1
00:00:02,520 --> 00:00:17,280
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:26,239 --> 00:00:30,760
Speaker 2: Algorithms are everywhere. They determine the price you pay for

3
00:00:30,800 --> 00:00:35,360
Speaker 2: your Uber, what gets fed to you on TikTok and Instagram,

4
00:00:35,800 --> 00:00:40,639
Speaker 2: and even the prices you pay in the supermarket. Is

5
00:00:40,680 --> 00:00:45,680
Speaker 2: all of this algorithmic impact helping or harming people? To

6
00:00:45,760 --> 00:00:49,360
Speaker 2: answer that question, let's bring in Cass Sunstein. He is

7
00:00:49,400 --> 00:00:53,280
Speaker 2: the author of a new book, Algorithmic Harm, Protecting People

8
00:00:53,360 --> 00:00:58,520
Speaker 2: in the Age of Artificial Intelligence, co written with ornbar Gil.

9
00:00:58,920 --> 00:01:02,600
Speaker 2: Cass is also a professor at Harvard Law School and

10
00:01:02,720 --> 00:01:06,520
Speaker 2: is perhaps best known for his books on Star Wars

11
00:01:06,720 --> 00:01:12,080
Speaker 2: and co authoring Nudge with Nobel Laureate Dick Taylor. So, pass,

12
00:01:12,160 --> 00:01:16,440
Speaker 2: let's just jump right into this and start by defining

13
00:01:16,920 --> 00:01:18,960
Speaker 2: what is algorithmic harm?

14
00:01:20,160 --> 00:01:23,760
Speaker 3: Okay, So let's use Star Wars. So let's say the

15
00:01:23,880 --> 00:01:28,840
Speaker 3: Jedi Knights use algorithms, and they give people things that

16
00:01:28,920 --> 00:01:32,639
Speaker 3: fit with their tastes and interests and information, and people

17
00:01:32,720 --> 00:01:35,920
Speaker 3: get If they're interested in books on behavioral economics, that's

18
00:01:35,959 --> 00:01:38,480
Speaker 3: what they get at a price that suits them. If

19
00:01:38,480 --> 00:01:41,280
Speaker 3: they're interested in a book on Star Wars, that's what

20
00:01:41,319 --> 00:01:44,240
Speaker 3: they get at a price that suits them. The Sith

21
00:01:44,319 --> 00:01:48,760
Speaker 3: by contrast, take advantage with algorithms of the fact that

22
00:01:48,840 --> 00:01:54,160
Speaker 3: some consumers lack information and some consumers suffer from behavioral biases.

23
00:01:54,440 --> 00:01:57,320
Speaker 3: So we're going to focus on consumers first. If people

24
00:01:57,560 --> 00:02:01,200
Speaker 3: don't know much, let's say about healthcare product, an algorithm

25
00:02:01,280 --> 00:02:03,720
Speaker 3: might know that that they're likely not to know much

26
00:02:04,480 --> 00:02:07,360
Speaker 3: and might say, we have a fantastic baldness cure for you.

27
00:02:08,200 --> 00:02:11,760
Speaker 3: Here it goes, and people will be duped and exploited.

28
00:02:12,040 --> 00:02:16,120
Speaker 3: So that's exploitation of absence of information. That's algorithmic harm.

29
00:02:16,360 --> 00:02:19,520
Speaker 3: If people are super optimistic and they think that some

30
00:02:19,680 --> 00:02:22,240
Speaker 3: new product is going to last forever, when it tends

31
00:02:22,240 --> 00:02:26,200
Speaker 3: to break on first usage, then the algorithm can know

32
00:02:26,240 --> 00:02:30,280
Speaker 3: those are unrealistically optimistic people and exploit their behavioral bias.

33
00:02:31,440 --> 00:02:37,680
Speaker 2: So I referenced a few obvious areas where algorithms are

34
00:02:37,760 --> 00:02:42,640
Speaker 2: taking place. Uber pricing is one the books you see

35
00:02:43,040 --> 00:02:48,280
Speaker 2: on Amazon is algorithmically driven. Clearly, a lot of social

36
00:02:48,360 --> 00:02:53,240
Speaker 2: media for better or WORSEUS algorithmically driven, and even things

37
00:02:53,320 --> 00:02:56,600
Speaker 2: like the sort of music you like on Pandora. What

38
00:02:56,680 --> 00:03:01,960
Speaker 2: are some of the less obvious examples of how algorithms

39
00:03:02,080 --> 00:03:06,400
Speaker 2: are affecting consumers and regular people? Every day.

