1
00:00:05,160 --> 00:00:09,720
Speaker 1: Why are we humans so easy to deceive? What are

2
00:00:09,800 --> 00:00:12,520
Speaker 1: the tricks of the trade, and how can we train

3
00:00:12,600 --> 00:00:16,640
Speaker 1: ourselves to be more aware of these? And what does

4
00:00:16,680 --> 00:00:20,240
Speaker 1: any of this have to do with fahnose or forging

5
00:00:20,320 --> 00:00:28,200
Speaker 1: letters or the shell game. Welcome to Inner Cosmos with

6
00:00:28,320 --> 00:00:31,400
Speaker 1: me David Eagleman. I'm a neuroscientist and an author at

7
00:00:31,440 --> 00:00:36,159
Speaker 1: Stanford and in these episodes, I examine the intersection between

8
00:00:36,240 --> 00:00:41,920
Speaker 1: our brains and our lives, and today's episode is about deception.

9
00:00:49,800 --> 00:00:53,559
Speaker 1: You presumably wouldn't do something to cheat a stranger out

10
00:00:53,600 --> 00:00:56,760
Speaker 1: of twenty dollars, So why are there people who would

11
00:00:56,960 --> 00:00:57,279
Speaker 1: do that?

12
00:00:57,720 --> 00:00:59,200
Speaker 2: And what can we do.

13
00:00:59,560 --> 00:01:04,120
Speaker 1: To be a little more thoughtful and aware and immune

14
00:01:04,440 --> 00:01:09,080
Speaker 1: against deception? So in a previous episode I talked about

15
00:01:09,120 --> 00:01:13,000
Speaker 1: a really impactful event when I was a neuroscience graduate

16
00:01:13,040 --> 00:01:16,480
Speaker 1: student getting my PhD. I was a second year student

17
00:01:16,520 --> 00:01:19,480
Speaker 1: in the department and this new young woman came in

18
00:01:19,520 --> 00:01:22,560
Speaker 1: as a first year student. We'll call her Tanya, and

19
00:01:22,640 --> 00:01:26,720
Speaker 1: everyone could see that Tanya was great. She had great grades,

20
00:01:27,200 --> 00:01:31,480
Speaker 1: top standardized test scores, terrific letters of recommendation, and in

21
00:01:31,520 --> 00:01:35,560
Speaker 1: the interviews she even won over my graduate advisor, who

22
00:01:35,640 --> 00:01:39,440
Speaker 1: was famously spiky towards people. And I tell her full

23
00:01:39,480 --> 00:01:42,920
Speaker 1: story in episode sixteen, but the short version is that

24
00:01:43,000 --> 00:01:47,480
Speaker 1: she faked everything on her graduate school application. She faked

25
00:01:47,720 --> 00:01:52,240
Speaker 1: the school transcript and the GRE scores and the letters

26
00:01:52,240 --> 00:01:56,800
Speaker 1: of recommendation, and she was only caught because an administrator

27
00:01:56,840 --> 00:01:59,520
Speaker 1: at the school was so impressed with her that she

28
00:01:59,640 --> 00:02:03,040
Speaker 1: decided to call the professors who had written the letters

29
00:02:03,080 --> 00:02:06,880
Speaker 1: of recommendation to ask how they'd produced a student like Tanya.

30
00:02:07,400 --> 00:02:11,360
Speaker 1: And that's how the whole house of cards came tumbling down. Now,

31
00:02:11,400 --> 00:02:14,359
Speaker 1: for those of you who listened to episode sixteen, you'll

32
00:02:14,400 --> 00:02:19,160
Speaker 1: remember that Tanya's story then got much weirder because she

33
00:02:19,280 --> 00:02:22,200
Speaker 1: went to Yale University and tried to pull exactly the

34
00:02:22,240 --> 00:02:25,079
Speaker 1: same trick, and when she was caught there, they put

35
00:02:25,080 --> 00:02:28,320
Speaker 1: her in jail. And then she and her mother got

36
00:02:28,360 --> 00:02:32,440
Speaker 1: caught doing a drug deal with two undercover agents. And

37
00:02:32,480 --> 00:02:36,720
Speaker 1: then Tanya decided to try murdering a girl who looked

38
00:02:36,800 --> 00:02:40,480
Speaker 1: vaguely like her to avoid going to prison. Now that

39
00:02:40,600 --> 00:02:43,800
Speaker 1: plot failed, but only barely. So that's the quick recap

40
00:02:43,840 --> 00:02:46,079
Speaker 1: of the story. But the part I want to concentrate

41
00:02:46,120 --> 00:02:50,400
Speaker 1: on today is why did none of us see this coming?

42
00:02:50,919 --> 00:02:53,680
Speaker 1: We all thought she was great. And this was a

43
00:02:53,840 --> 00:02:59,000
Speaker 1: neuroscience graduate program full of people who were aspiring learners

44
00:02:59,040 --> 00:03:03,120
Speaker 1: about the human brain and faculty who were presumably already

45
00:03:03,400 --> 00:03:06,320
Speaker 1: experts in the brain. And yet every single one of

46
00:03:06,400 --> 00:03:09,639
Speaker 1: us thought that Tanya was great. None of us even

47
00:03:09,680 --> 00:03:13,799
Speaker 1: had the briefest glimpse of doubt or suspicion when she

48
00:03:13,919 --> 00:03:19,240
Speaker 1: started school, And we were all maximally surprised when we

49
00:03:19,280 --> 00:03:22,200
Speaker 1: saw how completely we had been fooled.

50
00:03:23,120 --> 00:03:25,120
Speaker 2: So why were we so blind?

51
00:03:26,200 --> 00:03:29,280
Speaker 1: Well, first of all, none of us would have thought

52
00:03:29,360 --> 00:03:33,360
Speaker 1: about faking our transcripts and writing fake letters and so on.

53
00:03:33,440 --> 00:03:34,160
Speaker 3: That kind of.

54
00:03:34,720 --> 00:03:40,040
Speaker 1: Deception didn't exist in our mental models, and so it

55
00:03:40,160 --> 00:03:42,600
Speaker 1: was totally invisible to us when it was sitting there

56
00:03:42,680 --> 00:03:45,440
Speaker 1: right in front of us. And one of the themes

57
00:03:45,480 --> 00:03:48,040
Speaker 1: of this podcast and of my next book is that

58
00:03:48,480 --> 00:03:52,280
Speaker 1: we need to get better at seeing outside the garden

59
00:03:52,320 --> 00:03:56,640
Speaker 1: walls of our own internal models. This is really what

60
00:03:56,760 --> 00:04:01,840
Speaker 1: the passage into maturity is about, seeing the limitations of

61
00:04:01,880 --> 00:04:06,200
Speaker 1: our own thinking and realizing that what's going on in

62
00:04:06,240 --> 00:04:09,960
Speaker 1: someone else's head might be very different than what's going

63
00:04:10,000 --> 00:04:13,640
Speaker 1: on inside hours even if we're not the kind of

64
00:04:13,720 --> 00:04:18,240
Speaker 1: person to do something, even if it seems absolutely unimaginable

65
00:04:18,279 --> 00:04:20,200
Speaker 1: to us, it doesn't mean that.

66
00:04:20,120 --> 00:04:21,560
Speaker 2: It seems that way to someone else.

67
00:04:22,080 --> 00:04:24,560
Speaker 1: And if you heard episodes twenty and twenty one, you'll

68
00:04:24,560 --> 00:04:27,159
Speaker 1: know that we dove into some of the really awful

69
00:04:27,200 --> 00:04:31,480
Speaker 1: things that happened during wartime. And again, just because you

70
00:04:31,680 --> 00:04:35,880
Speaker 1: can't imagine hacking your neighbors to death with a machete

71
00:04:35,960 --> 00:04:39,360
Speaker 1: or shooting your neighbors or bayonetting them, it doesn't mean

72
00:04:39,880 --> 00:04:44,880
Speaker 1: that someone else can't imagine that and won't foment violence

73
00:04:44,960 --> 00:04:48,800
Speaker 1: without having much compunction about it. So an understanding of

74
00:04:49,040 --> 00:04:55,440
Speaker 1: history requires an expansion of our mental models, and that's

75
00:04:55,440 --> 00:04:59,520
Speaker 1: what's required for navigating day to day life as well,

76
00:04:59,600 --> 00:05:03,719
Speaker 1: because if not everyone is just like you on the inside.

77
00:05:03,800 --> 00:05:09,000
Speaker 1: For example, psychopaths make up about one percent of the population,

78
00:05:09,520 --> 00:05:11,240
Speaker 1: and by the way, they make up about twenty to

79
00:05:11,240 --> 00:05:15,880
Speaker 1: thirty percent of the prison population. They don't care about you,

80
00:05:15,920 --> 00:05:18,960
Speaker 1: they don't simulate what it is like to be you,

81
00:05:19,640 --> 00:05:22,320
Speaker 1: and they can be violent towards you because they just

82
00:05:22,400 --> 00:05:25,320
Speaker 1: see you as an obstacle to.

83
00:05:25,400 --> 00:05:27,080
Speaker 2: Flow around to get what they want.

84
00:05:27,520 --> 00:05:30,240
Speaker 1: And I'm going to do an episode on psychopathy soon,

85
00:05:30,279 --> 00:05:31,839
Speaker 1: but the point I want to make right now is

86
00:05:31,839 --> 00:05:35,520
Speaker 1: that if you are not a psychopath, this is very

87
00:05:35,520 --> 00:05:39,120
Speaker 1: difficult to imagine someone behaving that way. But you'll be

88
00:05:39,320 --> 00:05:43,040
Speaker 1: smarter in your daily life if you understand how other

89
00:05:43,160 --> 00:05:47,160
Speaker 1: people can be different from you. Now, sometimes people are

90
00:05:47,200 --> 00:05:50,279
Speaker 1: different in wonderful ways, like when you see some situation

91
00:05:50,960 --> 00:05:53,960
Speaker 1: in which someone is braver than you, or just more

92
00:05:54,080 --> 00:05:57,680
Speaker 1: charitable with a higher percentage of their money, or more

93
00:05:57,800 --> 00:06:00,599
Speaker 1: willing to do the right thing, like to climb the

94
00:06:00,680 --> 00:06:03,560
Speaker 1: side of a building to save the toddler hanging off

95
00:06:03,600 --> 00:06:07,080
Speaker 1: the balcony, even though you would be more scared. But

96
00:06:07,279 --> 00:06:11,160
Speaker 1: sometimes we see people different from us in the other direction,

97
00:06:11,279 --> 00:06:15,200
Speaker 1: people who cheat and lie and steal, and it's hard

98
00:06:15,240 --> 00:06:17,640
Speaker 1: to understand because we don't have a good model of that,

99
00:06:17,800 --> 00:06:21,920
Speaker 1: and so we're often caught completely by surprise. I'll give

100
00:06:21,960 --> 00:06:24,640
Speaker 1: you an example of this when I was young. When

101
00:06:24,680 --> 00:06:27,880
Speaker 1: I was sixteen years old, I was traveling with my

102
00:06:28,000 --> 00:06:31,800
Speaker 1: parents in Barcelona, and I was spending an afternoon walking

103
00:06:31,800 --> 00:06:34,880
Speaker 1: around by myself, and I saw a crowd of people

104
00:06:34,920 --> 00:06:38,400
Speaker 1: playing a shell game. You know, this is the game

105
00:06:38,440 --> 00:06:41,839
Speaker 1: where a person puts a small ball under one of

106
00:06:41,920 --> 00:06:45,120
Speaker 1: three cups and then rotates the cups around and around

107
00:06:45,640 --> 00:06:48,360
Speaker 1: and then you have to guess which cup the ball

108
00:06:48,440 --> 00:06:52,040
Speaker 1: is under. So I stopped to watch because there was

109
00:06:52,080 --> 00:06:55,760
Speaker 1: a small crowd and the dealer was moving the cups around,

110
00:06:55,839 --> 00:06:58,280
Speaker 1: and there was this pedestrian like me who had put

111
00:06:58,320 --> 00:07:02,320
Speaker 1: down some money. And pedestrian watched the cups go around

112
00:07:02,360 --> 00:07:04,919
Speaker 1: and round, and when they stopped, he pointed to a

113
00:07:04,960 --> 00:07:06,160
Speaker 1: cup and it was.

114
00:07:06,120 --> 00:07:06,960
Speaker 2: The wrong cup.

115
00:07:07,600 --> 00:07:10,200
Speaker 1: But I could see where the cups had moved, and

116
00:07:10,280 --> 00:07:12,240
Speaker 1: I knew it was the cup on the left, but

117
00:07:12,280 --> 00:07:15,400
Speaker 1: this pedestrian pointed to the middle. So the dealer uncovered

118
00:07:15,440 --> 00:07:18,400
Speaker 1: the middle cup, and the whole crowd made a whooping sound,

119
00:07:18,800 --> 00:07:21,520
Speaker 1: and so the pedestrian put down more money to play

120
00:07:21,560 --> 00:07:24,800
Speaker 1: another round, and the dealer shows the ball under the

121
00:07:24,880 --> 00:07:27,680
Speaker 1: left cup, and then he rotates the cups around faster

122
00:07:27,720 --> 00:07:30,800
Speaker 1: and faster, but I kept my eyes locked on the

123
00:07:30,800 --> 00:07:34,520
Speaker 1: correct cup, and again the pedestrian guessed wrong, but I

124
00:07:34,600 --> 00:07:37,160
Speaker 1: knew where the ball was. So this happens a few

125
00:07:37,200 --> 00:07:40,760
Speaker 1: more rounds, and the pedestrian gives up, and the dealer

126
00:07:40,840 --> 00:07:42,920
Speaker 1: looks at me and motions for me to put up

127
00:07:42,960 --> 00:07:45,960
Speaker 1: some money, so I did so he shows me the

128
00:07:45,960 --> 00:07:48,600
Speaker 1: ball and rotates the cups around and around, and I

129
00:07:48,760 --> 00:07:50,560
Speaker 1: keep my eye on it, and when he stops, I

130
00:07:50,680 --> 00:07:54,679
Speaker 1: point to the correct cup and the cup was empty.

131
00:07:55,200 --> 00:07:57,960
Speaker 2: I got it wrong, what was going on?

132
00:07:58,160 --> 00:08:00,240
Speaker 1: So he motions for me to put down more money,

133
00:08:00,280 --> 00:08:02,840
Speaker 1: and I want to win my lost money back, so

134
00:08:02,880 --> 00:08:05,520
Speaker 1: I put down more and he runs the rotations again,

135
00:08:05,800 --> 00:08:08,200
Speaker 1: and I point to the cup where the ball should be,

136
00:08:08,680 --> 00:08:12,440
Speaker 1: and again it's empty. And before I know it, someone

137
00:08:12,440 --> 00:08:14,640
Speaker 1: in the crowd makes a whoop sound, and suddenly the

138
00:08:14,680 --> 00:08:17,840
Speaker 1: dealer folds up the board and the entire crowd disappears,

139
00:08:18,440 --> 00:08:21,280
Speaker 1: and I'm standing there all by myself in the street,

140
00:08:21,360 --> 00:08:24,160
Speaker 1: and I felt like such a fool because I had

141
00:08:24,200 --> 00:08:27,280
Speaker 1: just been deceived. Now this is embarrassing for me to

142
00:08:27,320 --> 00:08:29,520
Speaker 1: tell the story, and even all these years later, there

143
00:08:29,600 --> 00:08:31,400
Speaker 1: is some pain in the remembrance.

144
00:08:31,720 --> 00:08:32,600
Speaker 3: But my hope in.

145
00:08:32,520 --> 00:08:36,480
Speaker 1: Relating the story is that at least one teenage listener

146
00:08:36,600 --> 00:08:39,840
Speaker 1: gets an expansion of their mental model from this and

147
00:08:40,120 --> 00:08:42,600
Speaker 1: doesn't have to play this game. Just in case you

148
00:08:42,600 --> 00:08:45,720
Speaker 1: don't know, there is no honest version of the shell game.

149
00:08:45,760 --> 00:08:49,040
Speaker 1: It's always performed by hucksters who use sleight of hand

150
00:08:49,080 --> 00:08:51,840
Speaker 1: to move the ball from one cup to another, and

151
00:08:51,880 --> 00:08:55,760
Speaker 1: the whole crowd is in on the deception. Now, when

152
00:08:55,800 --> 00:08:57,680
Speaker 1: I did some research on this, I found the shell

153
00:08:57,720 --> 00:09:01,480
Speaker 1: game is very old, so the game of people trying

154
00:09:01,520 --> 00:09:03,920
Speaker 1: to deceive other people is ancient.

155
00:09:04,920 --> 00:09:05,160
Speaker 3: Now.

156
00:09:05,240 --> 00:09:10,160
Speaker 1: Sometimes deception is planned in advance like this, and sometimes

157
00:09:10,200 --> 00:09:12,400
Speaker 1: people are just trying to get out of bad situation.

158
00:09:12,559 --> 00:09:15,800
Speaker 1: They make it worse. Sometimes it's hard to tell. I mean,

159
00:09:15,800 --> 00:09:18,800
Speaker 1: look at the company, Thearrhannos, which you've probably heard of.

