1
00:00:00,000 --> 00:00:02,340
Hey there, ready for a deep dive into something

2
00:00:02,340 --> 00:00:03,340
really fascinating?

3
00:00:03,340 --> 00:00:04,480
Always up for that.

4
00:00:04,480 --> 00:00:07,700
Today we're gonna be talking about artificial intelligence

5
00:00:07,700 --> 00:00:10,980
and how it's like shaking up the world of behavioral science.

6
00:00:10,980 --> 00:00:13,900
Yeah, it's definitely moving faster than,

7
00:00:13,900 --> 00:00:15,880
well, it's moving fast, it's crucial

8
00:00:15,880 --> 00:00:16,900
we understand what's happening.

9
00:00:16,900 --> 00:00:17,740
It really is.

10
00:00:17,740 --> 00:00:19,420
We've got some seriously interesting research

11
00:00:19,420 --> 00:00:22,600
and insights here, all from a panel discussion

12
00:00:22,600 --> 00:00:24,200
at the London School of Economics.

13
00:00:24,200 --> 00:00:26,760
Oh wow, the LSE, they always have their finger on the pulse.

14
00:00:26,760 --> 00:00:29,980
Exactly, so we're talking about some serious heavy hitters

15
00:00:29,980 --> 00:00:33,880
in both AI and behavioral science researchers,

16
00:00:33,880 --> 00:00:35,800
folks applying this stuff in the real world.

17
00:00:35,800 --> 00:00:38,460
Okay, yeah, so what are some of the big takeaways?

18
00:00:38,460 --> 00:00:39,920
Well, one of the things that really jumped out

19
00:00:39,920 --> 00:00:42,960
was this idea of AI being used to create,

20
00:00:42,960 --> 00:00:46,240
get this, synthetic research participants.

21
00:00:46,240 --> 00:00:47,860
Whoa, hold on, synthetic?

22
00:00:47,860 --> 00:00:50,240
Like, artificial people?

23
00:00:50,240 --> 00:00:51,700
You got it, instead of recruiting

24
00:00:51,700 --> 00:00:53,880
hundreds of people for a study,

25
00:00:53,880 --> 00:00:56,160
researchers can now use AI to create

26
00:00:56,160 --> 00:00:58,100
like digital versions of people.

27
00:00:58,100 --> 00:01:00,040
They can give them specific traits, backgrounds,

28
00:01:00,040 --> 00:01:02,640
all sorts of stuff, and then basically simulate

29
00:01:02,640 --> 00:01:04,480
how different folks might respond

30
00:01:04,480 --> 00:01:06,320
in like a research setting.

31
00:01:06,320 --> 00:01:08,780
So you're saying instead of finding people to study,

32
00:01:08,780 --> 00:01:11,860
they can just create them, that's wild.

33
00:01:11,860 --> 00:01:14,000
It really is, and it could totally revolutionize

34
00:01:14,000 --> 00:01:15,200
how research is done.

35
00:01:15,200 --> 00:01:18,280
Think about it, testing theories and interventions

36
00:01:18,280 --> 00:01:21,540
way faster without the headaches of traditional studies.

37
00:01:21,540 --> 00:01:24,880
Yeah, I see the appeal, but how accurate are these AI people?

38
00:01:24,880 --> 00:01:27,840
Can they really be that realistic, that complex?

39
00:01:27,840 --> 00:01:29,680
That's the million dollar question, right?

40
00:01:29,680 --> 00:01:30,880
The tech is still pretty new,

41
00:01:30,880 --> 00:01:32,100
but it's getting better all the time.

42
00:01:32,100 --> 00:01:34,300
Like, how new are we talking?

43
00:01:34,300 --> 00:01:36,940
I mean, researchers are developing these

44
00:01:36,940 --> 00:01:39,940
really sophisticated algorithms that factor in everything,

45
00:01:39,940 --> 00:01:42,240
you know, demographics, personality traits.

46
00:01:42,240 --> 00:01:44,860
They're getting surprisingly good at mimicking us.

47
00:01:44,860 --> 00:01:46,660
Okay, that's kind of unsettling, isn't it?

48
00:01:46,660 --> 00:01:50,500
I mean, it's like AI's reading between the lines of our lives.

49
00:01:50,500 --> 00:01:54,900
It is a bit like that, but think about the good it could do.