40
00:03:07,320 --> 00:03:10,120
Speaker 3: Okay, So let's start with us straightforward once and then

41
00:03:10,160 --> 00:03:13,560
Speaker 3: we'll get a little subtle. So straightforwardly, it might be

42
00:03:13,680 --> 00:03:17,120
Speaker 3: that people are being asked to pay a price that

43
00:03:18,000 --> 00:03:21,160
Speaker 3: suits their economic situation. So if you have a lot

44
00:03:21,160 --> 00:03:23,639
Speaker 3: of money, the algorithm knows that maybe the price will

45
00:03:23,680 --> 00:03:26,080
Speaker 3: be twice as much as it would be if you

46
00:03:26,160 --> 00:03:31,080
Speaker 3: were less wealthy. That I think is basically okay. It

47
00:03:31,160 --> 00:03:34,239
Speaker 3: leads to greater efficiency in the system. It's like rich

48
00:03:34,320 --> 00:03:38,000
Speaker 3: people will pay more for the same product than poor people,

49
00:03:38,040 --> 00:03:41,160
Speaker 3: and the algorithm is aware of that. So that's not

50
00:03:41,280 --> 00:03:44,600
Speaker 3: that subtle. But it's important. Also, not that subtle is

51
00:03:45,080 --> 00:03:49,640
Speaker 3: targeting people based on what's known about their particular tastes

52
00:03:49,680 --> 00:03:52,600
Speaker 3: and preferences. Let's put wealth to one side, and so

53
00:03:52,680 --> 00:03:55,720
Speaker 3: it's known that certain people are super interested in dogs,

54
00:03:55,760 --> 00:03:58,440
Speaker 3: other people are interested in cats. And there we go,

55
00:03:59,000 --> 00:04:02,640
Speaker 3: and all that is very straightforward's happening. If consumers are

56
00:04:02,680 --> 00:04:07,240
Speaker 3: sophisticated and knowledgeable, that can be a great thing to

57
00:04:07,320 --> 00:04:10,240
Speaker 3: make markets work better. If they aren't, it can be

58
00:04:10,280 --> 00:04:14,200
Speaker 3: a terrible thing to make consumers get manipulated and hurt.

59
00:04:14,680 --> 00:04:18,159
Speaker 3: Here's something a little more subtle. If an algorithm knows

60
00:04:18,200 --> 00:04:22,400
Speaker 3: for example, that you like Olivia Rodrigo, and I hope

61
00:04:22,440 --> 00:04:24,840
Speaker 3: you do, because she's really good, then there are going

62
00:04:24,920 --> 00:04:27,400
Speaker 3: to be a lot of Olivia or Rodrigo songs that

63
00:04:27,440 --> 00:04:30,040
Speaker 3: are going to be put into your system. And let's

64
00:04:30,040 --> 00:04:33,159
Speaker 3: say no one's really like Olivia Rodrigo, but let's suppose

65
00:04:33,160 --> 00:04:36,000
Speaker 3: there are others who are vaguely like her, and you're

66
00:04:36,000 --> 00:04:38,320
Speaker 3: going to hear a lot of that. Now, that might

67
00:04:38,440 --> 00:04:42,000
Speaker 3: seem not like algorithmic harm, that might seem like a

68
00:04:42,040 --> 00:04:46,120
Speaker 3: triumph of freedom and markets, but it might mean that

69
00:04:46,240 --> 00:04:49,760
Speaker 3: piece of people's tastes will calcify, and we're going to

70
00:04:49,800 --> 00:04:54,080
Speaker 3: get very bulkanized culturally with respect to what people see

71
00:04:54,080 --> 00:04:56,679
Speaker 3: in here. So they're going to be Olivia or Arigo people,

72
00:04:56,720 --> 00:04:58,440
Speaker 3: and then they're going to be led Zeppelin people, and

73
00:04:58,480 --> 00:05:00,840
Speaker 3: they're going to be Frank Sinatra people. And there was

74
00:05:00,880 --> 00:05:03,360
Speaker 3: another singer called Bach. I guess I don't know much

75
00:05:03,360 --> 00:05:06,560
Speaker 3: about him, but there's Bach, and there would be Bach people.