160
00:09:18,840 --> 00:09:21,800
Speaker 1: They were a health tech company founded by a young

161
00:09:21,840 --> 00:09:25,280
Speaker 1: woman named Elizabeth Holmes, and they were out to develop

162
00:09:25,400 --> 00:09:29,720
Speaker 1: a biological test that could measure a whole bunch of

163
00:09:29,760 --> 00:09:33,240
Speaker 1: things using just a drop or two of blood. So

164
00:09:33,320 --> 00:09:37,240
Speaker 1: they raised seven hundred and twenty four million dollars from investors,

165
00:09:37,760 --> 00:09:41,880
Speaker 1: including the media mogul Robert Murdoch and former Secretary of

166
00:09:41,920 --> 00:09:45,560
Speaker 1: State Henry Kissinger, and all kinds of big players were

167
00:09:45,679 --> 00:09:51,080
Speaker 1: enthusiastic about this revolutionary technology. But this all came crashing

168
00:09:51,160 --> 00:09:55,480
Speaker 1: down in twenty fifteen when it surfaced that Sarahnos had

169
00:09:55,520 --> 00:09:59,680
Speaker 1: been lying about what their technology could do. Holmes was

170
00:09:59,760 --> 00:10:03,760
Speaker 1: charged with fraud and conspiracy, and she was found guilty

171
00:10:03,800 --> 00:10:06,960
Speaker 1: in twenty twenty two. Now you probably know this story,

172
00:10:06,960 --> 00:10:10,560
Speaker 1: but the interesting backside of this story is So many

173
00:10:10,600 --> 00:10:13,240
Speaker 1: people here in Silicon Valley have said things to me

174
00:10:13,840 --> 00:10:18,959
Speaker 1: suggesting that they would have never fallen for Farnos, like, oh,

175
00:10:19,000 --> 00:10:21,800
Speaker 1: I would have known right away that couldn't work. But

176
00:10:21,920 --> 00:10:25,480
Speaker 1: that's silly, and I generally don't believe them because it's

177
00:10:25,559 --> 00:10:28,960
Speaker 1: easy to get douped. And if the other people who

178
00:10:28,960 --> 00:10:31,200
Speaker 1: believe in it and invest in it and sit on

179
00:10:31,240 --> 00:10:35,079
Speaker 1: the board are billionaires and big shots, what's to make

180
00:10:35,120 --> 00:10:38,800
Speaker 1: you think that wouldn't have a great gravitational pull on you.

181
00:10:39,080 --> 00:10:41,400
Speaker 1: Because the truth is, when we look at things that

182
00:10:41,760 --> 00:10:45,320
Speaker 1: other people believe in, or even simply things that match

183
00:10:45,360 --> 00:10:51,120
Speaker 1: our expectations, we often don't do any further looking into it. Now,

184
00:10:51,120 --> 00:10:53,959
Speaker 1: this puts us in a tough spot because we have

185
00:10:54,080 --> 00:10:57,960
Speaker 1: to trust other people. We really have no choice, because

186
00:10:58,000 --> 00:11:01,839
Speaker 1: we can't disbelieve every thing or have time to check

187
00:11:01,880 --> 00:11:05,679
Speaker 1: on everything. So how do we work around this bias?

188
00:11:06,080 --> 00:11:08,880
Speaker 1: How can we take some of the tools of science,

189
00:11:09,320 --> 00:11:13,160
Speaker 1: which are all about clear thinking and import these into

190
00:11:13,200 --> 00:11:17,040
Speaker 1: our daily lives. Well, I'm no expert on this. I

191
00:11:17,080 --> 00:11:20,080
Speaker 1: often err on the side of believing everyone, but I

192
00:11:20,120 --> 00:11:24,360
Speaker 1: knew who to call. My colleagues Dan Simon's and Christopher

193
00:11:24,360 --> 00:11:27,959
Speaker 1: Shubrie recently wrote a terrific book all about deception called

194
00:11:28,360 --> 00:11:30,760
Speaker 1: Nobody's Fool. So I wrung them up.

195
00:11:31,559 --> 00:11:34,760
Speaker 4: The most interesting thing we found about what all the

196
00:11:34,800 --> 00:11:38,679
Speaker 4: cons and deceptions have in common is that the con artists,

197
00:11:38,760 --> 00:11:41,040
Speaker 4: the scammers, the swindlers, whatever you want to call them,

198
00:11:41,559 --> 00:11:44,800
Speaker 4: all seem to be taking advantage of sort of the

199
00:11:44,840 --> 00:11:50,920
Speaker 4: same set of our cognitive proclivities and our attentional biases.

200
00:11:50,960 --> 00:11:53,560
Speaker 4: What we like to pay attention to, what attracts us,

201
00:11:54,160 --> 00:11:58,520
Speaker 4: and what mistakes we tend to make, and decision making

202
00:11:58,559 --> 00:12:00,400
Speaker 4: that we may not be aware of.

203
00:12:01,080 --> 00:12:05,240
Speaker 1: That's Christopher Shabri, a professor of psychology and director of

204
00:12:05,320 --> 00:12:08,720
Speaker 1: Decision Sciences at Geisinger Research Institute.

205
00:12:08,760 --> 00:12:10,800
Speaker 4: They may not be consciously aware of, but somehow they

206
00:12:10,840 --> 00:12:13,840
Speaker 4: have sort of learned that adapted their schemes to things

207
00:12:13,920 --> 00:12:16,800
Speaker 4: that tend to exploit these loopholes in our thinking. Not

208
00:12:16,920 --> 00:12:21,959
Speaker 4: loopholes that are sort of design flaws necessarily, they're actually usually,

209
00:12:22,040 --> 00:12:24,440
Speaker 4: you know, good things about how we think. But when

210
00:12:24,440 --> 00:12:26,320
Speaker 4: someone is really trying to take advantage of us, they

211
00:12:26,320 --> 00:12:28,920
Speaker 4: can cleverly exploit those and gain the advantage.

212
00:12:29,360 --> 00:12:33,600
Speaker 1: So what are the important lessons for all of us

213
00:12:33,720 --> 00:12:36,760
Speaker 1: to think about? Given that I'd say there are a few,

214
00:12:37,840 --> 00:12:41,480
Speaker 1: and that's Dan Simons, a professor of psychology at the

215
00:12:41,600 --> 00:12:48,120
Speaker 1: University of Illinois. One is that we have a tendency

216
00:12:48,160 --> 00:12:52,400
Speaker 1: to assume that only the most gullible or naive or

217
00:12:52,800 --> 00:12:57,080
Speaker 1: uneducated people fall for scams. And that's partly because we

218
00:12:57,200 --> 00:13:00,959
Speaker 1: generally only see the results of cons and scams after

219
00:13:00,960 --> 00:13:03,000
Speaker 1: they're over right, So it's easy to see what the

220
00:13:03,000 --> 00:13:05,800
Speaker 1: red flags were when you knew it was a scam,

221
00:13:05,880 --> 00:13:08,240
Speaker 1: you found out it was a scam in the same

222
00:13:08,280 --> 00:13:12,160
Speaker 1: way that we can easily spot, you know, the obvious

223
00:13:12,200 --> 00:13:15,160
Speaker 1: ones in advance, things like the Nigerian prints email scam

224
00:13:15,160 --> 00:13:17,640
Speaker 1: that we know about. We can spot those red flags

225
00:13:17,679 --> 00:13:20,880
Speaker 1: because we've seen them before. But we tend to assume

226
00:13:20,880 --> 00:13:23,560
Speaker 1: that all scams prey on the people who are gullible.

227
00:13:23,840 --> 00:13:26,560
Speaker 1: And one of the key insights we've across all of

228
00:13:26,559 --> 00:13:29,320
Speaker 1: the sorts of scams that we've encountered is that scams

229
00:13:29,320 --> 00:13:32,800
Speaker 1: can affect anybody. Cons can affect anybody if they're targeted

230
00:13:33,080 --> 00:13:37,240
Speaker 1: in the right way to our wants and desires and needs. Yeah,

231
00:13:37,240 --> 00:13:39,880
Speaker 1: you know, I thought about this a lot with Sarahnos

232
00:13:40,160 --> 00:13:45,320
Speaker 1: here in Silicon Valley retrospectively, everyone acted like I would

233
00:13:45,320 --> 00:13:47,960
Speaker 1: have never fallen for that. But it's obvious that a

234
00:13:48,000 --> 00:13:51,560
Speaker 1: lot of good and smart people got sucked up into that,

235
00:13:52,120 --> 00:13:55,160
Speaker 1: And so how do you how do you interpret that? Well?

236
00:13:55,200 --> 00:13:58,119
Speaker 5: I think it's the way we see something like Sarahnose

237
00:13:58,600 --> 00:14:01,120
Speaker 5: is in hindsight, after the fact, in the same way

238
00:14:01,160 --> 00:14:03,040
Speaker 5: that we might watch a heist movie or a who

239
00:14:03,120 --> 00:14:05,839
Speaker 5: Done It movie, where we know there's a heist, right,

240
00:14:05,880 --> 00:14:08,520
Speaker 5: we know there's a con artist, we know that. In

241
00:14:09,760 --> 00:14:11,520
Speaker 5: the context of the movie, we know to look for

242
00:14:11,559 --> 00:14:14,120
Speaker 5: those red flags. We're trying to figure it out, and

243
00:14:14,280 --> 00:14:17,000
Speaker 5: the characters in the movie aren't. But when it's viewed

244
00:14:17,000 --> 00:14:20,360
Speaker 5: from the outside, it's kind of obvious, right, So Pharaohnose

245
00:14:20,480 --> 00:14:22,760
Speaker 5: after the fact, Yeah, there are lots of red flags

246
00:14:22,760 --> 00:14:24,880
Speaker 5: along the way, and they've been reported thoroughly, and it's

247
00:14:25,000 --> 00:14:28,040
Speaker 5: great narrative. But when you're immersed in it and you're

248
00:14:28,040 --> 00:14:30,360
Speaker 5: trying to figure out what's the next best investment or

249
00:14:30,400 --> 00:14:31,920
Speaker 5: what do I want to get in on really quick?

250
00:14:31,920 --> 00:14:34,520
Speaker 5: If you're a venture capitalist trying to kind of get

251
00:14:34,600 --> 00:14:37,320
Speaker 5: in on the next big thing, spotting all those red

252
00:14:37,320 --> 00:14:40,960
Speaker 5: flags is more difficult because you're incentivized to act with

253
00:14:41,040 --> 00:14:43,560
Speaker 5: efficiency and to try and catch things before they take

254
00:14:43,600 --> 00:14:46,480
Speaker 5: off and before people know about them. So those are

255
00:14:46,520 --> 00:14:49,440
Speaker 5: the contexts in which they're the marks, rather than watching

256
00:14:49,520 --> 00:14:51,520
Speaker 5: some interesting, engaging movie.

257
00:14:51,560 --> 00:14:54,320
Speaker 4: In the case of Pharaohose. Also, you know, there were

258
00:14:54,360 --> 00:14:57,200
Speaker 4: people who didn't invest and who didn't join the board

259
00:14:57,200 --> 00:14:59,040
Speaker 4: of directors and so on. They don't get as much

260
00:14:59,080 --> 00:15:02,800
Speaker 4: publicity as the unfortunate ones who did and look like

261
00:15:02,880 --> 00:15:05,800
Speaker 4: marks and suckers and so on in retrospect. But some

262
00:15:05,840 --> 00:15:09,760
Speaker 4: professional investors who specialized in biotech and healthcare investing, they

263
00:15:09,760 --> 00:15:13,200
Speaker 4: asked a lot more questions about the product, about the technology,

264
00:15:13,240 --> 00:15:15,640
Speaker 4: about clinical data, about all of that stuff, and then

265
00:15:16,080 --> 00:15:20,480
Speaker 4: they walked away. And I think one other important point

266
00:15:20,520 --> 00:15:22,480
Speaker 4: about their nose is I think, although I don't know,

267
00:15:22,480 --> 00:15:24,240
Speaker 4: because I'm not inside the heads of all those people,

268
00:15:24,280 --> 00:15:27,239
Speaker 4: but I think a lot of people didn't even consciously

269
00:15:27,400 --> 00:15:29,640
Speaker 4: consider the idea that there might have been a scam

270
00:15:29,720 --> 00:15:33,680
Speaker 4: or a fraud going on. Everything seemed good, everybody was optimistic,

271
00:15:33,760 --> 00:15:36,480
Speaker 4: there was a great vision. Little things that seem kind

272
00:15:36,480 --> 00:15:38,280
Speaker 4: of odd. Maybe you can explain away, this is just

273
00:15:38,320 --> 00:15:42,080
Speaker 4: a quirky company. The CEO is a little odd. You know, Well,

274
00:15:42,120 --> 00:15:43,840
Speaker 4: they've got all these famous people on the board. That

275
00:15:43,920 --> 00:15:47,240
Speaker 4: must be a good thing. Simply considering the possibility in

276
00:15:47,280 --> 00:15:49,760
Speaker 4: a big decision making situation that you maybe are being

277
00:15:49,800 --> 00:15:52,600
Speaker 4: scammed or there's something going on that you're not aware of,

278
00:15:53,080 --> 00:15:55,280
Speaker 4: you know, could be the first step towards like starting

279
00:15:55,280 --> 00:15:57,960
Speaker 4: to see those red flags or look for those red flags,

280
00:15:57,960 --> 00:16:00,280
Speaker 4: and maybe you can actually find some of them if

281
00:16:00,320 --> 00:16:02,480
Speaker 4: you were even thinking about the possibility that they might

282
00:16:02,520 --> 00:16:06,440
Speaker 4: be out there somewhere. We require a lot of trust

283
00:16:06,840 --> 00:16:09,440
Speaker 4: just to get by in life. And so how do

284
00:16:09,560 --> 00:16:13,120
Speaker 4: you guys think about striking a balance of trust and

285
00:16:13,160 --> 00:16:14,200
Speaker 4: a little bit of suspicion.

286
00:16:15,240 --> 00:16:17,760
Speaker 5: Well, trust is essential, right. In fact, we tend to

287
00:16:17,840 --> 00:16:21,560
Speaker 5: assume that when we hear something from somebody it's true

288
00:16:21,640 --> 00:16:23,960
Speaker 5: and until we take time to think about it otherwise.

289
00:16:24,000 --> 00:16:26,320
Speaker 5: And most of the time that's a great thing because

290
00:16:26,320 --> 00:16:29,120
Speaker 5: in most conversations, nobody's trying to lie to you. In

291
00:16:29,280 --> 00:16:31,840
Speaker 5: most interactions, nobody's trying to con you. I mean, the

292
00:16:31,880 --> 00:16:34,400
Speaker 5: odds of any of us being a victim of a

293
00:16:34,440 --> 00:16:37,200
Speaker 5: bernie made off or a their noose is pretty low.

294
00:16:37,680 --> 00:16:40,520
Speaker 5: The odds of any of us receiving fake information on

295
00:16:40,560 --> 00:16:43,240
Speaker 5: social media is pretty high. But we tend to be

296
00:16:43,280 --> 00:16:46,440
Speaker 5: trusting of the information we get, and it's a good

297
00:16:46,480 --> 00:16:47,000
Speaker 5: thing that we are.

298
00:16:47,120 --> 00:16:47,240
Speaker 4: Right.

299
00:16:47,240 --> 00:16:50,080
Speaker 5: If we were constantly skeptical of everything we encountered, we

300
00:16:50,080 --> 00:16:52,840
Speaker 5: could just never do anything. We could never have a conversation.

301
00:16:53,240 --> 00:16:56,440
Speaker 5: You can't check everything. You can't be a perpetual skeptic

302
00:16:56,520 --> 00:16:59,880
Speaker 5: or cynic about everything. You're not going to go check

303
00:17:00,720 --> 00:17:04,520
Speaker 5: in the grocery store if you buy an organic apple, right,

304
00:17:04,560 --> 00:17:05,840
Speaker 5: You're not going to go out to the farm and

305
00:17:05,840 --> 00:17:09,080
Speaker 5: make sure they didn't use any pesticides. Right, It's too much.

306
00:17:09,760 --> 00:17:12,120
Speaker 5: We can't really check that. We have to accept that

307
00:17:12,320 --> 00:17:13,960
Speaker 5: some of the time we're going to have to be trusting.

308
00:17:14,560 --> 00:17:16,240
Speaker 5: And the key is to kind of figure out when

309
00:17:16,280 --> 00:17:18,840
Speaker 5: are those times when we're at the greatest risk, When

310
00:17:18,880 --> 00:17:22,880
Speaker 5: are those times when the consequences could be bad enough

311
00:17:22,920 --> 00:17:24,520
Speaker 5: that we really would want to check and see if

312
00:17:24,520 --> 00:17:25,160
Speaker 5: we're being scammed.

313
00:17:25,840 --> 00:17:28,280
Speaker 2: So before we go on to some other topics, just

314
00:17:28,560 --> 00:17:31,719
Speaker 2: can you give a few examples of hoaxes or swindles

315
00:17:31,800 --> 00:17:35,720
Speaker 2: or scams so that our listeners can understand what we're

316
00:17:35,720 --> 00:17:36,240
Speaker 2: talking about.

317
00:17:37,480 --> 00:17:39,439
Speaker 4: I have a good one that I think a lot

318
00:17:39,440 --> 00:17:41,679
Speaker 4: of people probably haven't heard of, but they really should have,

319
00:17:42,200 --> 00:17:45,800
Speaker 4: which is called sometimes the president scam or the CEO scam,

320
00:17:46,000 --> 00:17:48,840
Speaker 4: and I didn't discover it. It's been it was going

321
00:17:48,880 --> 00:17:50,960
Speaker 4: on for a while and it was documented elsewhere. But

322
00:17:51,280 --> 00:17:53,560
Speaker 4: I think it's a great example of some of some

323
00:17:53,640 --> 00:17:59,000
Speaker 4: of the key ideas. So this French Israeli fraudster named

324
00:17:59,080 --> 00:18:03,920
Speaker 4: Jilbert Shickley developed a scam in which he would call

325
00:18:04,040 --> 00:18:08,040
Speaker 4: up sort of mid level employees of French companies, pretending

326
00:18:08,080 --> 00:18:11,760
Speaker 4: to be the CEO of the company, reaching down through

327
00:18:11,800 --> 00:18:13,880
Speaker 4: the ranks and calling up some middle manager and giving

328
00:18:13,920 --> 00:18:17,720
Speaker 4: them a task to do directly for him. And the

329
00:18:17,800 --> 00:18:21,359
Speaker 4: task always wound up in money being transferred directly to

330
00:18:21,440 --> 00:18:26,160
Speaker 4: some bank account or person or something like that, where

331
00:18:26,240 --> 00:18:28,280
Speaker 4: of course it wound up with Shickley and whoever his

332
00:18:28,359 --> 00:18:32,240
Speaker 4: associates were in their bank account somehow. And I think

333
00:18:32,240 --> 00:18:35,880
Speaker 4: it's kind of an audacious con because it's one guy

334
00:18:35,960 --> 00:18:38,360
Speaker 4: with a telephone calling people up who he's never met

335
00:18:38,400 --> 00:18:43,359
Speaker 4: before and talking them into essentially giving him a lot

336
00:18:43,400 --> 00:18:47,160
Speaker 4: of money. But it does illustrate sort of the idea

337
00:18:47,160 --> 00:18:49,280
Speaker 4: of truth bias that Dan was just talking about, that

338
00:18:49,720 --> 00:18:51,879
Speaker 4: if you don't believe that the person on the phone

339
00:18:51,920 --> 00:18:54,760
Speaker 4: is the CEO calling you, the whole thing goes nowhere.