50
00:01:54,900 --> 00:01:57,740
Imagine using that info to design better treatments

51
00:01:57,740 --> 00:02:01,260
for things like addiction, procrastination, even prejudice.

52
00:02:01,260 --> 00:02:03,580
Okay, yeah, I see the potential there,

53
00:02:03,580 --> 00:02:05,960
but maybe this is just me.

54
00:02:05,960 --> 00:02:08,040
What happens when they get too good?

55
00:02:08,040 --> 00:02:09,000
Where's the line?

56
00:02:09,000 --> 00:02:10,800
Right, and that's something the experts at LSE

57
00:02:10,800 --> 00:02:12,380
were really grappling with.

58
00:02:12,380 --> 00:02:15,720
On one hand, this could minimize harm to real people,

59
00:02:15,720 --> 00:02:17,800
you know, even open doors to research that

60
00:02:17,800 --> 00:02:19,260
would be impossible otherwise.

61
00:02:19,260 --> 00:02:20,420
I see that for sure.

62
00:02:20,420 --> 00:02:21,720
But on the other hand,

63
00:02:21,720 --> 00:02:24,540
you've got concerns about privacy, consent.

64
00:02:24,540 --> 00:02:28,260
The potential for misuse is huge.

65
00:02:28,260 --> 00:02:29,620
Yeah, it's a lot to unpack.

66
00:02:29,620 --> 00:02:30,900
What else did they have to say?

67
00:02:30,900 --> 00:02:32,620
Well, one of the things that I found super interesting

68
00:02:32,620 --> 00:02:34,660
was just how many potential applications

69
00:02:34,660 --> 00:02:36,260
there are for AI in this field.

70
00:02:36,260 --> 00:02:37,460
Okay, so we're not just talking about

71
00:02:37,460 --> 00:02:39,140
synthetic research subjects then.

72
00:02:39,140 --> 00:02:40,060
What else is out there?

73
00:02:40,060 --> 00:02:41,220
Think of it this way.

74
00:02:41,220 --> 00:02:44,420
Imagine an AI that can go through mountains of data

75
00:02:44,420 --> 00:02:47,380
and pick out these subtle patterns in human behavior,

76
00:02:47,380 --> 00:02:50,580
patterns that we, as humans, might completely miss.

77
00:02:50,580 --> 00:02:52,420
Hmm, okay, I'm listening.

78
00:02:52,420 --> 00:02:56,100
Like, imagine AI that can actually help us understand

79
00:02:56,100 --> 00:02:59,000
why people do the things they do, you know,

80
00:02:59,000 --> 00:03:00,180
in a much deeper way.

81
00:03:00,180 --> 00:03:02,060
Yeah, that's powerful stuff.

82
00:03:02,060 --> 00:03:03,620
And then what?

83
00:03:03,620 --> 00:03:05,380
What do we do with that information?

84
00:03:05,380 --> 00:03:07,920
Well, one thing we talked about was using it to design,

85
00:03:07,920 --> 00:03:10,260
like, way more effective interventions for stuff.

86
00:03:10,260 --> 00:03:11,780
We've touched on some of those already.

87
00:03:11,780 --> 00:03:13,820
Right, right, like addiction, procrastination.

88
00:03:13,820 --> 00:03:17,540
Exactly, or even like helping people make healthier choices,

89
00:03:17,540 --> 00:03:19,860
encouraging them to save more money.

90
00:03:19,860 --> 00:03:22,080
The possibilities are huge, really.

91
00:03:22,080 --> 00:03:24,760
Okay, I'm starting to see the bigger picture here.

92
00:03:24,760 --> 00:03:27,000
So we've got these synthetic research participants,

93
00:03:27,000 --> 00:03:30,080
we've got AI analyzing our every move.

94
00:03:30,080 --> 00:03:33,180
Is there anything AI can't do in this field?

95
00:03:33,180 --> 00:03:34,180
Well, it's funny you should ask that

96
00:03:34,180 --> 00:03:36,740
because one of the experts at LSE,

97
00:03:36,740 --> 00:03:38,100
I think her name was Susan Michie,

98
00:03:38,100 --> 00:03:39,720
she made a really interesting point.

99
00:03:39,720 --> 00:03:40,560
What was that?

100
00:03:40,560 --> 00:03:43,040
She basically said, yeah, AI has crazy potential,

101
00:03:43,040 --> 00:03:46,300
but we can't, like, put all our eggs in one basket, you know?