76
00:05:06,960 --> 00:05:11,359
Speaker 3: And that's culturally damaging, and it's also damaging for the

77
00:05:11,440 --> 00:05:14,360
Speaker 3: development of individual tastes and preferences.

78
00:05:14,960 --> 00:05:19,400
Speaker 2: So let's put this into a little broader context than

79
00:05:19,680 --> 00:05:23,280
Speaker 2: simply musical tastes and I like all of those, so

80
00:05:23,560 --> 00:05:28,760
Speaker 2: I haven't become bulkanized yet. But when we look at

81
00:05:29,040 --> 00:05:32,920
Speaker 2: consumption of news media, when we look at consumption of information,

82
00:05:33,800 --> 00:05:37,960
Speaker 2: it really seems like the country has self divided itself

83
00:05:38,360 --> 00:05:42,560
Speaker 2: into these happy little media bubbles that are either far

84
00:05:42,680 --> 00:05:45,520
Speaker 2: left leaning or far right leaning, which are is kind

85
00:05:45,520 --> 00:05:48,440
Speaker 2: of weird because I always learned the bulk of the

86
00:05:48,480 --> 00:05:53,360
Speaker 2: country and the traditional bell curve, most people are somewhere

87
00:05:53,680 --> 00:05:56,279
Speaker 2: in the middle. Hey, maybe they're center right or center left,

88
00:05:56,600 --> 00:05:59,839
Speaker 2: but they're not out on the tails. How does the

89
00:06:00,000 --> 00:06:05,719
Speaker 2: these algorithms affect our consumption of news and information?

90
00:06:07,560 --> 00:06:10,320
Speaker 3: About fifteen twenty years ago, there is a lot of

91
00:06:10,400 --> 00:06:16,159
Speaker 3: concern that through individual choices, people would create echo chambers

92
00:06:16,240 --> 00:06:19,440
Speaker 3: in which they would live. And that's a fair concern

93
00:06:19,600 --> 00:06:23,520
Speaker 3: and it has created a number of, let's say, challenges

94
00:06:23,760 --> 00:06:29,120
Speaker 3: for self government and learning. What you're pointing to is

95
00:06:29,160 --> 00:06:33,839
Speaker 3: also emphasized in the book, which is that algorithms can

96
00:06:34,040 --> 00:06:38,040
Speaker 3: echo chamber you. An algorithm might say, you know, you're

97
00:06:38,279 --> 00:06:42,000
Speaker 3: keenly interested in immigration, and you have this point of view,

98
00:06:42,480 --> 00:06:44,679
Speaker 3: so boy, are we going to funnel to you lots

99
00:06:44,680 --> 00:06:48,039
Speaker 3: of information because clicks are money, and you're going to

100
00:06:48,040 --> 00:06:50,720
Speaker 3: be clicking, clicking, clickling, click kicking, And that might be

101
00:06:50,839 --> 00:06:53,960
Speaker 3: a very good thing from the standpoint of the seller,

102
00:06:54,080 --> 00:06:56,480
Speaker 3: so to speak, or the user of the algorithm, but

103
00:06:56,600 --> 00:07:00,200
Speaker 3: from the standpoint of view, it's not so fantastic. And

104
00:07:00,320 --> 00:07:03,920
Speaker 3: from the standpoint of our society it's less than not

105
00:07:04,080 --> 00:07:08,400
Speaker 3: so fantastic, because people will be living in algorithm driven

106
00:07:09,080 --> 00:07:13,600
Speaker 3: universes that are very separate from one another, and they

107
00:07:13,680 --> 00:07:15,680
Speaker 3: can end up not liking each other very much.

108
00:07:16,200 --> 00:07:19,240
Speaker 2: Even worse than not liking each other. Their view of

109
00:07:19,320 --> 00:07:23,240
Speaker 2: the world aren't based on the same facts or the

110
00:07:23,320 --> 00:07:28,520
Speaker 2: same reality. Everybody knows about Facebook and to a lesser degree,

111
00:07:28,680 --> 00:07:33,360
Speaker 2: TikTok and Instagram and how it very much bulkanized people

112
00:07:33,480 --> 00:07:37,480
Speaker 2: into things. And we've seen that in the world of media.