340
00:18:55,200 --> 00:18:57,280
Speaker 4: But once you believe that, then the scam has a

341
00:18:57,359 --> 00:18:59,720
Speaker 4: chance to get through. And it also illustrates sort of

342
00:18:59,720 --> 00:19:02,640
Speaker 4: some of selection, sort of someone selection bias we see

343
00:19:02,640 --> 00:19:04,720
Speaker 4: in cons Like we hear about the ones that worked,

344
00:19:04,920 --> 00:19:06,399
Speaker 4: but we don't know about all the people who just

345
00:19:06,400 --> 00:19:08,560
Speaker 4: like hung up the phone or deleted the email when

346
00:19:08,600 --> 00:19:10,480
Speaker 4: he tried to talk to you know, to start talking

347
00:19:10,520 --> 00:19:12,680
Speaker 4: to them into it. Just like the millions and millions

348
00:19:12,720 --> 00:19:15,320
Speaker 4: of people who delete the Nigerian prints emails never get

349
00:19:15,320 --> 00:19:17,280
Speaker 4: it mentioned anywhere. You know, it's just a few people

350
00:19:17,320 --> 00:19:20,680
Speaker 4: who actually wind up going through with it. That CEO

351
00:19:20,800 --> 00:19:24,120
Speaker 4: scam or President scam went along for quite a long

352
00:19:24,160 --> 00:19:26,560
Speaker 4: time and sort of morphed and changed into different versions

353
00:19:26,600 --> 00:19:29,240
Speaker 4: where eventually people were pretending to be the Defense Minister

354
00:19:29,320 --> 00:19:34,520
Speaker 4: of France calling, you know, contacting wealthy individuals, especially with

355
00:19:34,600 --> 00:19:37,560
Speaker 4: French ties, and saying that the government of France needed

356
00:19:37,560 --> 00:19:41,919
Speaker 4: their help getting hostages, secret hostages out of Syria and

357
00:19:41,960 --> 00:19:44,720
Speaker 4: Pharmisis and so on, and wound up taking i think

358
00:19:44,760 --> 00:19:47,359
Speaker 4: something like eighty million dollars or eighty million euros in

359
00:19:47,400 --> 00:19:50,520
Speaker 4: total from a number of you know, French companies and

360
00:19:50,560 --> 00:19:53,240
Speaker 4: wealthy wealthy individuals by sort of similar tactics.

361
00:19:53,760 --> 00:19:55,719
Speaker 5: There's a new modern version of this which is much

362
00:19:55,800 --> 00:19:58,080
Speaker 5: dumber and much simpler. It doesn't require any sort of

363
00:19:58,080 --> 00:20:02,439
Speaker 5: sophisticated persuasion. People just send an email purportedly from the

364
00:20:02,440 --> 00:20:04,520
Speaker 5: boss of a company and saying, Hey, I'm in a

365
00:20:04,560 --> 00:20:06,480
Speaker 5: meeting right now, but I need to transfer these funds

366
00:20:06,520 --> 00:20:08,720
Speaker 5: right away or I need to close this sale. Can

367
00:20:08,760 --> 00:20:10,960
Speaker 5: you just go ahead and make this transfer for me?

368
00:20:11,680 --> 00:20:14,919
Speaker 5: And if the email happens to reach somebody who is

369
00:20:15,000 --> 00:20:18,159
Speaker 5: responsible for doing that, then they might go ahead and

370
00:20:18,200 --> 00:20:20,560
Speaker 5: do it without even double checking, when in reality the

371
00:20:20,560 --> 00:20:23,320
Speaker 5: money would just get paid to some other count of

372
00:20:23,320 --> 00:20:27,119
Speaker 5: the scammers. And this is so pervasive that I have

373
00:20:27,160 --> 00:20:29,560
Speaker 5: a cousin who teaches tennis. She runs a tennis club

374
00:20:29,600 --> 00:20:34,440
Speaker 5: and regularly gets her underlings. The other people teaching there

375
00:20:34,600 --> 00:20:37,879
Speaker 5: regularly get emails from her, not really from her, but

376
00:20:37,920 --> 00:20:40,119
Speaker 5: from her saying hey, I'm in a meeting, can you

377
00:20:40,200 --> 00:20:44,280
Speaker 5: run this? Can you make this payment? And her employees

378
00:20:44,520 --> 00:20:47,239
Speaker 5: and coworkers know that that's not true. She's pretty much

379
00:20:47,240 --> 00:20:49,720
Speaker 5: never in meetings, and they're not the sorts of people

380
00:20:49,720 --> 00:20:52,520
Speaker 5: who make purchases. But again, if you send it out

381
00:20:52,520 --> 00:20:55,119
Speaker 5: to enough people, you're going to happen to hit. Some

382
00:20:55,200 --> 00:20:57,359
Speaker 5: who in that moment are busy doing things are used

383
00:20:57,359 --> 00:20:59,119
Speaker 5: to getting emails from their boss asking them to do

384
00:20:59,160 --> 00:21:01,800
Speaker 5: something really quickly, and we'll go along with it without

385
00:21:01,880 --> 00:21:05,240
Speaker 5: questioning it. And this is a major source of business.

386
00:21:05,320 --> 00:21:05,560
Speaker 3: Run.

387
00:21:06,200 --> 00:21:09,000
Speaker 1: Do you guys have any other examples, just something that

388
00:21:09,240 --> 00:21:12,480
Speaker 1: some hoax or swinger or something that you came across

389
00:21:12,920 --> 00:21:15,160
Speaker 1: that you think is really illustrative.

390
00:21:15,600 --> 00:21:19,000
Speaker 4: I can give another example from the world of chess.

391
00:21:19,160 --> 00:21:21,240
Speaker 3: So I'm a chess player. Well, Dan is also a

392
00:21:21,320 --> 00:21:21,840
Speaker 3: chess player.

393
00:21:22,119 --> 00:21:24,560
Speaker 4: I'm probably a more serious player than Dan is, and

394
00:21:25,320 --> 00:21:29,159
Speaker 4: I'm a funnier player. Dan's a funny player. I'm a

395
00:21:29,160 --> 00:21:33,399
Speaker 4: serious player. I'll try to be funnier though. When you

396
00:21:33,440 --> 00:21:36,560
Speaker 4: play chess online, you don't see the person you're playing.

397
00:21:36,880 --> 00:21:39,879
Speaker 4: It's just a screen name and all you see is

398
00:21:39,920 --> 00:21:42,479
Speaker 4: the moves they play on an animated chess board, kind

399
00:21:42,520 --> 00:21:43,920
Speaker 4: of like you're playing a video game, right, You just

400
00:21:43,960 --> 00:21:47,080
Speaker 4: see the moose being made. And So I was playing

401
00:21:47,080 --> 00:21:49,120
Speaker 4: a game once. This has happened to me more than once.

402
00:21:49,119 --> 00:21:50,760
Speaker 4: But the occasion that I remember is I was playing

403
00:21:50,760 --> 00:21:51,959
Speaker 4: a game and it was a guy I had never

404
00:21:52,000 --> 00:21:55,240
Speaker 4: played before and the game started, and he had a

405
00:21:55,240 --> 00:21:57,880
Speaker 4: similar rating to mine, meaning, you know, we were both

406
00:21:57,880 --> 00:22:00,119
Speaker 4: pretty good players. I should be ready for a you know,

407
00:22:00,200 --> 00:22:03,000
Speaker 4: I should be ready for a good game. And every

408
00:22:03,040 --> 00:22:05,879
Speaker 4: move I made, it seemed like he always found like

409
00:22:05,920 --> 00:22:08,880
Speaker 4: a great response, and he never made a mistake. And

410
00:22:09,240 --> 00:22:11,080
Speaker 4: times when I thought I was winning, I really wasn't

411
00:22:11,080 --> 00:22:14,320
Speaker 4: winning because he found the escape. And moreover, he was

412
00:22:14,359 --> 00:22:16,919
Speaker 4: moving quite quickly, like he would make every move in

413
00:22:16,960 --> 00:22:18,280
Speaker 4: like five to ten seconds. And I was like, wow,

414
00:22:18,320 --> 00:22:19,800
Speaker 4: this guy's putting a lot of pressure on me. You know,

415
00:22:20,160 --> 00:22:22,040
Speaker 4: I'm thinking for a minute sometimes on these moves, and

416
00:22:22,080 --> 00:22:24,679
Speaker 4: he comes back in ten seconds all the time. And

417
00:22:24,720 --> 00:22:26,680
Speaker 4: in the end I got checkmated and I lost the game,

418
00:22:27,080 --> 00:22:29,800
Speaker 4: and I thought, wow, that guy like that guy played

419
00:22:29,800 --> 00:22:31,480
Speaker 4: a really good game against me. But then when I

420
00:22:31,520 --> 00:22:34,359
Speaker 4: looked at the game afterwards, chess dot com where we

421
00:22:34,359 --> 00:22:37,160
Speaker 4: played this game shows you exactly how much time both

422
00:22:37,160 --> 00:22:40,040
Speaker 4: players use in every move after the game, and I

423
00:22:40,080 --> 00:22:42,080
Speaker 4: noticed that he was making all of his moves within

424
00:22:42,119 --> 00:22:45,760
Speaker 4: that very tight band of five to ten seconds per move, basically,

425
00:22:46,040 --> 00:22:48,879
Speaker 4: never less than five seconds, never more than ten seconds.

426
00:22:48,960 --> 00:22:50,560
Speaker 4: Maybe it was twelve seconds, I don't remember, but if

427
00:22:50,560 --> 00:22:52,560
Speaker 4: you looked at my own graph. There were a couple

428
00:22:52,560 --> 00:22:54,119
Speaker 4: of moves that I took like one or two minutes

429
00:22:54,160 --> 00:22:56,719
Speaker 4: on and some moves I played almost instantly, like one second.

430
00:22:57,640 --> 00:23:01,199
Speaker 4: The consistency of his timing, and also the consistency of

431
00:23:01,240 --> 00:23:03,320
Speaker 4: the fact that he never made a mistake. All of

432
00:23:03,359 --> 00:23:06,000
Speaker 4: his moves were you know, almost the best move, if

433
00:23:06,040 --> 00:23:10,480
Speaker 4: not the best move according to computer analysis, really reveals

434
00:23:10,520 --> 00:23:12,439
Speaker 4: that all he was doing was being a conduit for

435
00:23:12,480 --> 00:23:14,960
Speaker 4: a computer, Like he was just typing my moves into

436
00:23:14,960 --> 00:23:17,240
Speaker 4: a computer and typing and putting back into chess dot

437
00:23:17,280 --> 00:23:19,480
Speaker 4: com the moves that the computer told him to play.

438
00:23:19,840 --> 00:23:23,240
Speaker 4: And here's an example where the behavior of a human

439
00:23:23,760 --> 00:23:26,280
Speaker 4: you know, really ought to be more noisy in some

440
00:23:26,400 --> 00:23:29,360
Speaker 4: fundamental way than the behavior of a computer. Humans never

441
00:23:29,400 --> 00:23:33,880
Speaker 4: play moves with robotic cadences. They never play the correct

442
00:23:33,880 --> 00:23:37,200
Speaker 4: move forty or fifty moves in a row, and there's

443
00:23:37,240 --> 00:23:40,280
Speaker 4: just much more variability in human decision making and almost

444
00:23:40,280 --> 00:23:44,320
Speaker 4: any human activity than there is in computer based activity.

445
00:23:44,320 --> 00:23:46,320
Speaker 4: So I filed a report. You know, you can report

446
00:23:46,320 --> 00:23:48,200
Speaker 4: any player, and sure enough, like a day or two later,

447
00:23:48,280 --> 00:23:49,919
Speaker 4: chess dot com came back and said, we're giving you

448
00:23:49,960 --> 00:23:52,600
Speaker 4: back the rating. Points you lost. This guy is violated,

449
00:23:52,640 --> 00:23:54,439
Speaker 4: you know, violated our fair play policy. And that kind

450
00:23:54,440 --> 00:23:56,320
Speaker 4: of thing happens all the time in online chess because

451
00:23:56,320 --> 00:23:59,280
Speaker 4: computers are so good that they can be used easily

452
00:23:59,320 --> 00:24:02,000
Speaker 4: to cheat. The real problem is are people using it

453
00:24:02,040 --> 00:24:03,840
Speaker 4: sort of like in over the board and like serious

454
00:24:03,840 --> 00:24:06,480
Speaker 4: tournaments and matches. And that's a whole other controversy. But

455
00:24:06,560 --> 00:24:09,360
Speaker 4: certainly was a case where you know, I was essentially

456
00:24:09,359 --> 00:24:11,359
Speaker 4: the victim of a little minor scam. I happened to

457
00:24:11,359 --> 00:24:14,679
Speaker 4: figure it out, but as a scam based on people

458
00:24:14,800 --> 00:24:19,000
Speaker 4: not noticing that, you know, not noticing the absence of variability,

459
00:24:19,040 --> 00:24:20,639
Speaker 4: which is a critical thing and a lot of and

460
00:24:20,720 --> 00:24:22,280
Speaker 4: a lot of cons I was.

461
00:24:22,280 --> 00:24:25,119
Speaker 5: Gonna say one thing about that that particular case. It's

462
00:24:25,160 --> 00:24:29,440
Speaker 5: a fairly minor scam, right, It's just one element of

463
00:24:29,960 --> 00:24:33,200
Speaker 5: the kinds of habits and hooks that we find really compelling,

464
00:24:33,880 --> 00:24:37,159
Speaker 5: the hook of consistency, right or that. Yeah, it's just

465
00:24:37,240 --> 00:24:40,520
Speaker 5: something that we don't look for the noise when we should.

466
00:24:40,520 --> 00:24:43,080
Speaker 5: But if you look at bigger scams, things like Bernie

467
00:24:43,080 --> 00:24:46,159
Speaker 5: made Us, Ponzi Ski or fairness, they rely on a

468
00:24:46,200 --> 00:24:48,760
Speaker 5: whole bunch of our cognitive tendencies and they appeal to

469
00:24:49,359 --> 00:24:51,359
Speaker 5: a lot of kinds of information that we find really

470
00:24:51,440 --> 00:24:54,120
Speaker 5: valuable and that do help us most of the time,

471
00:24:54,680 --> 00:24:55,679
Speaker 5: but they take advantage of.

472
00:24:55,680 --> 00:24:58,760
Speaker 3: Those to doup us.

473
00:24:59,119 --> 00:25:01,840
Speaker 1: So can you unpacked that a little bit about noise

474
00:25:02,119 --> 00:25:04,840
Speaker 1: in data and what we should be looking for.

475
00:25:05,520 --> 00:25:08,000
Speaker 5: Well, I mean, really, in any human behavior, anything that's

476
00:25:08,280 --> 00:25:12,879
Speaker 5: governed by interaction of people, we don't expect people to

477
00:25:12,920 --> 00:25:15,120
Speaker 5: perform the same way every single time. We don't expect

478
00:25:15,440 --> 00:25:18,040
Speaker 5: a three hundred hitter or three thirty three baseball hitter

479
00:25:18,119 --> 00:25:20,639
Speaker 5: to hit one out of every single every one out

480
00:25:20,640 --> 00:25:23,560
Speaker 5: of every three at bets, right, they will on average

481
00:25:24,480 --> 00:25:26,920
Speaker 5: average about one out of every three. But in any

482
00:25:26,960 --> 00:25:29,200
Speaker 5: game that doesn't guarantee they're going to get at least

483
00:25:29,240 --> 00:25:32,439
Speaker 5: one hit, right, we tend to confuse the sort of

484
00:25:32,480 --> 00:25:35,840
Speaker 5: on average performance with what happens every single time. So

485
00:25:35,960 --> 00:25:38,800
Speaker 5: take the case of Bernie Madoff's Ponzi scheme. Right, this

486
00:25:38,920 --> 00:25:41,240
Speaker 5: wasn't a sort of classic Ponzi scheme where he promised

487
00:25:41,240 --> 00:25:43,720
Speaker 5: fifty percent returns in six months like a you know,

488
00:25:43,800 --> 00:25:48,359
Speaker 5: current crypto scam. Would his returned eight to fourteen percent

489
00:25:48,400 --> 00:25:51,439
Speaker 5: or eight to twelve percent almost every year for the

490
00:25:51,560 --> 00:25:53,560
Speaker 5: entire life of the Ponzi scheme, with never a down

491
00:25:53,640 --> 00:25:55,520
Speaker 5: year and almost never a down month. It was like

492
00:25:55,560 --> 00:25:59,639
Speaker 5: a smooth, steady growth, And that's not what you expect

493
00:25:59,640 --> 00:26:01,760
Speaker 5: for some thing is complex. As a financial system, you

494
00:26:01,840 --> 00:26:05,040
Speaker 5: expect ups and downs. Sometimes you'll be up twenty five

495
00:26:05,040 --> 00:26:07,000
Speaker 5: percent in a year, sometimes you'll be down ten percent

496
00:26:07,040 --> 00:26:09,159
Speaker 5: in a year, and the average might be eight to

497
00:26:09,200 --> 00:26:11,920
Speaker 5: twelve percent, and that'd be pretty good, But you don't

498
00:26:11,960 --> 00:26:14,399
Speaker 5: expect the average to be true of every single case.