102
00:03:46,300 --> 00:03:48,180
Oh, interesting, so it's not a magic bullet.

103
00:03:48,180 --> 00:03:49,420
Exactly.

104
00:03:49,420 --> 00:03:50,900
She actually used this analogy.

105
00:03:50,900 --> 00:03:53,180
She said it's like a carpenter who's so psyched

106
00:03:53,180 --> 00:03:55,760
about his new power saw that he forgets

107
00:03:55,760 --> 00:03:57,700
about all the other tools he has.

108
00:03:57,700 --> 00:03:58,660
Okay, I get it.

109
00:03:58,660 --> 00:04:02,740
We can't let AI blind us to other important aspects

110
00:04:02,740 --> 00:04:04,500
of, you know, being human.

111
00:04:04,500 --> 00:04:06,480
Right, like we still don't fully understand

112
00:04:06,480 --> 00:04:09,860
how people make those big life-altering changes, right?

113
00:04:09,860 --> 00:04:11,540
The kind of stuff we were talking about before,

114
00:04:11,540 --> 00:04:14,180
like quitting smoking or going vegan

115
00:04:14,180 --> 00:04:17,100
or deciding to live a more minimalist lifestyle.

116
00:04:17,100 --> 00:04:20,220
Which, by the way, are exactly the kinds of changes we need

117
00:04:20,220 --> 00:04:23,220
if we're gonna tackle huge issues like climate change.

118
00:04:23,220 --> 00:04:25,260
It's like, we're so busy trying to predict

119
00:04:25,260 --> 00:04:26,720
what people will click on next

120
00:04:26,720 --> 00:04:28,260
that we forget about the bigger picture.

121
00:04:28,260 --> 00:04:30,260
100%, and speaking of the bigger picture,

122
00:04:30,260 --> 00:04:34,180
there's another crucial factor we gotta consider.

123
00:04:34,180 --> 00:04:35,000
Let's hear it.

124
00:04:35,000 --> 00:04:36,540
The environmental impact of all this.

125
00:04:36,540 --> 00:04:38,500
You know, all this amazing AI development.

126
00:04:38,500 --> 00:04:42,580
Oh, right, because like, training those massive AI models

127
00:04:42,580 --> 00:04:44,800
requires a ton of energy, right?

128
00:04:44,800 --> 00:04:47,580
Yeah, exactly, and it's something we,

129
00:04:47,580 --> 00:04:49,340
you know, it's something we can't ignore.

130
00:04:49,340 --> 00:04:51,540
AI might have the potential to help us fix

131
00:04:51,540 --> 00:04:53,320
some of the world's biggest problems,

132
00:04:53,320 --> 00:04:54,820
but we have to be smart about it,

133
00:04:54,820 --> 00:04:56,700
mindful of the resources it needs.

134
00:04:56,700 --> 00:04:58,680
Right, gotta make sure the cure isn't worse

135
00:04:58,680 --> 00:04:59,520
than the disease.

136
00:04:59,520 --> 00:05:00,360
Yeah!

137
00:05:00,360 --> 00:05:02,180
It's a lot to wrap your head around, isn't it?

138
00:05:02,180 --> 00:05:04,860
AI could be the key to like,

139
00:05:04,860 --> 00:05:06,340
unlocking all these amazing things,

140
00:05:06,340 --> 00:05:07,820
but it could also make things a lot worse

141
00:05:07,820 --> 00:05:09,160
if we're not careful.

142
00:05:09,160 --> 00:05:10,980
It really is a tightrope walk.

143
00:05:10,980 --> 00:05:14,420
The future of AI, it's not set in stone, you know?

144
00:05:14,420 --> 00:05:18,320
It's something we're all like, actively creating.

145
00:05:18,320 --> 00:05:20,140
Yeah, with every choice we make.

146
00:05:20,140 --> 00:05:21,440
So where do we go from here?

147
00:05:21,440 --> 00:05:25,520
How do we ensure that AI is actually used for good?

148
00:05:25,520 --> 00:05:26,620
That it doesn't become, you know,

149
00:05:26,620 --> 00:05:29,100
just another tool for manipulating people

150
00:05:29,100 --> 00:05:30,780
or like, destroying the planet?