113
00:07:37,600 --> 00:07:40,200
Speaker 2: You have Fox News over here, in MSNBC over there.

114
00:07:41,640 --> 00:07:48,400
Speaker 2: How significant of a threat does algorithmic news feeds present

115
00:07:48,640 --> 00:07:57,119
Speaker 2: to the country as a democracy, self regulating, self determined democracy.

116
00:07:56,880 --> 00:08:01,320
Speaker 3: Really significant? And there's algorithms and then there's large language models,

117
00:08:01,800 --> 00:08:05,440
Speaker 3: and they can both be used to create situations in which,

118
00:08:05,560 --> 00:08:09,280
Speaker 3: let's say, the people in some city let's call Los Angeles,

119
00:08:09,400 --> 00:08:13,000
Speaker 3: are seeing stuff that creates a reality that's very different

120
00:08:13,400 --> 00:08:15,960
Speaker 3: from the reality that people are saying, and let's say,

121
00:08:16,280 --> 00:08:20,200
Speaker 3: boise Idaho. And that can be a real problem for

122
00:08:20,920 --> 00:08:25,480
Speaker 3: understanding one another and also for mutual problem solving.

123
00:08:26,360 --> 00:08:29,880
Speaker 2: So let's apply this a little bit more to consumers

124
00:08:29,960 --> 00:08:35,720
Speaker 2: and markets. You describe two specific types of algorithmic discrimination.

125
00:08:36,520 --> 00:08:40,360
Speaker 2: One is price discrimination and the other is quality discrimination.

126
00:08:41,520 --> 00:08:44,240
Speaker 2: Why should we be aware of this distinction? Do they

127
00:08:44,400 --> 00:08:46,440
Speaker 2: both deserve regulatory attention?

128
00:08:47,840 --> 00:08:52,480
Speaker 3: So if there is price discrimination through algorithms in which

129
00:08:52,520 --> 00:08:56,120
Speaker 3: different people get different offers depending on what the algorithm

130
00:08:56,280 --> 00:09:00,120
Speaker 3: knows about their wealth and tastes, that's one thing, and

131
00:09:00,240 --> 00:09:03,160
Speaker 3: it might be okay. People don't stand up and cheer

132
00:09:03,280 --> 00:09:06,079
Speaker 3: and say hooray. But if people who have a lot

133
00:09:06,160 --> 00:09:09,800
Speaker 3: of resources are given an offer that's not as let's

134
00:09:09,800 --> 00:09:14,160
Speaker 3: say seductive, as one that is given to people who

135
00:09:14,160 --> 00:09:17,040
Speaker 3: don't have a lot of resources, just because the price

136
00:09:17,160 --> 00:09:19,400
Speaker 3: is higher for the roots than the poor, that's okay.

137
00:09:19,559 --> 00:09:24,720
Speaker 3: There's something efficient and market friendly about that. If it's

138
00:09:24,800 --> 00:09:29,920
Speaker 3: the case that people who are let's say, not caring

139
00:09:30,080 --> 00:09:33,280
Speaker 3: much about whether a tennis racket is going to break

140
00:09:33,480 --> 00:09:36,840
Speaker 3: after multiple uses, and other people who think that tennis

141
00:09:36,920 --> 00:09:40,640
Speaker 3: racket really has to be solid because I play every day,

142
00:09:40,880 --> 00:09:42,640
Speaker 3: and I'm going to play for the next five years.

143
00:09:43,080 --> 00:09:47,320
Speaker 3: Then some people are given the let's say, immortal tennis racket,

144
00:09:47,440 --> 00:09:50,640
Speaker 3: and other people are given the one that's more fragile.