499
00:26:15,160 --> 00:26:18,439
Speaker 5: And this plays out in many many contexts where usually

500
00:26:18,480 --> 00:26:21,119
Speaker 5: consistency is a sign that we have great understanding of

501
00:26:21,119 --> 00:26:22,640
Speaker 5: how things work. We can do it the same way

502
00:26:22,680 --> 00:26:25,760
Speaker 5: every single time. We want things to be reliable, but

503
00:26:26,840 --> 00:26:30,000
Speaker 5: the tendency to have things be too consistent can be

504
00:26:30,040 --> 00:26:33,320
Speaker 5: so appealing to us that we don't realize when noise

505
00:26:33,320 --> 00:26:36,000
Speaker 5: should be present. This is common in a lot of

506
00:26:36,040 --> 00:26:38,800
Speaker 5: science frunt as well, that you find results that are

507
00:26:38,800 --> 00:26:41,919
Speaker 5: just too consistent to be believable, but the people who

508
00:26:41,960 --> 00:26:43,879
Speaker 5: are making it up don't realize that they need to

509
00:26:43,880 --> 00:26:44,760
Speaker 5: make up noise too.

510
00:27:00,359 --> 00:27:02,919
Speaker 1: So a lot of the reasons that we fall for

511
00:27:03,040 --> 00:27:06,840
Speaker 1: hoaxes their scams is because of cognitive shortcuts that we're

512
00:27:06,880 --> 00:27:09,919
Speaker 1: taking so tell us about that and what we can

513
00:27:09,960 --> 00:27:11,000
Speaker 1: do about those shortcuts.

514
00:27:12,000 --> 00:27:14,760
Speaker 4: Well, one of the most important shortcuts, I think is

515
00:27:15,160 --> 00:27:18,000
Speaker 4: it's not even so much a shortcut, it's just an

516
00:27:18,119 --> 00:27:22,680
Speaker 4: un standard operating procedure. We are very good at paying

517
00:27:22,680 --> 00:27:23,440
Speaker 4: attention to things.

518
00:27:23,520 --> 00:27:23,679
Speaker 3: You know.

519
00:27:23,720 --> 00:27:26,320
Speaker 4: Attention is a wonderful thing. We can do things with

520
00:27:26,359 --> 00:27:28,560
Speaker 4: attention that we can't do without attention. Like we couldn't

521
00:27:28,560 --> 00:27:31,040
Speaker 4: even follow a soccer game or a football game or

522
00:27:31,040 --> 00:27:33,119
Speaker 4: a basketball game without attention, and otherwise would just be

523
00:27:33,119 --> 00:27:35,120
Speaker 4: a big blur of bodies moving around and the little

524
00:27:35,200 --> 00:27:37,640
Speaker 4: round thing like you know, flying back and forth. Occasionally

525
00:27:37,920 --> 00:27:40,240
Speaker 4: we'd have no hope of understanding it. But with attention

526
00:27:40,320 --> 00:27:42,440
Speaker 4: we can focus on selected aspects of it and sort

527
00:27:42,440 --> 00:27:44,399
Speaker 4: of put together the plot and the sequence of events

528
00:27:44,440 --> 00:27:46,480
Speaker 4: and what people are trying to do, and understand the

529
00:27:46,480 --> 00:27:48,560
Speaker 4: intentions behind it, all the way up to the strategies

530
00:27:48,600 --> 00:27:48,959
Speaker 4: and so on.

531
00:27:49,040 --> 00:27:49,760
Speaker 3: It's great.

532
00:27:50,080 --> 00:27:52,920
Speaker 4: However, the downside of attention is that when we're paying

533
00:27:52,920 --> 00:27:56,679
Speaker 4: attention to something, we may not notice other things that

534
00:27:56,720 --> 00:28:01,919
Speaker 4: we're not paying attention to. And a Hoover fraudster knows that,

535
00:28:02,400 --> 00:28:04,119
Speaker 4: and they know that if they can get our attention

536
00:28:04,320 --> 00:28:07,240
Speaker 4: on one thing, kind of like a magician, then we

537
00:28:07,359 --> 00:28:09,639
Speaker 4: might not notice other important things that are happening. And

538
00:28:09,640 --> 00:28:12,520
Speaker 4: of course they're not doing magic, they're actually trying to

539
00:28:12,560 --> 00:28:13,800
Speaker 4: deceive us for profit.

540
00:28:14,240 --> 00:28:15,800
Speaker 3: So many of.

541
00:28:15,760 --> 00:28:21,480
Speaker 4: The basic sort of deceptions in areas like marketing, where

542
00:28:21,480 --> 00:28:23,800
Speaker 4: it's sort of not even deception in some cases, it's

543
00:28:23,800 --> 00:28:25,600
Speaker 4: just kind of like the way business is done. You

544
00:28:25,720 --> 00:28:30,040
Speaker 4: get the recipient to focus on what you're showing them,

545
00:28:30,400 --> 00:28:32,640
Speaker 4: and you can count on them usually to not ask

546
00:28:32,720 --> 00:28:35,040
Speaker 4: questions about what you're not showing them. So, for example,

547
00:28:35,080 --> 00:28:39,280
Speaker 4: like a product demo video, like this startup company called Nicola,

548
00:28:39,320 --> 00:28:43,920
Speaker 4: which is still around trying to build electric vehicles trucks.

549
00:28:43,920 --> 00:28:46,840
Speaker 4: In this case, they created a demo video of one

550
00:28:46,840 --> 00:28:50,360
Speaker 4: of their trucks tooling down a highway, looked like going

551
00:28:50,360 --> 00:28:52,680
Speaker 4: at a nice rate of speed, and addingpressive music behind

552
00:28:52,720 --> 00:28:56,200
Speaker 4: it and so on, counting on people not to realize,

553
00:28:56,240 --> 00:28:59,080
Speaker 4: not to think, well, wait a minute, what happened before

554
00:28:59,120 --> 00:29:01,840
Speaker 4: the demo started. What was the angle that the camera

555
00:29:02,000 --> 00:29:04,240
Speaker 4: was at. Actually the camera was tilted a little bit,

556
00:29:04,280 --> 00:29:07,120
Speaker 4: so actually what was appeared to be rolling down on

557
00:29:07,160 --> 00:29:09,600
Speaker 4: a flat surface was rolling down a hill. So the

558
00:29:09,640 --> 00:29:12,080
Speaker 4: thing actually had no functioning motor you know, and so on.

559
00:29:12,160 --> 00:29:14,520
Speaker 4: It just rolled down a hill slowly, and then the

560
00:29:14,560 --> 00:29:17,200
Speaker 4: positioning of the camera and the video, you know, cutting

561
00:29:17,200 --> 00:29:18,880
Speaker 4: did the rest of the work. And those are things

562
00:29:18,920 --> 00:29:20,880
Speaker 4: we just don't think about. Right, We're focusing on the truck.

563
00:29:20,920 --> 00:29:23,400
Speaker 4: It looks nice, it's moving, nice background, and so on,

564
00:29:23,960 --> 00:29:26,320
Speaker 4: and we don't ask what's missing, Like, what information are

565
00:29:26,320 --> 00:29:28,160
Speaker 4: we missing about what's here? What information are they not

566
00:29:28,200 --> 00:29:29,920
Speaker 4: providing to us? Are they telling us about all the

567
00:29:29,920 --> 00:29:31,640
Speaker 4: times they tried to make the vehicle work but it

568
00:29:31,720 --> 00:29:33,520
Speaker 4: just didn't work at all, and just showing us the

569
00:29:33,520 --> 00:29:38,120
Speaker 4: one time that it did. So attention focus is really useful,

570
00:29:38,480 --> 00:29:41,600
Speaker 4: but it creates, you know, it sort of creates a loophole.

571
00:29:41,920 --> 00:29:45,040
Speaker 4: It creates a way for other people to, you know,

572
00:29:45,120 --> 00:29:46,800
Speaker 4: to exploit that well.

573
00:29:46,840 --> 00:29:50,360
Speaker 1: One cognitive shortcut that you mentioned in the book is prediction.

574
00:29:50,600 --> 00:29:53,560
Speaker 1: So how is it that we become victims of our

575
00:29:53,560 --> 00:29:54,800
Speaker 1: own life experience?

576
00:29:56,440 --> 00:29:58,560
Speaker 5: Yeah, And I think that's a great way of phrasing it,

577
00:29:58,600 --> 00:30:01,120
Speaker 5: that it's our life experience. It makes sense for us

578
00:30:01,120 --> 00:30:03,560
Speaker 5: to have expectations based on our past experience and to

579
00:30:03,640 --> 00:30:06,400
Speaker 5: use those predictions. And the vast majority of the time

580
00:30:07,200 --> 00:30:09,040
Speaker 5: we can use our past behavior to predict what's going

581
00:30:09,080 --> 00:30:10,920
Speaker 5: to happen in the future, right, that that's a really

582
00:30:10,920 --> 00:30:13,880
Speaker 5: important thing to be able to do. The challenge comes

583
00:30:13,920 --> 00:30:18,560
Speaker 5: in that we don't tend to question enough the information

584
00:30:18,600 --> 00:30:22,040
Speaker 5: we get when it's perfectly consistent with what we predicted. So,

585
00:30:22,760 --> 00:30:24,480
Speaker 5: and this is something that I think is really interesting

586
00:30:24,480 --> 00:30:27,520
Speaker 5: in the context of scientific errors. So let's say you

587
00:30:27,640 --> 00:30:30,840
Speaker 5: run an experiment and you've got an experimental group and

588
00:30:30,840 --> 00:30:32,640
Speaker 5: a placebo group, and you want to see which one

589
00:30:32,680 --> 00:30:35,760
Speaker 5: does better. Right, and you're predicting your new experimental intervention

590
00:30:35,840 --> 00:30:38,040
Speaker 5: is going to do great. And let's say that you

591
00:30:38,200 --> 00:30:41,600
Speaker 5: find that the placebo condition actually does better than the

592
00:30:41,600 --> 00:30:45,240
Speaker 5: experimental condition. Well, you're going to really dig into those results.

593
00:30:45,280 --> 00:30:46,680
Speaker 5: You're going to dig into the data. You're going to

594
00:30:46,680 --> 00:30:48,200
Speaker 5: look at your code. You're going to make sure that

595
00:30:48,240 --> 00:30:51,240
Speaker 5: everything was coded correctly, that there weren't any data points

596
00:30:51,240 --> 00:30:54,280
Speaker 5: that didn't make sense. You're to make sure you didn't

597
00:30:54,280 --> 00:30:56,120
Speaker 5: swap the names of the conditions so that you got

598
00:30:56,160 --> 00:30:58,400
Speaker 5: it wrong. You're going to look into it pretty carefully

599
00:30:58,440 --> 00:31:00,400
Speaker 5: because it didn't match what you were predicting.

600
00:31:01,280 --> 00:31:02,080
Speaker 3: Had it come out.

601
00:31:01,960 --> 00:31:04,160
Speaker 5: Exactly the way you predicted, you might not dig as

602
00:31:04,200 --> 00:31:07,280
Speaker 5: closely and that's been something that's led to a lot

603
00:31:07,280 --> 00:31:10,320
Speaker 5: of errors. Right, So you have a spreadsheet that produces

604
00:31:10,320 --> 00:31:12,000
Speaker 5: the right results, and you don't double check to make

605
00:31:12,040 --> 00:31:15,440
Speaker 5: sure you didn't fill down the column incorrectly because it

606
00:31:15,480 --> 00:31:18,280
Speaker 5: matched what you were predicting. So that sort of error

607
00:31:19,000 --> 00:31:21,800
Speaker 5: is I think a really common one. We're really good

608
00:31:21,920 --> 00:31:24,600
Speaker 5: at applying our critical faculties when we see something we

609
00:31:24,640 --> 00:31:27,040
Speaker 5: don't like that we didn't expect that we didn't predict.

610
00:31:27,480 --> 00:31:30,200
Speaker 5: Somebody shares something on social media that was counter to

611
00:31:30,240 --> 00:31:32,920
Speaker 5: your views, you can rip into that, and we're all

612
00:31:32,960 --> 00:31:35,640
Speaker 5: pretty good at doing that, But when it perfectly matches,

613
00:31:35,840 --> 00:31:38,080
Speaker 5: we're much more likely to just quickly pass it along

614
00:31:38,120 --> 00:31:41,840
Speaker 5: and retweet it and not necessarily think through carefully is

615
00:31:41,840 --> 00:31:43,280
Speaker 5: it really true.

616
00:31:43,760 --> 00:31:45,800
Speaker 1: Indeed, when we see things that are familiar in matching

617
00:31:45,800 --> 00:31:48,120
Speaker 1: our expectation, we don't look further into it. So how

618
00:31:48,160 --> 00:31:51,240
Speaker 1: do we work around that bias?

619
00:31:51,960 --> 00:31:54,520
Speaker 4: Well, I would say obviously the first thing is to

620
00:31:54,520 --> 00:31:57,080
Speaker 4: be where that we're doing this. That's the first step.

621
00:31:57,440 --> 00:32:03,120
Speaker 4: Second step is to and again like not every moment

622
00:32:03,120 --> 00:32:05,680
Speaker 4: of every day, but when you're making a big decision,

623
00:32:05,840 --> 00:32:07,600
Speaker 4: or when you think that the stakes are high, or

624
00:32:07,600 --> 00:32:11,440
Speaker 4: when someone might be trying to deceive you ask consciously,

625
00:32:11,480 --> 00:32:16,120
Speaker 4: explicitly whether you predicted what just happened, and if you

626
00:32:16,200 --> 00:32:20,000
Speaker 4: did predict it, then actually check it out as well.

627
00:32:20,080 --> 00:32:21,640
Speaker 4: I think a lot of times we don't even sort

628
00:32:21,680 --> 00:32:24,959
Speaker 4: of stop to wonder whether this is coming out exactly

629
00:32:24,960 --> 00:32:27,120
Speaker 4: the way I predict it, because you know, things rarely

630
00:32:27,160 --> 00:32:29,800
Speaker 4: happen exactly the way we predict, you know, especially in

631
00:32:29,800 --> 00:32:32,440
Speaker 4: an environment we don't have a lot of experience with before.

632
00:32:32,440 --> 00:32:34,640
Speaker 4: When we're doing a new experiment, testing a new theory,

633
00:32:34,920 --> 00:32:37,360
Speaker 4: should it really come out exactly like we predict I

634
00:32:37,360 --> 00:32:40,200
Speaker 4: mean maybe if we're the big best scientist ever, you know,

635
00:32:41,240 --> 00:32:43,960
Speaker 4: but often it doesn't go that way. So we should

636
00:32:44,000 --> 00:32:46,000
Speaker 4: be vigilant at those points also to see like, is

637
00:32:46,040 --> 00:32:47,800
Speaker 4: our code right, did we you know, did we make

638
00:32:47,800 --> 00:32:50,320
Speaker 4: a mistake or something like that. The example that Dan

639
00:32:50,600 --> 00:32:53,120
Speaker 4: gave about switching the columns, you know, or switching the

640
00:32:53,200 --> 00:32:56,000
Speaker 4: variable names or something like that is actually exactly what

641
00:32:56,080 --> 00:32:59,720
Speaker 4: happened in a you know, a fairly recently uncovered case

642
00:32:59,760 --> 00:33:03,080
Speaker 4: where where the data totally did not support the claim

643
00:33:03,120 --> 00:33:06,120
Speaker 4: that was being made. This was a study of the

644
00:33:06,200 --> 00:33:10,040
Speaker 4: idea that signing a declaration at the top versus at

645
00:33:10,080 --> 00:33:12,440
Speaker 4: the bottom would make you more honest in what you

646
00:33:12,520 --> 00:33:14,440
Speaker 4: declared on that form. So, in this case it was

647
00:33:14,880 --> 00:33:18,280
Speaker 4: an automobile insurance company. They were asking people to report

648
00:33:18,280 --> 00:33:20,320
Speaker 4: how many miles they had driven their vehicles in the

649
00:33:20,360 --> 00:33:24,160
Speaker 4: previous year. And the test was sign at the top

650
00:33:24,160 --> 00:33:26,160
Speaker 4: saying you're going to be honest in you're reporting, versus

651
00:33:26,160 --> 00:33:27,880
Speaker 4: sign at the bottom saying you've been honest in what

652
00:33:27,920 --> 00:33:30,680
Speaker 4: you reported. The idea was like, signing first would draw

653
00:33:30,680 --> 00:33:33,080
Speaker 4: your attention to honesty, and you'd you know, produce more

654
00:33:34,000 --> 00:33:36,440
Speaker 4: you know, more accurate, more honest results in that case.

655
00:33:36,760 --> 00:33:37,760
Speaker 3: And when.

656
00:33:39,000 --> 00:33:42,120
Speaker 4: When this experiment was done and the data file was

657
00:33:42,160 --> 00:33:45,840
Speaker 4: being looked at by some of the researchers, initially it

658
00:33:45,880 --> 00:33:47,480
Speaker 4: seemed like there was no effect at all, or the

659
00:33:47,480 --> 00:33:49,600
Speaker 4: effect was even the opposite of what they had expected.

660
00:33:50,040 --> 00:33:52,480
Speaker 4: But then one of them said, oh, well, accidentally switched

661
00:33:52,800 --> 00:33:56,360
Speaker 4: the columns, you know. So once the columns were switched back,

662
00:33:56,400 --> 00:33:58,880
Speaker 4: then the effect, you know, turned out to be right

663
00:33:58,960 --> 00:34:01,040
Speaker 4: basically exactly as Dan said, you know, you switched sort

664
00:34:01,040 --> 00:34:03,680
Speaker 4: of the you know, the treatment and the placebo in

665
00:34:03,680 --> 00:34:07,280
Speaker 4: this case, the sign first and sign and sign later columns. Well,

666
00:34:07,280 --> 00:34:09,520
Speaker 4: it turned out in retrospect that the entire data set

667
00:34:09,600 --> 00:34:13,759
Speaker 4: was fraudulent. But once they got the result that you know,

668
00:34:13,800 --> 00:34:15,960
Speaker 4: that fit the theory, or fit the prediction, or fit

669
00:34:16,000 --> 00:34:19,399
Speaker 4: the expectations, then apparently they stopped looking to see does

670
00:34:19,400 --> 00:34:21,040
Speaker 4: the rest of the data make sense? Are there any

671
00:34:21,040 --> 00:34:22,319
Speaker 4: obvious red flags in there?

