20.040 --> 0:18:23.040 universe as fraud and deception, and the question is what 0:18:23.119 --> 0:18:25.480 are we going to do about it. A first line 0:18:25.520 --> 0:18:29.520 of defense is to try to ensure consumer protection, not 0:18:29.720 --> 0:18:32.960 through heavy handed regulation. I'm a long time University of 0:18:33.040 --> 0:18:37.560 Chicago person. I have in my DNA not liking heavy 0:18:37.600 --> 0:18:42.600 handed regulation, but through helping people to know what they're 0:18:42.640 --> 0:18:47.000 buying and helping people not to suffer from a behavioral 0:18:47.119 --> 0:18:52.160 bias such as, let's say, incomplete attention or unrealistic optimism 0:18:52.200 --> 0:18:55.600 when they're buying things. So these are standard consumer protection 0:18:55.800 --> 0:18:59.120 things which many of our agencies in the US, home 0:18:59.160 --> 0:19:03.840 grown America. They've done that, and that's good and we 0:19:04.000 --> 0:19:06.639 need more of that. So that's first line of defense. 0:19:07.080 --> 0:19:12.120 Second line of defense isn't to say, you know, privacy, privacy, privacy, 0:19:12.240 --> 0:19:14.600 though maybe that's a good song to sing. It's to 0:19:14.680 --> 0:19:19.000 say right to algorithmic transparency. So this is something which 0:19:19.119 --> 0:19:23.520 neither the US, nor Europe, nor Asia, nor South America 0:19:23.800 --> 0:19:27.359 nor Africa has been very advanced on. So this is 0:19:27.440 --> 0:19:29.800 a coming thing where we need to know what the 0:19:29.880 --> 0:19:34.960 algorithms are doing. So it's public what's Amazon's algorithm doing? 0:19:35.400 --> 0:19:37.879 That would be good to know, and it shouldn't be 0:19:38.000 --> 0:19:43.480 the case that some efforts to ensure transparency invade Amazon's 0:19:43.600 --> 0:19:44.480 legitimate rights. 0:19:45.760 --> 0:19:51.080 Really, really fascinating, thanks Cas. Anybody who is participating in 0:19:51.240 --> 0:19:56.920 the American economy and society, consumers, investors, even just regular 0:19:57.040 --> 0:20:01.440 readers of news, needs to be away of how algorithms 0:20:01.480 --> 0:20:05.480 are affecting what they see, the prices they pay, and 0:20:05.560 --> 0:20:09.000 the sort of information they're getting. So with a little 0:20:09.040 --> 0:20:13.520 bit of forethought and the book Algorithmic Harm, you can 0:20:13.680 --> 0:20:18.760 protect yourself from the worst aspects of algorithms. And AI, 0:20:19.640 --> 0:20:23.720 I'm Barry Redults you're listening to Bloomberg's at the Money