145
00:09:51,040 --> 00:09:55,240
Speaker 3: That's also okay, so long as we're dealing with people

146
00:09:55,280 --> 00:09:57,720
Speaker 3: who have a level of sophistication. They know what they're

147
00:09:57,760 --> 00:10:00,600
Speaker 3: getting and they know what they need. It's the case

148
00:10:00,720 --> 00:10:05,400
Speaker 3: that for either pricing or for quality, the algorithm is

149
00:10:05,520 --> 00:10:09,960
Speaker 3: aware of the fact that certain consumers are particularly likely

150
00:10:10,080 --> 00:10:14,480
Speaker 3: not to have relevant information, then everything goes haywire. And

151
00:10:14,600 --> 00:10:18,640
Speaker 3: if this isn't frightening enough, note that algorithms are an

152
00:10:18,760 --> 00:10:24,400
Speaker 3: increasingly or an increasingly excellent position to know. This person

153
00:10:24,480 --> 00:10:28,040
Speaker 3: with whom I'm dealing doesn't know a lot about whether

154
00:10:28,160 --> 00:10:31,160
Speaker 3: products are going to last, and I can exploit that.

155
00:10:31,840 --> 00:10:34,880
Speaker 3: Or this person is very focused on today and tomorrow

156
00:10:35,360 --> 00:10:38,400
Speaker 3: and next year doesn't really matter. The person's present biased

157
00:10:38,559 --> 00:10:41,679
Speaker 3: and I can exploit that. And that's something that can

158
00:10:42,040 --> 00:10:47,160
Speaker 3: damage vulnerable consumers a lot, either with respect to quality

159
00:10:47,320 --> 00:10:48,480
Speaker 3: or with respect to pricing.

160
00:10:48,960 --> 00:10:52,160
Speaker 2: So let's flesh that out a little more. I'm very

161
00:10:52,280 --> 00:10:56,800
Speaker 2: much aware that when Facebook sells ads, because I've been

162
00:10:56,920 --> 00:11:02,640
Speaker 2: pitched these from Facebook. They could target an audience based

163
00:11:02,720 --> 00:11:06,800
Speaker 2: on not just their likes and dislikes, but their geography,

164
00:11:07,000 --> 00:11:11,599
Speaker 2: their search history, their credit score, their purchase history. Like,

165
00:11:11,840 --> 00:11:15,240
Speaker 2: they know more about you than you know about yourself.

166
00:11:15,920 --> 00:11:21,280
Speaker 2: It seems like we've created an opportunity for some potentially

167
00:11:21,679 --> 00:11:27,679
Speaker 2: abusive behavior. Where is the line crossed from Hey, we

168
00:11:27,920 --> 00:11:30,640
Speaker 2: know that you like dogs and so we're going to

169
00:11:30,760 --> 00:11:34,360
Speaker 2: market dog food to you to we know everything there

170
00:11:34,440 --> 00:11:37,360
Speaker 2: is about you, and we're going to exploit your behavior

171
00:11:37,480 --> 00:11:41,520
Speaker 2: biases and some of your emotional weaknesses.

172
00:11:42,120 --> 00:11:46,520
Speaker 3: Okay, So suppose there's a population of Facebook users who

173
00:11:46,640 --> 00:11:53,480
Speaker 3: are super well informed about food and really rational about food.

174
00:11:54,080 --> 00:11:58,920
Speaker 3: So they particularly happen to be fond of sushi, and

175
00:11:59,480 --> 00:12:03,240
Speaker 3: Facebook is going hard at them with respect to offers

176
00:12:03,360 --> 00:12:08,520
Speaker 3: for sushi and so forth. Now let's suppose there's another population,

177
00:12:09,080 --> 00:12:12,199
Speaker 3: which is they know what they like about food, but

178
00:12:12,320 --> 00:12:18,320
Speaker 3: they have kind of hopes and false beliefs both about

179
00:12:19,040 --> 00:12:22,040
Speaker 3: the effect of food on health. Then you can really

180
00:12:22,160 --> 00:12:25,360
Speaker 3: market to them things that will lead to poor choices.

181
00:12:25,840 --> 00:12:29,440
Speaker 3: And I've made a stark distinction between fully rational which

182
00:12:29,520 --> 00:12:34,280
Speaker 3: is kind of economic speak, and you know, imperfectly informed

183
00:12:34,320 --> 00:12:38,520
Speaker 3: and behaviorally biased people. Also economic speak, but it's really intuitive.