672
00:34:22,360 --> 00:34:22,839
Speaker 3: And so on.

673
00:34:23,520 --> 00:34:25,759
Speaker 4: I think a perfect example of at least, you know,

674
00:34:25,840 --> 00:34:30,080
Speaker 4: some authors of that paper being satisfied that they're you know,

675
00:34:30,120 --> 00:34:33,080
Speaker 4: the theory had been confirmed and not looking more deeply enough.

676
00:34:33,440 --> 00:34:35,920
Speaker 4: Of course, you know, researchers are taught to look at

677
00:34:35,920 --> 00:34:37,920
Speaker 4: the distributions of their variables, look at all of this

678
00:34:38,040 --> 00:34:40,600
Speaker 4: kinds of stuff and so on, before getting too excited

679
00:34:40,640 --> 00:34:43,359
Speaker 4: about just you know, confirming their hypothesis. But sometimes that's

680
00:34:43,360 --> 00:34:47,359
Speaker 4: hard even for experienced scientists to do so. In our

681
00:34:47,400 --> 00:34:51,120
Speaker 4: own everyday life, we should be more aware of when

682
00:34:51,120 --> 00:34:54,120
Speaker 4: our expectations are being sort of exquisitely satisfied. That could

683
00:34:54,160 --> 00:34:58,000
Speaker 4: be someone deliberately designing something to you know, to take

684
00:34:58,040 --> 00:34:58,680
Speaker 4: advantage of.

685
00:34:58,680 --> 00:35:02,200
Speaker 5: Us, say in a much more mundane case, you know,

686
00:35:02,280 --> 00:35:05,120
Speaker 5: before you repost something or share it on social media,

687
00:35:05,320 --> 00:35:06,760
Speaker 5: just ask yourself, is it really true?

688
00:35:07,280 --> 00:35:07,440
Speaker 3: Right?

689
00:35:07,440 --> 00:35:08,839
Speaker 5: And what would I need to know to be sure

690
00:35:08,880 --> 00:35:11,239
Speaker 5: that it was really true, And that's something you can do,

691
00:35:11,280 --> 00:35:13,719
Speaker 5: whether or not you agree with it, and it just

692
00:35:13,800 --> 00:35:16,640
Speaker 5: takes a second. But once you ask that question, you

693
00:35:16,719 --> 00:35:18,719
Speaker 5: might realize, I have no idea how i'd know if

694
00:35:18,719 --> 00:35:20,400
Speaker 5: that were actually true. You know, I'd have to do

695
00:35:20,400 --> 00:35:22,000
Speaker 5: a lot of digging. And then you know, maybe just

696
00:35:22,000 --> 00:35:23,880
Speaker 5: don't reshare things that you haven't been able to verify

697
00:35:24,640 --> 00:35:27,640
Speaker 5: that might might actually help prevent the spread of information.

698
00:35:28,320 --> 00:35:30,279
Speaker 4: I think most people really do want to only share

699
00:35:30,320 --> 00:35:32,279
Speaker 4: true stuff. I don't think people deliberately want to spread

700
00:35:32,320 --> 00:35:33,920
Speaker 4: false information a lot of time. I think they're just

701
00:35:33,960 --> 00:35:35,920
Speaker 4: not often thinking like whether it might be false that

702
00:35:36,360 --> 00:35:38,240
Speaker 4: they're being swept along by other cues.

703
00:35:38,280 --> 00:35:41,000
Speaker 1: Besides that, I wanted to jump back to the science

704
00:35:41,680 --> 00:35:43,759
Speaker 1: the practice of science for just a second, which is

705
00:35:44,040 --> 00:35:46,560
Speaker 1: I was just talking some colleagues who got a big

706
00:35:46,640 --> 00:35:49,560
Speaker 1: data set and they wanted to prove something in particular

707
00:35:49,600 --> 00:35:51,600
Speaker 1: about it. They got this big thing of police records

708
00:35:51,600 --> 00:35:54,719
Speaker 1: and they had a particular thing that they wanted to demonstrate,

709
00:35:55,440 --> 00:35:59,799
Speaker 1: and they analyzed it and couldn't find evidence for their hypothesis,

710
00:36:00,040 --> 00:36:02,239
Speaker 1: figured out another way to look at it statistically, and

711
00:36:02,320 --> 00:36:05,080
Speaker 1: then another way, and they still could fine and finally

712
00:36:05,200 --> 00:36:09,040
Speaker 1: came up with some way, some statistical trick, and they

713
00:36:09,040 --> 00:36:13,000
Speaker 1: were very proud, they said, to have found finally this

714
00:36:13,160 --> 00:36:15,520
Speaker 1: evidence of this bias that they were looking for in

715
00:36:15,560 --> 00:36:18,279
Speaker 1: these police records. But it made me wonder about it,

716
00:36:18,320 --> 00:36:21,279
Speaker 1: because they clearly went in to find this thing. And

717
00:36:21,320 --> 00:36:25,319
Speaker 1: the question is did they do the right thing by

718
00:36:25,760 --> 00:36:29,160
Speaker 1: continuing to search and search and search with different statistical methods,

719
00:36:29,360 --> 00:36:32,359
Speaker 1: or is it purely that they were trying to make

720
00:36:32,400 --> 00:36:35,040
Speaker 1: the duck quack in a particular way. How do you

721
00:36:35,080 --> 00:36:37,000
Speaker 1: think about these issues?

722
00:36:37,800 --> 00:36:40,239
Speaker 5: It's a really complicated problem because, of course you want

723
00:36:40,280 --> 00:36:43,280
Speaker 5: to be able to explore your data right. The problem

724
00:36:43,360 --> 00:36:45,719
Speaker 5: comes when you don't think about all of the alternative

725
00:36:45,719 --> 00:36:47,520
Speaker 5: paths you could take to get to the outcome that

726
00:36:47,560 --> 00:36:50,560
Speaker 5: you want to report, right. And Andrew Gellman refers to

727
00:36:50,560 --> 00:36:52,640
Speaker 5: this as the garden of forking paths. And I think

728
00:36:52,640 --> 00:36:55,560
Speaker 5: it doesn't imply any sort of malicious intent or intent

729
00:36:55,640 --> 00:36:58,000
Speaker 5: to deceive at all. But we make lots of choices

730
00:36:58,040 --> 00:37:01,600
Speaker 5: along the way that can influence the result, and sometimes

731
00:37:01,600 --> 00:37:04,560
Speaker 5: we don't even think about what those choices were. So

732
00:37:04,640 --> 00:37:07,239
Speaker 5: I think the problem comes not in exploring your data

733
00:37:07,239 --> 00:37:10,400
Speaker 5: really fully. It comes in only reporting the thing that worked,

734
00:37:10,800 --> 00:37:14,120
Speaker 5: the one example, the one analysis that was successful, And

735
00:37:14,160 --> 00:37:15,960
Speaker 5: what you really want to know is, hey, is this

736
00:37:16,040 --> 00:37:19,120
Speaker 5: hypothesis robust to a whole bunch of different ways of

737
00:37:19,160 --> 00:37:21,640
Speaker 5: testing it? And it sounds like in that particular case

738
00:37:21,680 --> 00:37:23,799
Speaker 5: it wasn't at all Right. All of the other ways

739
00:37:23,840 --> 00:37:25,480
Speaker 5: you look at it, you don't find anything. You only

740
00:37:25,480 --> 00:37:28,680
Speaker 5: find it if you look in this one particular way. Well,

741
00:37:29,160 --> 00:37:31,160
Speaker 5: that would be an important thing to know, right, That

742
00:37:31,239 --> 00:37:33,799
Speaker 5: would be important for the science in the field to

743
00:37:33,920 --> 00:37:37,080
Speaker 5: know that this only works in this one study if

744
00:37:37,080 --> 00:37:39,080
Speaker 5: you measure it this way, and if you fish around enough,

745
00:37:39,080 --> 00:37:42,200
Speaker 5: you'll find something that could be consistent, which means that

746
00:37:42,200 --> 00:37:44,040
Speaker 5: maybe we shouldn't trust that a whole lot until we

747
00:37:44,040 --> 00:37:46,839
Speaker 5: can replicate with that particular way of analyzing the data

748
00:37:46,840 --> 00:37:49,839
Speaker 5: and see if that holds up consistently. We should also

749
00:37:49,880 --> 00:37:51,560
Speaker 5: check to make sure that it holds up for real

750
00:37:51,600 --> 00:37:54,080
Speaker 5: reasons as opposed to just something odd about how you've

751
00:37:54,080 --> 00:37:57,760
Speaker 5: constructed the measure. Right, it might be that it's completely

752
00:37:57,800 --> 00:38:00,359
Speaker 5: reliable when you measure it that way, but it's sort

753
00:38:00,360 --> 00:38:03,560
Speaker 5: of an artifact of the structure of data of that sort.

754
00:38:04,160 --> 00:38:06,279
Speaker 5: So I think the most powerful way to do that

755
00:38:06,320 --> 00:38:07,920
Speaker 5: to say, hey, I want to make a claim that

756
00:38:07,960 --> 00:38:11,400
Speaker 5: we've discovered some relationship with some bias. Well, if you

757
00:38:11,400 --> 00:38:14,200
Speaker 5: want to claim that it's a general truth about how

758
00:38:14,200 --> 00:38:16,439
Speaker 5: the world works, you want to be able to show

759
00:38:16,440 --> 00:38:18,279
Speaker 5: that it works under a range of different ways of

760
00:38:18,320 --> 00:38:20,680
Speaker 5: measuring it and under a range of different conditions, not

761
00:38:20,800 --> 00:38:23,880
Speaker 5: just the one special one that you identified. And I

762
00:38:23,880 --> 00:38:25,320
Speaker 5: think this has been an issue in our field for

763
00:38:25,360 --> 00:38:28,040
Speaker 5: a long time, is that you know, there's obviously a

764
00:38:28,080 --> 00:38:30,120
Speaker 5: goal to try and support the theories that you're working under.

765
00:38:30,400 --> 00:38:32,719
Speaker 5: That's a natural thing to be doing, and it's not

766
00:38:32,760 --> 00:38:34,120
Speaker 5: necessarily a terrible.

767
00:38:33,800 --> 00:38:34,399
Speaker 3: Thing to be doing.

768
00:38:35,120 --> 00:38:39,239
Speaker 5: But we haven't been completely straightforward as a field in

769
00:38:39,760 --> 00:38:43,239
Speaker 5: reporting all the things we've tried, and depending on the

770
00:38:43,320 --> 00:38:46,000
Speaker 5: kind of approach you're taking, that can be really misleading.

771
00:38:46,040 --> 00:38:47,960
Speaker 5: If you don't report everything the way it was done,

772
00:38:48,000 --> 00:38:50,640
Speaker 5: it's cherry picking in a sense. You're taking out the

773
00:38:50,640 --> 00:38:53,200
Speaker 5: results that you wanted and ignoring all the ones that

774
00:38:53,239 --> 00:38:57,080
Speaker 5: didn't work, and you're left with the reader only having

775
00:38:57,120 --> 00:38:59,560
Speaker 5: in the paper that you read about it, focusing on

776
00:38:59,600 --> 00:39:01,400
Speaker 5: the information that you've told them. And it's just like

777
00:39:01,440 --> 00:39:04,239
Speaker 5: a magician who's directed you to the thing they want

778
00:39:04,280 --> 00:39:07,160
Speaker 5: you to see and hidden all of the secret methods.

779
00:39:06,840 --> 00:39:07,719
Speaker 3: That get you there.

780
00:39:08,840 --> 00:39:11,040
Speaker 4: Yeah, you, in a way, as the researcher who did that,

781
00:39:11,040 --> 00:39:13,440
Speaker 4: have inadvertently become the con artist, although you had no

782
00:39:13,480 --> 00:39:15,880
Speaker 4: intention to do it, and you're not actually trying to deceive,

783
00:39:15,880 --> 00:39:18,080
Speaker 4: but you're accidentally sort of using some of those very

784
00:39:18,080 --> 00:39:22,239
Speaker 4: same techniques that people could use, you know, to do

785
00:39:22,320 --> 00:39:25,160
Speaker 4: worse things than publish a paper that didn't you know,

786
00:39:25,160 --> 00:39:27,920
Speaker 4: that didn't actually have good evidence for its conclusions.

787
00:39:28,560 --> 00:39:30,680
Speaker 1: Yeah, it strikes me that one of the things as

788
00:39:30,680 --> 00:39:34,719
Speaker 1: I was reading this excellent book, a lot of this

789
00:39:34,880 --> 00:39:37,520
Speaker 1: just has to do with taking the tools of good

790
00:39:37,719 --> 00:39:41,239
Speaker 1: science to the way that we interpret the world around us.

791
00:39:41,320 --> 00:39:45,040
Speaker 1: So the things about asking more questions and digging deeper

792
00:39:45,080 --> 00:39:48,439
Speaker 1: and so on. But it's interesting that even scientists don't

793
00:39:48,440 --> 00:39:49,840
Speaker 1: always do good science.

794
00:39:50,840 --> 00:39:53,880
Speaker 5: Yeah, we're all capable of being fooled, right in. Scientists

795
00:39:54,160 --> 00:39:57,359
Speaker 5: and maybe journalists are trained to dig more and ask

796
00:39:57,440 --> 00:39:59,840
Speaker 5: more questions and to think critically about what they're hearing.

797
00:40:00,640 --> 00:40:03,120
Speaker 5: But you know, we're human. We have the same sorts

798
00:40:03,160 --> 00:40:05,680
Speaker 5: of habits and ways of thinking. We tend to like

799
00:40:05,760 --> 00:40:08,600
Speaker 5: results that support what we predicted, and we're drawn to

800
00:40:08,640 --> 00:40:11,600
Speaker 5: the same kinds of information, and if somebody's looking to

801
00:40:11,640 --> 00:40:15,800
Speaker 5: sort of hide what they're doing, they can fool scientists.

802
00:40:15,840 --> 00:40:18,880
Speaker 5: And there are plenty of fraudulent papers out there that

803
00:40:19,560 --> 00:40:22,200
Speaker 5: got through peer review even though there were red flags.

804
00:40:22,239 --> 00:40:24,800
Speaker 5: And just like that sort of heist movie that you

805
00:40:24,960 --> 00:40:27,040
Speaker 5: watch from the outside and you see all of the

806
00:40:27,080 --> 00:40:28,839
Speaker 5: red flags along the way that people are falling for,

807
00:40:28,920 --> 00:40:30,839
Speaker 5: but you're not falling for them because you're watching them.

808
00:40:31,480 --> 00:40:34,400
Speaker 5: In hindsight, they're all obvious, right, but in that moment,

809
00:40:34,600 --> 00:40:37,319
Speaker 5: you don't necessarily see them as red flags until you know, oh, wait,

810
00:40:37,360 --> 00:40:39,640
Speaker 5: that paper was fraudulent. I found out through other means.

811
00:40:40,239 --> 00:40:41,960
Speaker 5: Now I can see all the red flags that are there.

812
00:40:42,360 --> 00:40:45,440
Speaker 1: Yes, So as I was reading the book, the way

813
00:40:45,480 --> 00:40:48,000
Speaker 1: I was thinking about it was, you know, the brain

814
00:40:48,120 --> 00:40:51,200
Speaker 1: is of course locked in silence and darkness inside the skull,

815
00:40:51,239 --> 00:40:54,279
Speaker 1: and it's just trying to make an internal model of

816
00:40:54,320 --> 00:40:56,800
Speaker 1: the world out there, a mental model, and we're always

817
00:40:57,200 --> 00:41:01,240
Speaker 1: we're always very limited in what our internal models can detect,

818
00:41:01,360 --> 00:41:05,640
Speaker 1: can see, And so one of the most important things

819
00:41:05,680 --> 00:41:11,920
Speaker 1: to expanding our model is to ask questions. And in

820
00:41:11,960 --> 00:41:14,680
Speaker 1: a sense this is the same as paying attention to something.

821
00:41:14,880 --> 00:41:17,400
Speaker 1: We ask a question that forces us to attend to

822
00:41:17,480 --> 00:41:20,960
Speaker 1: some aspect and then that updates our model a little bit.

823
00:41:21,239 --> 00:41:25,160
Speaker 1: So in chapter four, you guys had an example of

824
00:41:25,200 --> 00:41:29,320
Speaker 1: a chess grand master that asks his students to always

825
00:41:29,360 --> 00:41:31,720
Speaker 1: ask three questions when they're looking at the board?

826
00:41:31,800 --> 00:41:33,160
Speaker 2: Can you tell us about that?

827
00:41:34,560 --> 00:41:36,160
Speaker 4: Would I would love to tell you. I would love

828
00:41:36,200 --> 00:41:40,800
Speaker 4: to tell you about that. So I actually took a

829
00:41:40,920 --> 00:41:43,760
Speaker 4: during COVID, I took a summer chess camp on Zoom

830
00:41:43,800 --> 00:41:48,000
Speaker 4: with this guy and it was me and like twelve

831
00:41:48,040 --> 00:41:51,080
Speaker 4: people aged ten and under, which was a fun you know,

832
00:41:51,120 --> 00:41:53,240
Speaker 4: which was a fun again because who goes to summer

833
00:41:53,280 --> 00:41:55,680
Speaker 4: chess camps, right, It's like, you know, kids gauge ten

834
00:41:55,719 --> 00:41:59,400
Speaker 4: and under. Very good players by the way. And you

835
00:41:59,400 --> 00:42:01,759
Speaker 4: know one thing that he often that the coach would

836
00:42:01,760 --> 00:42:03,600
Speaker 4: often do. His name is jako Ogard. He's one of

837
00:42:03,680 --> 00:42:05,480
Speaker 4: the most famous chess coaches in the world, and it

838
00:42:05,560 --> 00:42:07,239
Speaker 4: was privileged to be able to sort of be in.