151
00:05:30,780 --> 00:05:32,580
Well, there are a lot of ideas thrown around,

152
00:05:32,580 --> 00:05:33,620
but one that stuck with me,

153
00:05:33,620 --> 00:05:36,060
I think it was from Oliver Hauser.

154
00:05:36,060 --> 00:05:38,220
He said, we need to stop thinking about AI

155
00:05:38,220 --> 00:05:40,340
as some kind of like, magical solution.

156
00:05:40,340 --> 00:05:41,420
Oh, interesting.

157
00:05:41,420 --> 00:05:43,140
What should we be thinking of it as?

158
00:05:43,140 --> 00:05:45,660
He said we should think of it more like a new medication.

159
00:05:45,660 --> 00:05:48,140
Like, it needs to go through rigorous testing

160
00:05:48,140 --> 00:05:49,620
before we unleash it on the world.

161
00:05:49,620 --> 00:05:50,780
Yeah, exactly.

162
00:05:50,780 --> 00:05:53,000
We need to do the research, understand the benefits,

163
00:05:53,000 --> 00:05:55,580
the risks, you know, the whole nine yards

164
00:05:55,580 --> 00:05:57,860
before we fully incorporate it into our lives.

165
00:05:57,860 --> 00:05:58,680
Yeah, that makes sense.

166
00:05:58,680 --> 00:06:00,420
We wouldn't just start mass producing some new drug

167
00:06:00,420 --> 00:06:02,900
without understanding the side effects, right?

168
00:06:02,900 --> 00:06:03,740
Right, exactly.

169
00:06:03,740 --> 00:06:06,440
And we need clear guidelines, regulations, the whole thing,

170
00:06:06,440 --> 00:06:09,460
just like we do with any other powerful tool, really.

171
00:06:09,460 --> 00:06:11,520
So it's not just about the technology.

172
00:06:11,520 --> 00:06:12,740
It's about us, right?

173
00:06:12,740 --> 00:06:15,620
Like humans and how we choose to use it.

174
00:06:15,620 --> 00:06:18,020
We need to be careful, think ahead,

175
00:06:18,020 --> 00:06:19,060
all that good stuff.

176
00:06:19,060 --> 00:06:21,260
Absolutely, and that's where behavioral science comes in.

177
00:06:21,260 --> 00:06:22,820
It's like essential, really.

178
00:06:22,820 --> 00:06:23,660
How so?

179
00:06:23,660 --> 00:06:25,740
Because if we really understand human psychology,

180
00:06:25,740 --> 00:06:29,140
we can design these AI systems to be, you know,

181
00:06:29,140 --> 00:06:31,100
not just effective, but ethical,

182
00:06:31,100 --> 00:06:33,780
aligned with our values, not just some, you know,

183
00:06:33,780 --> 00:06:35,540
some scary out of control thing.

184
00:06:35,540 --> 00:06:37,120
Okay, yeah, that makes sense.

185
00:06:37,120 --> 00:06:38,580
So it's not AI versus humans.

186
00:06:38,580 --> 00:06:40,780
It's AI working with humans, right,

187
00:06:40,780 --> 00:06:43,200
to like achieve some greater good.

188
00:06:43,200 --> 00:06:44,620
Exactly, it's a team effort,

189
00:06:44,620 --> 00:06:48,060
and it's gonna take, you know, ongoing dialogue,

190
00:06:48,060 --> 00:06:50,220
careful consideration, all of that.

191
00:06:50,220 --> 00:06:52,860
It's like we're building this crazy powerful tool, right?

192
00:06:52,860 --> 00:06:54,860
And we need to make sure we have the instruction manual

193
00:06:54,860 --> 00:06:55,700
to go with it.

194
00:06:55,700 --> 00:06:57,360
Totally, and part of that manual, I think,

195
00:06:57,360 --> 00:07:00,240
is understanding how AI could be, you know,

196
00:07:00,240 --> 00:07:03,620
designed to kind of play us, like tap into our weaknesses.

197
00:07:03,620 --> 00:07:06,060
Right, like our tendency to trust a friendly face

198
00:07:06,060 --> 00:07:08,100
or follow the crowd, all that good stuff.

199
00:07:08,100 --> 00:07:10,100
Exactly, we have to be so aware of that,

200
00:07:10,100 --> 00:07:11,700
especially as AI gets, you know,

201
00:07:11,700 --> 00:07:13,820
more and more integrated into our lives.