184
00:12:38,920 --> 00:12:41,040
Speaker 3: There's a radio show maybe This Will Bring It Home,

185
00:12:41,080 --> 00:12:43,719
Speaker 3: that I listen to when I drive into work, and

186
00:12:43,880 --> 00:12:48,079
Speaker 3: there's a lot of marketing about a product that is

187
00:12:48,120 --> 00:12:52,280
Speaker 3: supposed to relieve pain. And I don't want to criticize

188
00:12:52,320 --> 00:12:55,599
Speaker 3: any producer of any product, but I have reason to

189
00:12:55,679 --> 00:12:59,319
Speaker 3: believe that the relevant product doesn't help much. But the

190
00:13:00,480 --> 00:13:04,400
Speaker 3: the station that is marketing this product to people, this

191
00:13:04,600 --> 00:13:09,359
Speaker 3: pain relief product, must know that the audience is vulnerable

192
00:13:09,679 --> 00:13:12,719
Speaker 3: to it, and they must know exactly how to get

193
00:13:12,760 --> 00:13:15,840
Speaker 3: at them. And that's not in the interest of that's

194
00:13:15,880 --> 00:13:17,640
Speaker 3: not going to make America great again.

195
00:13:18,240 --> 00:13:23,000
Speaker 2: To say the very least. So we've been talking about algorithms,

196
00:13:23,760 --> 00:13:29,000
Speaker 2: but obviously the subtext is artificial intelligence, which seems to

197
00:13:29,040 --> 00:13:33,839
Speaker 2: be the natural extension and for the development of algos

198
00:13:34,480 --> 00:13:39,239
Speaker 2: tell us how as AI becomes more sophisticated and pervasive,

199
00:13:39,760 --> 00:13:43,520
Speaker 2: how is this going to impact our lives as employees,

200
00:13:43,600 --> 00:13:46,079
Speaker 2: as consumers, as citizens.

201
00:13:48,440 --> 00:13:54,439
Speaker 3: Chat GPT, chances are, knows a lot about everyone who

202
00:13:54,640 --> 00:13:58,719
Speaker 3: uses set. So I actually asked chat GPT recently, I

203
00:13:58,840 --> 00:14:01,559
Speaker 3: use it some not huge. I asked it to say

204
00:14:01,640 --> 00:14:05,360
Speaker 3: some things about myself, and it said a few things

205
00:14:05,480 --> 00:14:10,160
Speaker 3: that were kind of scarily precise about me based on

206
00:14:10,640 --> 00:14:13,719
Speaker 3: some number dozens, not one hundreds. I don't think of

207
00:14:13,880 --> 00:14:18,240
Speaker 3: engagements with chatch ept. So large language models that track

208
00:14:18,360 --> 00:14:22,360
Speaker 3: your prompts can know a lot about you, and if

209
00:14:22,560 --> 00:14:26,440
Speaker 3: they're able also to know your name, they can, you know, instantly,

210
00:14:26,560 --> 00:14:30,080
Speaker 3: basically learn a ton about you online, and we need

211
00:14:30,160 --> 00:14:33,880
Speaker 3: to have privacy protections that are working there. Still, it's

212
00:14:33,960 --> 00:14:38,400
Speaker 3: the case that AI broadly is able to use algorithms,

213
00:14:38,720 --> 00:14:42,560
Speaker 3: and generative AI can go well beyond the algorithms we've

214
00:14:42,600 --> 00:14:48,640
Speaker 3: gotten familiar with both to make the beauty of algorithmic engagement,

215
00:14:48,800 --> 00:14:52,200
Speaker 3: that is, here's what you like, here's what you want,

216
00:14:52,360 --> 00:14:56,560
Speaker 3: we're going to help you, and the ugliness of algorithms,

217
00:14:57,280 --> 00:14:59,840
Speaker 3: here's how we can exploit you to get you to buy.

218
00:15:00,640 --> 00:15:03,200
Speaker 3: And of course I'm thinking of investments too, so in

219
00:15:03,280 --> 00:15:05,200
Speaker 3: your neck of the woods, it would be a child's

220
00:15:05,200 --> 00:15:09,600
Speaker 3: play to get people super excited about investments, which AI

221
00:15:09,840 --> 00:15:14,240
Speaker 3: knows the people with whom it's engaging are particularly susceptible to,

222
00:15:14,720 --> 00:15:17,320
Speaker 3: even though they're really dumb, engagements.