839
00:42:07,200 --> 00:42:08,320
Speaker 3: His camp for a few hours.

840
00:42:08,600 --> 00:42:11,200
Speaker 4: He would constantly say, we need more, you know, you

841
00:42:11,239 --> 00:42:13,280
Speaker 4: need to think of more moves. Right in a chess position,

842
00:42:13,360 --> 00:42:15,399
Speaker 4: there's like thirty to forty moves you can typically play,

843
00:42:15,600 --> 00:42:18,160
Speaker 4: you know, with all your pieces and people are often

844
00:42:18,160 --> 00:42:21,160
Speaker 4: become too focused on one. So he would say, think

845
00:42:21,200 --> 00:42:23,080
Speaker 4: of more candidate moves. Think of more moves you might

846
00:42:23,120 --> 00:42:25,000
Speaker 4: want to play and analyze. And if you're having trouble

847
00:42:25,000 --> 00:42:27,560
Speaker 4: thinking of them, he has specific questions that you can

848
00:42:27,640 --> 00:42:30,080
Speaker 4: use to try to generate ideas, and one of them

849
00:42:30,160 --> 00:42:33,399
Speaker 4: is what's your worst place piece? Maybe you should move

850
00:42:33,400 --> 00:42:35,080
Speaker 4: it if that's the one that's in the worst position.

851
00:42:35,440 --> 00:42:37,600
Speaker 4: Or what's the opponent's idea, Well, maybe you should come

852
00:42:37,640 --> 00:42:39,839
Speaker 4: up with a move that stops their idea. The third

853
00:42:39,880 --> 00:42:41,520
Speaker 4: one is what are the weaknesses? Maybe you should come

854
00:42:41,600 --> 00:42:43,239
Speaker 4: up with a move that attacks something that's weak. I

855
00:42:43,280 --> 00:42:45,440
Speaker 4: mean this is this will make sense to people who

856
00:42:45,520 --> 00:42:48,000
Speaker 4: play chess, but these are they're in almost all fields.

857
00:42:48,000 --> 00:42:50,520
Speaker 4: There are sort of general kinds of things you can

858
00:42:50,520 --> 00:42:53,319
Speaker 4: look at and principles you can use to generate, you know,

859
00:42:53,360 --> 00:42:55,360
Speaker 4: to generate more information and as you say, like you

860
00:42:55,840 --> 00:42:57,520
Speaker 4: I like your way putting it to sort of improve

861
00:42:57,600 --> 00:43:00,440
Speaker 4: your mental model of what's really going on, because in

862
00:43:00,520 --> 00:43:02,880
Speaker 4: order to in order to play good chess moves, you

863
00:43:02,960 --> 00:43:05,000
Speaker 4: have to have a good internal model of what's going

864
00:43:05,040 --> 00:43:07,839
Speaker 4: on on the board someplace in your brain. You've got

865
00:43:07,880 --> 00:43:09,840
Speaker 4: to have it and that's a way of sort of

866
00:43:09,880 --> 00:43:13,640
Speaker 4: generating more ideas, more analysis that then updates that then

867
00:43:13,719 --> 00:43:14,520
Speaker 4: updates the model.

868
00:43:15,400 --> 00:43:17,040
Speaker 5: And most of the time the models that we have

869
00:43:17,160 --> 00:43:18,680
Speaker 5: for how the world works are pretty.

870
00:43:18,400 --> 00:43:18,759
Speaker 3: Good, right.

871
00:43:18,800 --> 00:43:22,359
Speaker 5: I mean, we're not, you know, constantly getting conned. We're

872
00:43:22,360 --> 00:43:25,200
Speaker 5: not you know, we don't have trouble getting around, we

873
00:43:25,200 --> 00:43:28,520
Speaker 5: don't have trouble communicating with other people. Most of the time,

874
00:43:28,520 --> 00:43:31,680
Speaker 5: our models of how the world is working are great.

875
00:43:31,800 --> 00:43:34,560
Speaker 5: They work very effectively. And it's only in those cases

876
00:43:34,600 --> 00:43:36,520
Speaker 5: where we need to dig a little more to update

877
00:43:36,520 --> 00:43:39,320
Speaker 5: our model for the possibility we're being teated or deceived

878
00:43:40,120 --> 00:43:42,879
Speaker 5: that we need to ask a lot more questions. Right,

879
00:43:43,280 --> 00:43:45,000
Speaker 5: most of the time, we've built up these models from

880
00:43:45,040 --> 00:43:48,279
Speaker 5: a ton of experience, and they generally do okay, Yeah.

881
00:43:48,080 --> 00:43:52,280
Speaker 1: And we have expectations about what we're looking for given

882
00:43:52,320 --> 00:43:55,840
Speaker 1: these models, such that most of the time we're filling

883
00:43:55,880 --> 00:43:58,839
Speaker 1: in the blanks. And that is at the heart of

884
00:43:58,880 --> 00:44:01,600
Speaker 1: all these hoaxes and scams that you talk about throughout

885
00:44:01,600 --> 00:44:03,840
Speaker 1: the book, is we're filling in the blanks, and the

886
00:44:03,920 --> 00:44:06,799
Speaker 1: things that aren't said, we assume we.

887
00:44:06,719 --> 00:44:08,160
Speaker 2: Know what they mean.

888
00:44:08,480 --> 00:44:11,760
Speaker 1: So I'm curious what you guys think about the Turing

889
00:44:11,880 --> 00:44:16,520
Speaker 1: tests and for the listeners in case someone doesn't know.

890
00:44:16,719 --> 00:44:19,760
Speaker 1: The Turing test was proposed by Alan Turning to figure

891
00:44:19,760 --> 00:44:22,719
Speaker 1: out when a machine has become as smart as the human.

892
00:44:22,800 --> 00:44:26,239
Speaker 1: The idea is that you are the evaluator and you're

893
00:44:26,280 --> 00:44:28,760
Speaker 1: talking to a machine, and you're talking to a human,

894
00:44:29,160 --> 00:44:31,640
Speaker 1: let's say, by text, and you don't know which is which,

895
00:44:31,680 --> 00:44:33,640
Speaker 1: and the question is can you tell the difference between

896
00:44:33,640 --> 00:44:36,279
Speaker 1: the human and the machine. And the interesting part is,

897
00:44:36,360 --> 00:44:38,799
Speaker 1: because we bring so much to the table in any

898
00:44:38,840 --> 00:44:42,319
Speaker 1: conversation and we fill in the blanks, what do you

899
00:44:42,320 --> 00:44:45,040
Speaker 1: guys think Is that a good, meaningful test or is

900
00:44:45,080 --> 00:44:46,120
Speaker 1: it flawed in that way?

901
00:44:46,920 --> 00:44:49,080
Speaker 4: I think we have a lot of evidence that it's

902
00:44:49,120 --> 00:44:53,560
Speaker 4: not so good from chat, JPT and large language models,

903
00:44:53,560 --> 00:44:56,640
Speaker 4: which I don't think are actually intelligent in the way

904
00:44:56,680 --> 00:44:59,080
Speaker 4: that we should. I mean, I realized there might be

905
00:44:59,080 --> 00:45:00,800
Speaker 4: some dispute about this. Some people make some sort of

906
00:45:00,840 --> 00:45:04,600
Speaker 4: extravagant claims about signs of general intelligence, but I don't

907
00:45:04,600 --> 00:45:06,920
Speaker 4: think they're actually intelligent, or at least not in a

908
00:45:07,040 --> 00:45:09,279
Speaker 4: useful way. And yet they are extremely convincing. I mean,

909
00:45:09,320 --> 00:45:11,360
Speaker 4: I think they show that you can sort of dissociate,

910
00:45:11,840 --> 00:45:15,239
Speaker 4: you know, producing what humans expect to see next, which

911
00:45:15,280 --> 00:45:18,680
Speaker 4: is basically, you know, basically what large language models do

912
00:45:18,760 --> 00:45:21,040
Speaker 4: because they've been trained to sort of, you know, output

913
00:45:21,120 --> 00:45:23,319
Speaker 4: the most probable next token or word and so on.

914
00:45:23,560 --> 00:45:27,480
Speaker 4: You can dissociate that capability from having sort of a

915
00:45:27,520 --> 00:45:30,359
Speaker 4: true understanding of what's going on. For example, you could

916
00:45:30,440 --> 00:45:32,600
Speaker 4: ask an LLLM to play chess with you, and it

917
00:45:32,640 --> 00:45:34,279
Speaker 4: wouldn't do very well. It would produce a lot of

918
00:45:34,280 --> 00:45:36,080
Speaker 4: stuff that sounds like chess and so on. But if

919
00:45:36,080 --> 00:45:38,239
Speaker 4: you really knew the game of chess, you would know

920
00:45:38,320 --> 00:45:40,279
Speaker 4: that this is sort of gibberish. If you don't know,

921
00:45:40,400 --> 00:45:43,040
Speaker 4: it sounds perfectly good, right, You just you just sort

922
00:45:43,080 --> 00:45:45,040
Speaker 4: of you fill it in with sort of the assumption

923
00:45:45,120 --> 00:45:47,840
Speaker 4: that this guy sounds like he knows what he's talking about,

924
00:45:48,200 --> 00:45:51,359
Speaker 4: you know, which is which is not not you know,

925
00:45:51,440 --> 00:45:54,480
Speaker 4: not necessarily intelligence things.

926
00:45:54,280 --> 00:45:59,759
Speaker 5: Like chat GPT they speak with absolute confidence and certainty, right,

927
00:46:00,200 --> 00:46:03,280
Speaker 5: and that's regardless of whether or not they're generating true content.

928
00:46:03,920 --> 00:46:06,920
Speaker 5: You know, they're the consonant bullshitter right in that they

929
00:46:07,440 --> 00:46:10,600
Speaker 5: are equally confident when they're completely wrong and when they're

930
00:46:10,600 --> 00:46:14,239
Speaker 5: completely right. Because there's no grounding to any sort of

931
00:46:14,280 --> 00:46:16,719
Speaker 5: reality in the world. All they're doing is predicting what

932
00:46:16,760 --> 00:46:17,320
Speaker 5: comes next.

933
00:46:18,280 --> 00:46:20,319
Speaker 1: That's true, although I have to say one of the

934
00:46:20,360 --> 00:46:24,440
Speaker 1: things about chat GPT that I've really come to appreciate

935
00:46:24,560 --> 00:46:28,600
Speaker 1: is that almost any question that you ask it, it'll say, look,

936
00:46:28,680 --> 00:46:31,960
Speaker 1: some people think this, some people think that, and in conclusion,

937
00:46:32,080 --> 00:46:35,120
Speaker 1: we need to balance these points of view. And at

938
00:46:35,160 --> 00:46:37,759
Speaker 1: first I found that really annoying, but I came to

939
00:46:37,920 --> 00:46:42,400
Speaker 1: understand and appreciate that it is trying, you know, because

940
00:46:42,440 --> 00:46:45,239
Speaker 1: of the reinforcement learning and so on, it's trying to

941
00:46:46,160 --> 00:46:49,680
Speaker 1: give different perspectives instead of sounding totally confident about just

942
00:46:49,760 --> 00:46:50,600
Speaker 1: a single answer.

943
00:46:52,320 --> 00:46:53,920
Speaker 4: You kind of wish politicians would talk to you that

944
00:46:53,960 --> 00:46:57,879
Speaker 4: way sometimes exactly instead of the way they do it. Yeah,

945
00:46:57,920 --> 00:47:01,879
Speaker 4: well it's it's it's reasonable. I just don't think chet

946
00:47:01,920 --> 00:47:04,600
Speaker 4: GPT believes that both of those things are equally you know,

947
00:47:04,640 --> 00:47:06,440
Speaker 4: are equally likely to be true and so on in

948
00:47:06,480 --> 00:47:09,200
Speaker 4: any in any meaningful way. I think, you know, it's

949
00:47:09,880 --> 00:47:13,280
Speaker 4: the part of the danger I think of of models

950
00:47:13,320 --> 00:47:16,200
Speaker 4: like this is if you don't understand how they work,

951
00:47:17,040 --> 00:47:19,000
Speaker 4: and if you just see this is a I like

952
00:47:19,040 --> 00:47:20,560
Speaker 4: a lot of times you see these posts on Twitter

953
00:47:20,640 --> 00:47:24,160
Speaker 4: that say and A n AI did this that's just

954
00:47:24,239 --> 00:47:26,839
Speaker 4: designed to impress you, right and to mislead you by

955
00:47:26,880 --> 00:47:29,840
Speaker 4: thinking that therefore the results must be created by genius

956
00:47:29,880 --> 00:47:32,479
Speaker 4: and totally accurate. But if you actually understand something about

957
00:47:32,480 --> 00:47:34,600
Speaker 4: how it works, then you will have the reaction you did.

958
00:47:34,640 --> 00:47:36,600
Speaker 3: You'll say, well, this is this is kind.

959
00:47:36,440 --> 00:47:38,319
Speaker 4: Of interesting that it's trying to give you know, sort

960
00:47:38,360 --> 00:47:41,799
Speaker 4: of too equally like probable you know, schools of thought here,

961
00:47:42,000 --> 00:47:44,480
Speaker 4: or you know common you know, views on the topic

962
00:47:44,600 --> 00:47:47,040
Speaker 4: or something like that, but you understand something about how

963
00:47:47,080 --> 00:47:49,520
Speaker 4: they work, so it doesn't you know, they their output

964
00:47:49,600 --> 00:47:51,799
Speaker 4: sort of is more sensible to you than it might

965
00:47:51,840 --> 00:47:53,160
Speaker 4: be to people who don't know.

966
00:48:08,239 --> 00:48:10,840
Speaker 1: I want to come back to just something you mentioned

967
00:48:10,840 --> 00:48:13,160
Speaker 1: about politicians. One of the things I was thinking about

968
00:48:13,239 --> 00:48:16,520
Speaker 1: as I was reading the book was politicians often get

969
00:48:16,560 --> 00:48:19,480
Speaker 1: points deducted if they change their mind on something.

970
00:48:19,520 --> 00:48:20,880
Speaker 2: They're called flip floppers.

971
00:48:21,520 --> 00:48:26,000
Speaker 1: And it's such a shame because we know that if

972
00:48:26,000 --> 00:48:30,040
Speaker 1: people are using scientific reasoning, they might reasonably change their

973
00:48:30,160 --> 00:48:32,840
Speaker 1: mind about some issue.

974
00:48:32,719 --> 00:48:33,560
Speaker 2: At some point.

975
00:48:34,080 --> 00:48:36,479
Speaker 1: The question is, how would you guys see a way

976
00:48:37,040 --> 00:48:40,520
Speaker 1: to change something about the way we educate the public

977
00:48:40,719 --> 00:48:44,240
Speaker 1: so that politicians who change their mind on a topic

978
00:48:44,680 --> 00:48:47,760
Speaker 1: are not considered flip floppers and deducted points.

979
00:48:48,560 --> 00:48:53,239
Speaker 4: Now, I guess I would say that this refers back

980
00:48:53,280 --> 00:48:56,960
Speaker 4: to our taste for consistency. So there's something appealing in

981
00:48:57,120 --> 00:49:00,360
Speaker 4: consistency of a wide variety and consistency and a person's

982
00:49:00,400 --> 00:49:03,919
Speaker 4: behavior is also appealing to us. And many times that's good.

983
00:49:04,000 --> 00:49:04,120
Speaker 3: Right.

984
00:49:04,160 --> 00:49:06,200
Speaker 4: You want someone who keeps their word. You want someone

985
00:49:06,239 --> 00:49:08,040
Speaker 4: who says what they are going to do and then

986
00:49:08,040 --> 00:49:09,600
Speaker 4: does what they said they would do. You want someone

987
00:49:09,600 --> 00:49:12,240
Speaker 4: who's always on time. Like those are generally positive things.

988
00:49:12,560 --> 00:49:16,240
Speaker 4: But when you're talking about complex subjects like what should

989
00:49:16,239 --> 00:49:19,000
Speaker 4: climate policy be? You know, what should tax policy be?

990
00:49:19,160 --> 00:49:22,000
Speaker 4: Any of these kinds of things and so on, facts

991
00:49:22,080 --> 00:49:25,680
Speaker 4: change over time, you know, and that alone, never mind

992
00:49:25,880 --> 00:49:27,840
Speaker 4: you know, using scientific reasoning, but just the changing of

993
00:49:27,880 --> 00:49:31,480
Speaker 4: facts might, you know, might change people's views. I think

994
00:49:32,920 --> 00:49:37,960
Speaker 4: teaching people about the trap of consistency, you know, may

995
00:49:38,000 --> 00:49:39,879
Speaker 4: start to help. I'm not sure there's like an easy

996
00:49:39,960 --> 00:49:42,440
Speaker 4: nudge that is going to make people suddenly prefer the

997
00:49:42,440 --> 00:49:45,000
Speaker 4: flip flopper, because, after all, sometimes the flip flopper is

998
00:49:45,040 --> 00:49:47,160
Speaker 4: just being expedient, right. It's not as simple as saying, like, well,

999
00:49:47,160 --> 00:49:49,320
Speaker 4: we should anytime someone flip flops. We should just assume

1000
00:49:49,400 --> 00:49:52,600
Speaker 4: that they are have done some complex ratios nation and

1001
00:49:52,680 --> 00:49:55,719
Speaker 4: integration who do information, and now they've changed their mind.

1002
00:49:55,760 --> 00:49:57,600
Speaker 4: It could be they're just saying something to a different

1003
00:49:57,600 --> 00:50:00,560
Speaker 4: audience because that's what is expedient for them at the moment.