202
00:07:13,820 --> 00:07:16,140
Okay, so how do we do that?

203
00:07:16,140 --> 00:07:18,540
How do we make sure it's a force for good,

204
00:07:18,540 --> 00:07:22,100
not some slippery slope to, like, I don't know, dystopia?

205
00:07:22,100 --> 00:07:24,180
Well, for one, we can design these systems

206
00:07:24,180 --> 00:07:26,660
to like encourage critical thinking.

207
00:07:26,660 --> 00:07:29,420
Imagine an AI that not only tries to persuade you,

208
00:07:29,420 --> 00:07:31,620
but also prompts you to question stuff,

209
00:07:31,620 --> 00:07:32,820
look at different sides.

210
00:07:32,820 --> 00:07:33,660
I like that.

211
00:07:33,660 --> 00:07:35,500
So instead of just telling you what to think,

212
00:07:35,500 --> 00:07:37,100
it helps you think for yourself.

213
00:07:37,100 --> 00:07:40,100
Yes, and the experts at the LSE,

214
00:07:40,100 --> 00:07:42,340
they were really big on transparency.

215
00:07:42,340 --> 00:07:45,260
Like people deserve to know how these algorithms work,

216
00:07:45,260 --> 00:07:47,180
how they're shaping their choices.

217
00:07:47,180 --> 00:07:48,380
Right, but there's a balance, right?

218
00:07:48,380 --> 00:07:49,900
Too much information could be overwhelming.

219
00:07:49,900 --> 00:07:52,260
It's like, who actually reads those terms and conditions?

220
00:07:52,260 --> 00:07:53,740
Exactly, and that's where, you know,

221
00:07:53,740 --> 00:07:56,460
understanding human behavior comes in again.

222
00:07:56,460 --> 00:07:59,060
We can design these systems so they're transparent,

223
00:07:59,060 --> 00:08:00,660
but also user-friendly,

224
00:08:00,660 --> 00:08:03,500
no more hiding behind legalese, you know?

225
00:08:03,500 --> 00:08:04,620
I am all for that.

226
00:08:04,620 --> 00:08:06,660
This has been such a fascinating conversation,

227
00:08:06,660 --> 00:08:08,220
but honestly, I feel like we've only

228
00:08:08,220 --> 00:08:09,380
just scratched the surface here.

229
00:08:09,380 --> 00:08:11,840
Me too, the possibilities, the challenges,

230
00:08:11,840 --> 00:08:13,620
it's all still unfolding.

231
00:08:13,620 --> 00:08:15,540
But one thing's for sure,

232
00:08:15,540 --> 00:08:18,900
the future of AI is like totally bound up

233
00:08:18,900 --> 00:08:20,740
with how well we understand ourselves.

234
00:08:20,740 --> 00:08:22,780
And that's something we're all figuring out together

235
00:08:22,780 --> 00:08:24,380
every single day.

236
00:08:24,380 --> 00:08:25,400
So to our listeners,

237
00:08:25,400 --> 00:08:27,860
next time you see that perfectly targeted ad

238
00:08:27,860 --> 00:08:30,240
or scroll through your personalized newsfeed,

239
00:08:30,240 --> 00:08:32,820
remember there's probably some really sophisticated AI

240
00:08:32,820 --> 00:08:33,740
behind it.

241
00:08:33,740 --> 00:08:36,800
But also remember, we're not just along for the ride.

242
00:08:36,800 --> 00:08:40,020
We have a say in how this technology shapes our world.

243
00:08:40,020 --> 00:08:41,380
It's a two-way street.

244
00:08:41,380 --> 00:08:44,540
We're shaping AI and it's shaping us.

245
00:08:44,540 --> 00:08:46,140
It really makes you think.

246
00:08:46,140 --> 00:08:47,700
So we wanna hear from you.

247
00:08:47,700 --> 00:08:50,740
If you could ask an AI anything about human behavior,

248
00:08:50,740 --> 00:08:51,860
what would it be?

249
00:08:51,860 --> 00:08:53,700
Head over to our website, find us on social media,

250
00:08:53,700 --> 00:08:54,540
let us know.

251
00:08:54,540 --> 00:08:56,340
Until next time, keep exploring,

252
00:08:56,340 --> 00:09:12,340
keep asking questions and keep diving deep.