223
00:15:17,480 --> 00:15:23,440
Speaker 2: Really really interesting. So since we're talking about investing, I

224
00:15:23,600 --> 00:15:28,800
Speaker 2: can't help but bring up both AI and algorithms trying

225
00:15:28,920 --> 00:15:33,920
Speaker 2: to increase so called market efficiency, and I always go

226
00:15:34,200 --> 00:15:38,080
Speaker 2: back to Uber's surge pricing. As soon as it starts

227
00:15:38,160 --> 00:15:42,160
Speaker 2: to rain, the prices go up in the city. It's

228
00:15:42,360 --> 00:15:47,360
Speaker 2: obviously not an emergency. It's just an annoyance. However, we

229
00:15:47,520 --> 00:15:52,160
Speaker 2: do see situations of price gouging after a storm, after

230
00:15:52,240 --> 00:15:54,760
Speaker 2: a hurricane, people only have so many batteries and so

231
00:15:54,960 --> 00:15:58,680
Speaker 2: much plywood, and they kind of crank up prices. How

232
00:15:58,760 --> 00:16:02,400
Speaker 2: do we determine what is the line between something like

233
00:16:02,680 --> 00:16:07,600
Speaker 2: surge pricing and something like, you know, abusive price gouging.

234
00:16:08,160 --> 00:16:12,560
Speaker 3: Okay, so you're in a terrific area of behavioral economics.

235
00:16:12,880 --> 00:16:16,800
Speaker 3: So we know that in circumstances in which let's say

236
00:16:16,920 --> 00:16:22,360
Speaker 3: demand goes up high because everyone needs a shovel and

237
00:16:22,400 --> 00:16:26,600
Speaker 3: as a snowstorm, people are really mad if the prices

238
00:16:26,720 --> 00:16:30,280
Speaker 3: go up, though it might be just a sensible market adjustment.

239
00:16:31,000 --> 00:16:35,600
Speaker 3: So as a first approximation, if there's a spectacular need

240
00:16:35,760 --> 00:16:41,360
Speaker 3: for something, let's say, shovels or umbrellas, the market inflation

241
00:16:41,560 --> 00:16:44,800
Speaker 3: of the cost, while it's morally abhorrent to many, and

242
00:16:44,920 --> 00:16:49,520
Speaker 3: maybe in principle morally abhorrent from the standpoint of standard economics,

243
00:16:49,760 --> 00:16:53,480
Speaker 3: it's okay. Now, if it's the case that people under

244
00:16:53,920 --> 00:16:56,640
Speaker 3: short term pressure from the fact that there's a lot

245
00:16:56,720 --> 00:17:02,320
Speaker 3: of rain are especially vulnerable and some kind of emotionally

246
00:17:02,440 --> 00:17:06,680
Speaker 3: intense state, so they'll pay kind of anything for an umbrella,

247
00:17:07,480 --> 00:17:11,880
Speaker 3: then there's a behavioral bias which is motivating people's willingness

248
00:17:11,920 --> 00:17:14,040
Speaker 3: to pay a lot more than the product is worth.

249
00:17:14,840 --> 00:17:18,119
Speaker 2: So let's talk a little bit about disclosures and the

250
00:17:18,240 --> 00:17:21,880
Speaker 2: sort of mandates that are required when we look across

251
00:17:21,960 --> 00:17:24,840
Speaker 2: the pond, when we look at Europe, they're much more

252
00:17:24,880 --> 00:17:29,480
Speaker 2: aggressive about protecting privacy and making sure big tech companies

253
00:17:29,520 --> 00:17:32,640
Speaker 2: are disclosing all the things they have to disclose. How

254
00:17:32,760 --> 00:17:36,840
Speaker 2: far behind is the US and that generally and are

255
00:17:36,960 --> 00:17:40,399
Speaker 2: we behind when it comes to disclosures about algorithms or

256
00:17:40,480 --> 00:17:42,000
Speaker 2: AI I.

257
00:17:42,040 --> 00:17:45,560
Speaker 3: Think we're behind them in the sense that we're less

258
00:17:45,640 --> 00:17:50,200
Speaker 3: privacy focused. But it's not clear that that's bad, And

259
00:17:50,359 --> 00:17:53,920
Speaker 3: even if it isn't good, it's not clear that it's terrible.