1004
00:50:00,640 --> 00:50:03,040
Speaker 4: It's a tough problem to sort out, but certainly I

1005
00:50:03,040 --> 00:50:05,080
Speaker 4: don't think we should have the bias against changing one's

1006
00:50:05,120 --> 00:50:06,640
Speaker 4: mind that it is really seems to be built into

1007
00:50:06,640 --> 00:50:08,960
Speaker 4: the political system, because I do think it contributes to

1008
00:50:08,960 --> 00:50:11,719
Speaker 4: some polarization also, Right, in order to be a consistent conservative,

1009
00:50:11,719 --> 00:50:13,000
Speaker 4: you always have to do this, in order to be

1010
00:50:13,080 --> 00:50:17,640
Speaker 4: a strong you know, it's definitely it definitely has its cost.

1011
00:50:18,440 --> 00:50:21,560
Speaker 5: Well, and you know, scientists also aren't necessarily the greatest

1012
00:50:21,600 --> 00:50:23,160
Speaker 5: at updating their beliefs in light of evidence.

1013
00:50:23,400 --> 00:50:23,600
Speaker 2: Right.

1014
00:50:24,600 --> 00:50:26,840
Speaker 5: There are plenty of people who get evidence that should

1015
00:50:26,880 --> 00:50:29,560
Speaker 5: contradict their claims, but they continue arguing for the old

1016
00:50:29,760 --> 00:50:33,719
Speaker 5: position without fully updating their beliefs accordingly, Right, every time

1017
00:50:33,760 --> 00:50:36,879
Speaker 5: somebody encounters have failure to replicate their own work, right,

1018
00:50:37,480 --> 00:50:39,560
Speaker 5: what's their reaction, Well, I'm going to try and find

1019
00:50:39,560 --> 00:50:41,200
Speaker 5: all of the things that were done differently in that

1020
00:50:41,760 --> 00:50:44,880
Speaker 5: replication attempt that could maybe excuse why they didn't find

1021
00:50:45,520 --> 00:50:48,360
Speaker 5: what I found, And what they really should do is say, Okay,

1022
00:50:48,640 --> 00:50:50,279
Speaker 5: you know that may or may not be right. I

1023
00:50:50,360 --> 00:50:52,600
Speaker 5: might disagree with it, but it should make me a

1024
00:50:52,640 --> 00:50:55,879
Speaker 5: little less likely to believe in my original result. And

1025
00:50:55,920 --> 00:50:58,080
Speaker 5: maybe there are alternatives and I need to go test those,

1026
00:50:58,120 --> 00:51:00,880
Speaker 5: and if those alternatives don't work out, then I should

1027
00:51:00,880 --> 00:51:02,759
Speaker 5: be changing my beliefs. But you don't see that all

1028
00:51:02,760 --> 00:51:05,360
Speaker 5: that often. You don't see the person going back to

1029
00:51:05,360 --> 00:51:07,440
Speaker 5: the study and saying, Okay, I'm going to prove that

1030
00:51:07,520 --> 00:51:10,799
Speaker 5: it was this moderation that explained why they get They

1031
00:51:10,800 --> 00:51:13,200
Speaker 5: didn't get it, but I did. More often than not,

1032
00:51:13,320 --> 00:51:16,600
Speaker 5: you have a dismissal Right, Well, that's really not all

1033
00:51:16,600 --> 00:51:19,920
Speaker 5: that different than what a politician does when they refuse

1034
00:51:19,960 --> 00:51:21,920
Speaker 5: to change their views in light of new data.

1035
00:51:22,200 --> 00:51:23,400
Speaker 3: Yeah, you know.

1036
00:51:23,440 --> 00:51:25,919
Speaker 1: I think this is back to this issue of complexity

1037
00:51:25,960 --> 00:51:29,680
Speaker 1: in science, which is that it essentially works by the

1038
00:51:29,719 --> 00:51:33,640
Speaker 1: advocacy system, which is that you're supposed to defend an

1039
00:51:33,719 --> 00:51:36,200
Speaker 1: idea all the way down until you can't defend it

1040
00:51:36,200 --> 00:51:38,239
Speaker 1: anymore and then you give up on it. But so

1041
00:51:38,400 --> 00:51:41,520
Speaker 1: when someone says, oh, I didn't replicate your thing, you

1042
00:51:41,600 --> 00:51:43,680
Speaker 1: might be the only one who says, hey, I'm willing

1043
00:51:43,719 --> 00:51:47,040
Speaker 1: to defend this and really fight for the idea until

1044
00:51:47,080 --> 00:51:49,759
Speaker 1: I reached some point. The question is what is the

1045
00:51:49,800 --> 00:51:54,040
Speaker 1: proper point? And as we said, scientists are humans too,

1046
00:51:54,120 --> 00:51:57,400
Speaker 1: and so they really care about their reputation and their

1047
00:51:57,440 --> 00:51:58,480
Speaker 1: previous publications.

1048
00:51:59,160 --> 00:52:01,440
Speaker 5: And that's fine, So it's what's what's the point? But

1049
00:52:01,480 --> 00:52:04,160
Speaker 5: also what evidence do you use to defend it?

1050
00:52:04,360 --> 00:52:05,040
Speaker 3: Yeah? Right?

1051
00:52:05,200 --> 00:52:08,040
Speaker 5: So if if the evidence you used to defend it is, hey,

1052
00:52:08,040 --> 00:52:10,319
Speaker 5: we can still get our result and we can show

1053
00:52:10,360 --> 00:52:13,400
Speaker 5: why you didn't get it, that's great, right, Then you

1054
00:52:13,560 --> 00:52:16,719
Speaker 5: then you've bolstered your position. You've let everybody to update

1055
00:52:16,719 --> 00:52:19,560
Speaker 5: with their beliefs. Are If your defense is they're incompetent,

1056
00:52:20,400 --> 00:52:22,680
Speaker 5: that's not a great defense. It might be true, but

1057
00:52:23,280 --> 00:52:25,040
Speaker 5: you'd have to show why and then show that if

1058
00:52:25,040 --> 00:52:26,640
Speaker 5: you do it in the incompetent way, you don't get

1059
00:52:26,680 --> 00:52:27,080
Speaker 5: the result.

1060
00:52:28,480 --> 00:52:32,240
Speaker 4: Well, according to the adage that science progresses one funeral

1061
00:52:32,239 --> 00:52:34,799
Speaker 4: at a time, that point is the point of death. Right,

1062
00:52:34,920 --> 00:52:37,200
Speaker 4: And so there is a generational aspect to some of

1063
00:52:37,239 --> 00:52:39,160
Speaker 4: these to some of these things. I think there's there, seriously,

1064
00:52:39,200 --> 00:52:41,960
Speaker 4: is right, there's sort of generations. Generations often are associated

1065
00:52:41,960 --> 00:52:44,520
Speaker 4: with particular schools of thought or views that sort of

1066
00:52:44,560 --> 00:52:46,640
Speaker 4: pass and.

1067
00:52:45,760 --> 00:52:46,400
Speaker 3: And and evolved.

1068
00:52:46,400 --> 00:52:49,080
Speaker 4: It would be great if we could get there before then, right,

1069
00:52:50,080 --> 00:52:54,280
Speaker 4: So we should get there before that point. But it's

1070
00:52:54,400 --> 00:52:56,200
Speaker 4: I don't think I'm not so bleak. I'm not so

1071
00:52:56,280 --> 00:52:58,520
Speaker 4: blik on this maybe as some people, because I think

1072
00:52:58,520 --> 00:52:59,920
Speaker 4: a lot of times what happens is people just sort

1073
00:52:59,920 --> 00:53:01,960
Speaker 4: of drop out of the debate. Right. They may not

1074
00:53:02,000 --> 00:53:04,440
Speaker 4: have changed their mind, but they don't become an important

1075
00:53:04,440 --> 00:53:06,799
Speaker 4: person in that in that area anymore, or at least

1076
00:53:06,840 --> 00:53:10,080
Speaker 4: the next generation who's doing all the interesting work doesn't

1077
00:53:10,120 --> 00:53:13,840
Speaker 4: take those people that seriously anymore. They're still alive where

1078
00:53:13,880 --> 00:53:16,399
Speaker 4: you know, their funeral hasn't happened yet, but they maybe

1079
00:53:16,440 --> 00:53:18,279
Speaker 4: moved on to another topic. They're doing something else in

1080
00:53:18,320 --> 00:53:20,640
Speaker 4: science maybe, but they're not like exerting like an iron

1081
00:53:21,280 --> 00:53:23,880
Speaker 4: you know, an iron fist, you know, rule over some area,

1082
00:53:23,920 --> 00:53:26,120
Speaker 4: like when a science works that way, that's pathological.

1083
00:53:26,200 --> 00:53:26,359
Speaker 1: Right.

1084
00:53:26,440 --> 00:53:29,040
Speaker 4: But you know when when someone's views like you know,

1085
00:53:29,160 --> 00:53:30,799
Speaker 4: must be then then we're you know, you're not talking

1086
00:53:30,840 --> 00:53:32,359
Speaker 4: about science anymore, right. I think a lot of people

1087
00:53:32,440 --> 00:53:35,200
Speaker 4: just sort of move on instead of before it's too late.

1088
00:53:35,640 --> 00:53:37,759
Speaker 1: Yeah, And this is the great part about science is

1089
00:53:37,800 --> 00:53:41,120
Speaker 1: that it is the only endeavor that's constantly knocking down

1090
00:53:41,160 --> 00:53:41,759
Speaker 1: its own wall.

1091
00:53:41,920 --> 00:53:44,800
Speaker 2: So with enough time, the truth will out.

1092
00:53:45,239 --> 00:53:48,799
Speaker 1: Things are changing really rapidly ever since you, you know,

1093
00:53:48,880 --> 00:53:51,759
Speaker 1: finished writing the book and sent it off, And so

1094
00:53:51,920 --> 00:53:54,080
Speaker 1: one of the things that I want to ask about

1095
00:53:54,080 --> 00:53:57,040
Speaker 1: are things like what's happening with AI and deep fakes.

1096
00:53:57,120 --> 00:54:00,200
Speaker 1: What are the new kinds of hoaxes that you see

1097
00:54:00,360 --> 00:54:02,640
Speaker 1: coming down the line.

1098
00:54:02,920 --> 00:54:05,319
Speaker 5: Well, I'll raise one that's not new, And I think

1099
00:54:05,360 --> 00:54:08,360
Speaker 5: it's important to point out that none of the hoaxes

1100
00:54:08,400 --> 00:54:11,200
Speaker 5: that have happened over the thousands of years are really

1101
00:54:11,280 --> 00:54:13,480
Speaker 5: fundamentally all that different in the way that they take

1102
00:54:13,520 --> 00:54:16,480
Speaker 5: advantage of our tendencies, right, And that's the thing that

1103
00:54:16,480 --> 00:54:19,160
Speaker 5: we noticed across all of these different domains, from chess

1104
00:54:19,200 --> 00:54:22,600
Speaker 5: to sports, to art to science, that they all take

1105
00:54:22,640 --> 00:54:25,720
Speaker 5: advantage of the same sorts of tendencies. And new scams

1106
00:54:25,719 --> 00:54:27,239
Speaker 5: are going to do that too, they just might do

1107
00:54:27,280 --> 00:54:30,800
Speaker 5: it more effectively. And even the Nigerian Prince email scam

1108
00:54:31,080 --> 00:54:35,279
Speaker 5: was originally a Nigerian prince mail scam back when people

1109
00:54:35,320 --> 00:54:35,960
Speaker 5: sent letters.

1110
00:54:36,520 --> 00:54:38,080
Speaker 3: It's more effective in.

1111
00:54:38,000 --> 00:54:40,400
Speaker 5: That they don't have to spend as much time and

1112
00:54:40,440 --> 00:54:44,640
Speaker 5: effort finding potential victims. Right, some of these scams with

1113
00:54:44,719 --> 00:54:47,840
Speaker 5: the advent of AI are going to become more effective.

1114
00:54:48,040 --> 00:54:48,200
Speaker 3: Right.

1115
00:54:48,239 --> 00:54:51,680
Speaker 5: So one that's common right now is it's either the

1116
00:54:51,880 --> 00:54:55,759
Speaker 5: you know your kid's been arrested scam or has been

1117
00:54:55,800 --> 00:54:58,160
Speaker 5: in an accident or is being held hostage. It's a

1118
00:54:58,200 --> 00:55:00,840
Speaker 5: horrible thing, preying on people's fears. They'll call up a

1119
00:55:00,920 --> 00:55:04,359
Speaker 5: parent or a grandparent and say that the kid needs

1120
00:55:04,400 --> 00:55:07,399
Speaker 5: to be bailed out immediately, right, and often they'll call up,

1121
00:55:07,600 --> 00:55:10,600
Speaker 5: you know, a relative like a cousin or something. Now,

1122
00:55:10,600 --> 00:55:14,879
Speaker 5: that scam's pretty effective because people want to quickly solve

1123
00:55:14,880 --> 00:55:17,640
Speaker 5: this problem, right, They want to quickly fix what is wrong,

1124
00:55:18,520 --> 00:55:21,240
Speaker 5: and often we don't have the preventive measures in place

1125
00:55:21,960 --> 00:55:25,040
Speaker 5: to stop that. But imagine how much more powerful that

1126
00:55:25,239 --> 00:55:27,719
Speaker 5: is if you're using AI voice synthesis to make the call,

1127
00:55:28,200 --> 00:55:30,920
Speaker 5: right and it actually sounds like it's coming from that person.

1128
00:55:31,800 --> 00:55:34,399
Speaker 5: That's going to ramp that one up, same principles, it's

1129
00:55:34,440 --> 00:55:35,080
Speaker 5: just more potent.

1130
00:55:36,280 --> 00:55:36,719
Speaker 3: I think a.

1131
00:55:38,440 --> 00:55:40,759
Speaker 4: Whole area which is rife with scams, which is a

1132
00:55:40,800 --> 00:55:43,560
Speaker 4: new area, but you know, sort of re scamming based

1133
00:55:43,560 --> 00:55:47,520
Speaker 4: on old principles as cryptocurrency, So there are thousands and

1134
00:55:47,560 --> 00:55:51,080
Speaker 4: thousands and thousands of cryptocurrencies and coins being issued and

1135
00:55:51,640 --> 00:55:54,160
Speaker 4: so on, and you know, as far as I can tell,

1136
00:55:54,400 --> 00:55:58,040
Speaker 4: most of that is mostly fraud. But yet it relies

1137
00:55:58,040 --> 00:55:59,640
Speaker 4: on all the same principles. You've got sort of like

1138
00:55:59,680 --> 00:56:03,120
Speaker 4: famili your celebrities advertising these things. You've got time pressure,

1139
00:56:03,160 --> 00:56:04,920
Speaker 4: there's a limited offering. You know, you've got to make

1140
00:56:04,920 --> 00:56:09,080
Speaker 4: a decision. Now, You've got these sort of like fake consistency,

1141
00:56:09,200 --> 00:56:11,319
Speaker 4: Like people will claim that like our crypto fund is,

1142
00:56:11,480 --> 00:56:13,520
Speaker 4: you know, never had a down month in all of

1143
00:56:13,520 --> 00:56:15,719
Speaker 4: its three months of existence or something like that, you know,

1144
00:56:15,800 --> 00:56:18,319
Speaker 4: but consistently going up, and all the same stuff just

1145
00:56:18,360 --> 00:56:21,040
Speaker 4: being applied to a whole new being applied to a

1146
00:56:21,040 --> 00:56:23,120
Speaker 4: whole new thing. And I don't really see that, you know,

1147
00:56:23,160 --> 00:56:26,560
Speaker 4: getting any better. I as far as AI in deep

1148
00:56:26,560 --> 00:56:30,359
Speaker 4: fakes and so on, I do have some optimism that

1149
00:56:30,920 --> 00:56:33,960
Speaker 4: it's going to increase the value of truly trusted sources

1150
00:56:33,960 --> 00:56:35,480
Speaker 4: who bother to check that stuff.

1151
00:56:35,920 --> 00:56:36,600
Speaker 3: Right, So I.

1152
00:56:36,640 --> 00:56:39,200
Speaker 4: Noticed during there were not long ago there but this

1153
00:56:39,239 --> 00:56:42,120
Speaker 4: is sort of pseudo attempted coup revolution weird thing that

1154
00:56:42,200 --> 00:56:46,279
Speaker 4: happened in Russia. You know, a sort of paramilitary group

1155
00:56:46,360 --> 00:56:49,080
Speaker 4: kind of turned on the military and started marching to

1156
00:56:49,160 --> 00:56:51,799
Speaker 4: Moscow and so on. And I was fascinated by this

1157
00:56:51,920 --> 00:56:53,640
Speaker 4: and paying attention to Twitter, and there were all kinds

1158
00:56:53,640 --> 00:56:56,879
Speaker 4: of reports on Twitter, people claiming to be like eyewitnesses

1159
00:56:56,920 --> 00:56:59,560
Speaker 4: to things and so on, and very little of that

1160
00:56:59,600 --> 00:57:02,839
Speaker 4: made it to the mainstream media or to legitimate sources.

1161
00:57:03,400 --> 00:57:06,080
Speaker 4: And I thought about it afterwards, I thought, why is that?

1162
00:57:06,200 --> 00:57:08,400
Speaker 4: You know, well, maybe some of it was true, but

1163
00:57:08,400 --> 00:57:10,520
Speaker 4: probably most of it just couldn't be verified. You know.

1164
00:57:10,560 --> 00:57:12,080
Speaker 4: It was like one guy said they saw something, they

1165
00:57:12,080 --> 00:57:13,799
Speaker 4: couldn't find someone else who saw the same thing, They

1166
00:57:13,800 --> 00:57:17,520
Speaker 4: couldn't find the underlying you know, whatever it was that

1167
00:57:17,640 --> 00:57:20,120
Speaker 4: was supposedly the source of the evidence. But nonetheless the

1168
00:57:20,160 --> 00:57:24,160
Speaker 4: story that emerged, although a little more vague and abstract

1169
00:57:24,240 --> 00:57:26,480
Speaker 4: with less detail, was probably much more likely to be

1170
00:57:26,520 --> 00:57:30,200
Speaker 4: true because it was sort of filtered through agents that

1171
00:57:30,280 --> 00:57:33,280
Speaker 4: bother to check and try to only pass on verifiable information.

1172
00:57:33,360 --> 00:57:35,080
Speaker 4: And they are now faced with the problem of how

1173
00:57:35,120 --> 00:57:37,320
Speaker 4: do you tell whether this video of Trump doing X

1174
00:57:37,720 --> 00:57:40,480
Speaker 4: is actually Trump doing it or some fake that someone created, right,

1175
00:57:40,520 --> 00:57:42,800
Speaker 4: But I don't know who else to put more trust in,

1176
00:57:43,200 --> 00:57:46,840
Speaker 4: you know, for sorting that out than journalists and or

1177
00:57:46,880 --> 00:57:48,760
Speaker 4: and there are some organizations that they work with who

1178
00:57:48,760 --> 00:57:51,200
Speaker 4: are experts at detecting these kinds of things and so on.

1179
00:57:51,720 --> 00:57:55,440
Speaker 4: So I think maybe it might paradoxically increase the value

1180
00:57:55,680 --> 00:57:59,120
Speaker 4: and the attention paid to more legitimate sources, which I

1181
00:57:59,120 --> 00:58:00,800
Speaker 4: think would probably be a good thing on balance.

1182
00:58:01,200 --> 00:58:03,240
Speaker 5: I mean, the pessimistic view is that these things get

1183
00:58:03,280 --> 00:58:05,600
Speaker 5: increased in scale, right, it makes it much easier to

1184
00:58:06,520 --> 00:58:10,080
Speaker 5: scam at large scale and make it sound plausible, right.

1185
00:58:10,200 --> 00:58:14,000
Speaker 5: But the optimistic take is exactly what Chris was saying,

1186
00:58:14,000 --> 00:58:17,640
Speaker 5: that once we realize that these things are possible at scale,

1187
00:58:18,320 --> 00:58:20,800
Speaker 5: maybe we start being more skeptical of most of the

1188
00:58:20,880 --> 00:58:24,280
Speaker 5: sort of rapid information that we get, and we withhold

1189
00:58:24,320 --> 00:58:26,600
Speaker 5: judgment just a little bit longer until we can have

1190
00:58:26,640 --> 00:58:29,160
Speaker 5: some verified sources. And the idea would be if we

1191
00:58:29,160 --> 00:58:32,240
Speaker 5: could actually have verified sources again, we haven't had that

1192
00:58:32,280 --> 00:58:34,440
Speaker 5: for a while. Now that anybody can start up a

1193
00:58:34,520 --> 00:58:35,919
Speaker 5: cable network and say whatever they want.

1194
00:58:36,560 --> 00:58:39,040
Speaker 1: This is something I've been wondering about recently, is all

1195
00:58:39,040 --> 00:58:41,600
Speaker 1: of these things that we're very concerned about, like deep fakes,

1196
00:58:42,320 --> 00:58:46,120
Speaker 1: will the younger generations be much less susceptible to them

1197
00:58:46,160 --> 00:58:48,400
Speaker 1: because they're well aware that if you see a video

1198
00:58:48,440 --> 00:58:50,200
Speaker 1: of something, it might be real, it might be fake,

1199
00:58:50,640 --> 00:58:53,480
Speaker 1: as opposed to know, those of us who are older

1200
00:58:53,680 --> 00:58:56,080
Speaker 1: are really concerned about it in a way that we

1201
00:58:56,160 --> 00:58:56,960
Speaker 1: might not need to be.

1202
00:58:57,720 --> 00:58:59,080
Speaker 3: I think that's a really good question.

1203
00:58:59,160 --> 00:59:01,320
Speaker 4: I think the jury still out on that, because I

1204
00:59:01,360 --> 00:59:04,560
Speaker 4: think in some ways younger people are a bit more

1205
00:59:04,680 --> 00:59:07,720
Speaker 4: naive about some things. They don't have certain experiences and

1206
00:59:07,720 --> 00:59:10,240
Speaker 4: so on. On the other hand, as you say, they

1207
00:59:10,280 --> 00:59:12,360
Speaker 4: may be more used to the idea that videos are

1208
00:59:12,360 --> 00:59:15,440
Speaker 4: not proof the way people who grew up in an

1209
00:59:15,480 --> 00:59:19,080
Speaker 4: era of less video and less awareness of video editing

1210
00:59:19,120 --> 00:59:21,640
Speaker 4: and so on might not be. I'm reminded of the

1211
00:59:21,680 --> 00:59:24,960
Speaker 4: sort of discussion that you heard, you know, fifteen to

1212
00:59:25,000 --> 00:59:27,520
Speaker 4: twenty years ago about the so called digital natives and

1213
00:59:27,560 --> 00:59:29,840
Speaker 4: how having grown up with technology, they were so smart

1214
00:59:29,840 --> 00:59:30,680
Speaker 4: in using it and so on.

1215
00:59:30,760 --> 00:59:32,520
Speaker 3: And then when I became a college.

1216
00:59:32,200 --> 00:59:34,240
Speaker 4: Professor, I found out that students didn't really know how

1217
00:59:34,240 --> 00:59:36,600
Speaker 4: to do a proper Google search, you know, and so on,

1218
00:59:36,640 --> 00:59:39,880
Speaker 4: even though they were supposedly natives, like it's not in America,

1219
00:59:40,000 --> 00:59:43,440
Speaker 4: not being able to, you know, to speak English correctly,

1220
00:59:43,720 --> 00:59:47,600
Speaker 4: So that gives me less optimism. But I think in general,

1221
00:59:47,760 --> 00:59:50,240
Speaker 4: across generations. I think there's going to be a rise

1222
00:59:50,480 --> 00:59:54,280
Speaker 4: in a rise in skepticism, may be somewhat of a

1223
00:59:54,320 --> 00:59:57,760
Speaker 4: decline of truth bias. Truth bias can't decline too far

1224
00:59:57,840 --> 01:00:00,439
Speaker 4: otherwise we just can't interact with anybody anymore. But maybe

1225
01:00:00,480 --> 01:00:03,240
Speaker 4: a sort of a decline or a specialization of truth

1226
01:00:03,280 --> 01:00:04,960
Speaker 4: bias where you have sort of a little bit more

1227
01:00:04,960 --> 01:00:07,520
Speaker 4: truth bias in some areas, like when you're talking to

1228
01:00:07,560 --> 01:00:09,800
Speaker 4: an actual human being standing in front of you, and

1229
01:00:09,960 --> 01:00:11,960
Speaker 4: less when you're watching a video on TikTok. Like that

1230
01:00:12,000 --> 01:00:14,480
Speaker 4: would be a nice balance to have, right and not

1231
01:00:14,560 --> 01:00:16,480
Speaker 4: to pick on TikTok, but there seems to be more

1232
01:00:16,480 --> 01:00:19,200
Speaker 4: nonsense there than most other places, just from what I've noticed.

1233
01:00:20,120 --> 01:00:25,160
Speaker 1: Okay, so zooming out, give us some practical advice for people,

1234
01:00:25,720 --> 01:00:29,520
Speaker 1: some tips they can take home. Well.

1235
01:00:29,600 --> 01:00:32,000
Speaker 5: I'd say one quick one is that whenever you're in

1236
01:00:32,040 --> 01:00:35,760
Speaker 5: a situation where the consequences could be big, be willing

1237
01:00:35,800 --> 01:00:38,360
Speaker 5: to ask more questions. And it can be socially awkward

1238
01:00:38,400 --> 01:00:40,800
Speaker 5: to do that, right, to kind of press for more information,

1239
01:00:41,520 --> 01:00:44,720
Speaker 5: but doing that's essential if the consequences of being deceived

1240
01:00:44,760 --> 01:00:47,880
Speaker 5: are big, and sometimes you can kind of get started

1241
01:00:47,920 --> 01:00:51,200
Speaker 5: on asking questions without actually, you know, being hostile and aggressive,

1242
01:00:51,440 --> 01:00:53,520
Speaker 5: like can you tell me more? Is a way of

1243
01:00:53,520 --> 01:00:55,080
Speaker 5: getting somebody to talk a little bit more, give you

1244
01:00:55,120 --> 01:00:57,360
Speaker 5: a little more information that might actually make it more

1245
01:00:57,400 --> 01:01:01,120
Speaker 5: comfortable to ask questions about that more information they give you. Right,

1246
01:01:01,240 --> 01:01:04,600
Speaker 5: So the sorts of skills that many of us develop

1247
01:01:04,600 --> 01:01:06,520
Speaker 5: an academia. Or you're giving a talk and you can

1248
01:01:06,920 --> 01:01:09,600
Speaker 5: stand up and ask the hostile question, or you could

1249
01:01:09,960 --> 01:01:13,280
Speaker 5: ask a question that reveals more information, and the goal

1250
01:01:13,320 --> 01:01:15,360
Speaker 5: is to try and reveal more information and remain a

1251
01:01:15,360 --> 01:01:19,320
Speaker 5: little uncertain until you have that information. One broader one

1252
01:01:19,360 --> 01:01:21,920
Speaker 5: is if somebody were trying to scam me in this situation,

1253
01:01:22,000 --> 01:01:24,720
Speaker 5: Let's say you're investing in something. If somebody were trying

1254
01:01:24,760 --> 01:01:26,000
Speaker 5: to scam me, how would I know?

1255
01:01:26,720 --> 01:01:26,840
Speaker 1: Right?

1256
01:01:26,920 --> 01:01:29,000
Speaker 5: So, if I'm thinking about investing in crypto and say,

1257
01:01:29,040 --> 01:01:30,760
Speaker 5: is that a scam? How would I know if that's

1258
01:01:30,760 --> 01:01:34,080
Speaker 5: a scam. If you can't answer that question, then you

1259
01:01:34,080 --> 01:01:35,080
Speaker 5: probably should walk away.

1260
01:01:35,960 --> 01:01:36,520
Speaker 3: So if you.

1261
01:01:36,440 --> 01:01:40,840
Speaker 5: Don't understand how blockchain works and how crypto coins work,

1262
01:01:41,000 --> 01:01:43,400
Speaker 5: you probably shouldn't be investing in crypto. Regardless of what

1263
01:01:43,440 --> 01:01:46,200
Speaker 5: a celebrity tells you. If it were a scam, how

1264
01:01:46,240 --> 01:01:49,080
Speaker 5: would you tell well, and be really hard to tell

1265
01:01:49,120 --> 01:01:52,000
Speaker 5: if you don't understand how it works intimately. I'll give

1266
01:01:52,040 --> 01:01:55,880
Speaker 5: two practical ideas also. I think one is don't make

1267
01:01:55,920 --> 01:01:59,520
Speaker 5: really important decisions all by yourself. We came across many

1268
01:01:59,560 --> 01:02:03,400
Speaker 5: example where people were about to make mistakes, big mistakes, like,

1269
01:02:03,440 --> 01:02:05,320
Speaker 5: for example, one of those guys was about to give

1270
01:02:06,240 --> 01:02:08,960
Speaker 5: to wire money to the fake French defense minister, and

1271
01:02:09,120 --> 01:02:11,200
Speaker 5: his friend walked into the room where he was having

1272
01:02:11,440 --> 01:02:13,560
Speaker 5: this call with him, and he immediately right away said

1273
01:02:13,600 --> 01:02:15,600
Speaker 5: this can't be real. This must be a scam. And

1274
01:02:15,640 --> 01:02:17,920
Speaker 5: why was the friend able to notice but the victim,

1275
01:02:18,200 --> 01:02:21,360
Speaker 5: the intended victim wasn't. Well, probably the friend had not

1276
01:02:21,400 --> 01:02:23,840
Speaker 5: been in on all the previous conversations, so there wasn't

1277
01:02:23,880 --> 01:02:26,920
Speaker 5: sort of that sunk costs idea, that idea of a relationship.

1278
01:02:26,360 --> 01:02:26,840
Speaker 3: And so on.

1279
01:02:27,320 --> 01:02:29,200
Speaker 4: And maybe it was just he had a different mindset

1280
01:02:29,200 --> 01:02:31,040
Speaker 4: that day, He had a different attitude, he was thinking

1281
01:02:31,040 --> 01:02:33,640
Speaker 4: different things, and he never got sucked into the whole thing.

1282
01:02:33,680 --> 01:02:37,520
Speaker 4: So ask a friend, get an outside view before you

1283
01:02:37,560 --> 01:02:39,400
Speaker 4: make a big decision. Should I really send all my

1284
01:02:39,400 --> 01:02:42,080
Speaker 4: life savings to this guy, you know, just because everybody

1285
01:02:42,080 --> 01:02:44,200
Speaker 4: says he's the greatest thing, or is there any other

1286
01:02:44,320 --> 01:02:46,640
Speaker 4: consideration I should be using when investing my money.

1287
01:02:46,680 --> 01:02:47,439
Speaker 3: So that's one.

1288
01:02:47,960 --> 01:02:51,480
Speaker 4: The second one is like, do your work on deadlines,

1289
01:02:51,520 --> 01:02:53,640
Speaker 4: but don't like give away your money on deadline. So

1290
01:02:53,640 --> 01:02:55,680
Speaker 4: if anybody ever says, like, you know, you've got to

1291
01:02:55,680 --> 01:02:58,240
Speaker 4: do this within a certain period of time, the police

1292
01:02:58,240 --> 01:03:00,000
Speaker 4: are coming to your house if you don't like pay

1293
01:03:00,120 --> 01:03:04,720
Speaker 4: this bill right away, or this offers exploding very quickly,

1294
01:03:05,040 --> 01:03:07,240
Speaker 4: you know, or there's only one of these things left

1295
01:03:07,400 --> 01:03:10,800
Speaker 4: or whatever, just be aware that, like that's a prime

1296
01:03:10,960 --> 01:03:13,960
Speaker 4: environment to not have time to ask questions, not have

1297
01:03:13,960 --> 01:03:15,800
Speaker 4: time to think about the information you're missing, not go

1298
01:03:15,840 --> 01:03:18,040
Speaker 4: through any of this, and realize that, like, you can

1299
01:03:18,080 --> 01:03:20,200
Speaker 4: still buy that thing the next day if you really

1300
01:03:20,240 --> 01:03:22,000
Speaker 4: want it, you can still invest your money next week

1301
01:03:22,040 --> 01:03:23,120
Speaker 4: after you have checked the guy out.

1302
01:03:23,560 --> 01:03:26,200
Speaker 3: You're not going to lose much. I would go with

1303
01:03:26,240 --> 01:03:26,640
Speaker 3: those two.

1304
01:03:31,520 --> 01:03:34,920
Speaker 1: So that was Dan Simon's and Christopher Shabri. Now what

1305
01:03:35,000 --> 01:03:38,280
Speaker 1: we learn from them is that a lot of protecting

1306
01:03:38,280 --> 01:03:42,640
Speaker 1: ourselves against deception is about taking the tools of science,

1307
01:03:42,680 --> 01:03:46,440
Speaker 1: which is nothing but the tools of thinking clearly, and

1308
01:03:46,520 --> 01:03:49,840
Speaker 1: applying those to our daily lives. So, for example, if

1309
01:03:49,880 --> 01:03:54,440
Speaker 1: somebody says something is true, whether from their position of

1310
01:03:54,480 --> 01:03:58,360
Speaker 1: authority or religious status, or with a trust me on

1311
01:03:58,400 --> 01:04:03,480
Speaker 1: this one vibe. The key is to trust, but verify.

1312
01:04:03,600 --> 01:04:06,200
Speaker 1: The important thing to get in the habit of is

1313
01:04:06,280 --> 01:04:10,680
Speaker 1: just asking the next question. And it's tough because life

1314
01:04:10,720 --> 01:04:14,560
Speaker 1: doesn't allow questioning everything. Our schedules just don't allow that,

1315
01:04:15,080 --> 01:04:17,920
Speaker 1: and we have to operate on trust for most of

1316
01:04:17,960 --> 01:04:20,480
Speaker 1: what we do. And sometimes we find ourselves in a

1317
01:04:20,520 --> 01:04:23,760
Speaker 1: situation where someone doesn't quite answer the question we've asked,

1318
01:04:24,080 --> 01:04:27,200
Speaker 1: and it feels impolite to keep pressing on it. And

1319
01:04:27,320 --> 01:04:30,760
Speaker 1: also what is life if we don't trust? But the

1320
01:04:30,880 --> 01:04:34,480
Speaker 1: fact is we can always get a little smarter, a

1321
01:04:34,520 --> 01:04:39,040
Speaker 1: little less gullible by knowing that reality can be different

1322
01:04:39,080 --> 01:04:42,760
Speaker 1: in different heads. And whether we're talking about Tanya my

1323
01:04:42,840 --> 01:04:46,360
Speaker 1: fellow graduate student, or the dealer of the shell Game,

1324
01:04:46,880 --> 01:04:52,240
Speaker 1: or Elizabeth Holmes and Farahos or whatever, it's incredibly useful

1325
01:04:52,720 --> 01:04:57,960
Speaker 1: to stretch beyond the parochial limits of our mental models

1326
01:04:58,040 --> 01:05:02,000
Speaker 1: of the world, because with with more knowledge comes a

1327
01:05:02,080 --> 01:05:06,960
Speaker 1: bit more immunity, and understanding the character of our brains

1328
01:05:07,040 --> 01:05:10,240
Speaker 1: allows us to move through the world a little bit

1329
01:05:10,280 --> 01:05:20,920
Speaker 1: more smoothly than we would without that knowledge. Go to

1330
01:05:20,960 --> 01:05:24,360
Speaker 1: Eagleman dot com slash podcast. For more information and to

1331
01:05:24,400 --> 01:05:28,680
Speaker 1: find further reading, send me an email at podcasts at

1332
01:05:28,680 --> 01:05:32,520
Speaker 1: eagleman dot com with questions or discussion, and I'll address

1333
01:05:32,600 --> 01:05:40,040
Speaker 1: those in a special episode. Until next time, I'm David Eagleman,

1334
01:05:40,240 --> 01:05:42,160
Speaker 1: and this is Inner Cosmos.