260
00:17:54,560 --> 00:17:57,960
Speaker 3: I think neither Europe nor the US has put their

261
00:17:58,119 --> 00:18:03,280
Speaker 3: regulatory finger on the actual problem. So let's take the

262
00:18:03,400 --> 00:18:08,600
Speaker 3: problem of algorithms not figuring out what people want, but

263
00:18:08,760 --> 00:18:13,359
Speaker 3: algorithms exploiting a lack of information or a behavioral bias

264
00:18:13,800 --> 00:18:16,600
Speaker 3: to get people to buy things at prices that aren't

265
00:18:16,600 --> 00:18:19,880
Speaker 3: good for them. That's a problem. It's in the same

266
00:18:20,040 --> 00:18:23,040
Speaker 3: universe as fraud and deception, and the question is what

267
00:18:23,119 --> 00:18:25,480
Speaker 3: are we going to do about it. A first line

268
00:18:25,520 --> 00:18:29,520
Speaker 3: of defense is to try to ensure consumer protection, not

269
00:18:29,720 --> 00:18:32,960
Speaker 3: through heavy handed regulation. I'm a long time University of

270
00:18:33,040 --> 00:18:37,560
Speaker 3: Chicago person. I have in my DNA not liking heavy

271
00:18:37,600 --> 00:18:42,600
Speaker 3: handed regulation, but through helping people to know what they're

272
00:18:42,640 --> 00:18:47,000
Speaker 3: buying and helping people not to suffer from a behavioral

273
00:18:47,119 --> 00:18:52,160
Speaker 3: bias such as, let's say, incomplete attention or unrealistic optimism

274
00:18:52,200 --> 00:18:55,600
Speaker 3: when they're buying things. So these are standard consumer protection

275
00:18:55,800 --> 00:18:59,120
Speaker 3: things which many of our agencies in the US, home

276
00:18:59,160 --> 00:19:03,840
Speaker 3: grown America. They've done that, and that's good and we

277
00:19:04,000 --> 00:19:06,639
Speaker 3: need more of that. So that's first line of defense.

278
00:19:07,080 --> 00:19:12,120
Speaker 3: Second line of defense isn't to say, you know, privacy, privacy, privacy,

279
00:19:12,240 --> 00:19:14,600
Speaker 3: though maybe that's a good song to sing. It's to

280
00:19:14,680 --> 00:19:19,000
Speaker 3: say right to algorithmic transparency. So this is something which

281
00:19:19,119 --> 00:19:23,520
Speaker 3: neither the US, nor Europe, nor Asia, nor South America

282
00:19:23,800 --> 00:19:27,359
Speaker 3: nor Africa has been very advanced on. So this is

283
00:19:27,440 --> 00:19:29,800
Speaker 3: a coming thing where we need to know what the

284
00:19:29,880 --> 00:19:34,960
Speaker 3: algorithms are doing. So it's public what's Amazon's algorithm doing?

285
00:19:35,400 --> 00:19:37,879
Speaker 3: That would be good to know, and it shouldn't be

286
00:19:38,000 --> 00:19:43,480
Speaker 3: the case that some efforts to ensure transparency invade Amazon's

287
00:19:43,600 --> 00:19:44,480
Speaker 3: legitimate rights.

288
00:19:45,760 --> 00:19:51,080
Speaker 2: Really, really fascinating, thanks Cas. Anybody who is participating in

289
00:19:51,240 --> 00:19:56,920
Speaker 2: the American economy and society, consumers, investors, even just regular

290
00:19:57,040 --> 00:20:01,440
Speaker 2: readers of news, needs to be away of how algorithms

291
00:20:01,480 --> 00:20:05,480
Speaker 2: are affecting what they see, the prices they pay, and

292
00:20:05,560 --> 00:20:09,000
Speaker 2: the sort of information they're getting. So with a little

293
00:20:09,040 --> 00:20:13,520
Speaker 2: bit of forethought and the book Algorithmic Harm, you can

294
00:20:13,680 --> 00:20:18,760
Speaker 2: protect yourself from the worst aspects of algorithms. And AI,

295
00:20:19,640 --> 00:20:23,720
Speaker 2: I'm Barry Redults you're listening to Bloomberg's at the Money