Latest / The You Project / #2223 The Art and Science of Certainty - Professor Adam Kucharski

Transcript

#2223 The Art and Science of Certainty - Professor Adam Kucharski · The You Project

0:00
Speaker 1: G'day, team. It's the You Project. What else would it
0:03
Speaker 1: be if I was hosting at Tiffany Ann Cook? Tiffany
0:06
Speaker 1: Ann Cook is over there at TYP Central looking a
0:10
Speaker 1: little bit sleepy because she's just eaten dinner at the
0:12
Speaker 1: ridiculous grandma hour of about 5.30. What are you eating
0:16
Speaker 1: dinner at 5.30? Are you 80 years old? What are
0:19
Speaker 1: you doing?
0:19
Speaker 2: I do like nana time. I'm going to be in bed.
0:22
Speaker 2: I'm going to be sliding into bed by 8, 8.30.
0:25
Speaker 2: I don't want any food in my tum-tum by then.
0:28
Speaker 1: Yeah, that's, yeah, that. Yeah, well, that kind of makes sense,
0:31
Speaker 1: but also 5.30, come on. You know, Mary, Adam, that's
0:35
Speaker 1: my mum, Mary was complaining before because dinner at the
0:39
Speaker 1: new place that she's at, the new care home, which
0:41
Speaker 1: is beautiful, and thank you very much, Yolambi, for looking
0:44
Speaker 1: after my gorgeous mum. But she's not a big fan
0:47
Speaker 1: of the 5 o'clock dinner every night.
0:50
Speaker 3: I'd be in heaven there.
0:52
Speaker 1: I'm like, that's basically a mid-afternoon snack. Anyway, yeah, we
0:58
Speaker 1: could get you a room. I think there's a few
0:59
Speaker 1: spare ones.
1:00
Speaker 3: Okay.
1:01
Speaker 1: I'll look into it. I'll do some research. Adam, welcome
1:04
Speaker 1: to the You Project. Apologies for that initial crap. How
1:08
Speaker 1: are you?
1:10
Speaker 3: I'm good. Thank you. Thanks for having me. And I sympathize.
1:13
Speaker 2: I've got two young children, so I know the feeling
1:16
Speaker 2: if it gets to 5.30 and you're like, should we
1:18
Speaker 2: just hit an.
1:18
Speaker 3: Economy of scale here and get dinner for everybody? Yeah.
1:21
Speaker 1: Yeah. Do you do a 5.30 dinner person very often?
1:25
Speaker 3: I'd rather not.
1:26
Speaker 2: Be, but I think sometimes it's like, yeah, do you
1:28
Speaker 2: really want to cook twice, clean up twice?
1:30
Speaker 1: It's just practical, right? So I only started, apologies, I
1:36
Speaker 1: only started looking at you and your work yesterday. I
1:40
Speaker 1: don't know that my brain is in a state at
1:43
Speaker 1: 6.12 p.m. on a Tuesday night to be able to
1:46
Speaker 1: handle you, and I mean that in the nicest possible way.
1:50
Speaker 1: Have you ever seen the Big Bang Theory?
1:51
Speaker 3: I have, yes.
1:55
Speaker 1: So you're Sheldon and I'm Penny. So let's just move
2:00
Speaker 1: forward with that understanding so you understand my level of understanding.
2:06
Speaker 1: There you go. But tell people what you do rather
2:10
Speaker 1: than me just read some bio. Tell people what you do,
2:14
Speaker 1: what you're about, and we'll dive in from there.
2:17
Speaker 3: Yeah, so... the Sheldon synopsis, I guess.
2:21
Speaker 2: So original backgrounds, applied maths, but I think like a
2:24
Speaker 2: lot of people, if you're kind of, I mean, a
2:27
Speaker 2: lot of maths is problem solving, it's finding patterns, it's
2:30
Speaker 2: understanding how things fit together. Increasingly wanted to solve real
2:33
Speaker 2: world problems. How do we get better data on what
2:37
Speaker 2: we care about? Worked more in health. So particular fields
2:42
Speaker 2: where we can kind of try and understand how the
2:44
Speaker 2: health issues spread in populations. And then as kind of
2:48
Speaker 2: going through research, going up to becoming a professor in
2:51
Speaker 2: that area.
2:51
Speaker 3: I think when you start working in health.
2:54
Speaker 2: And how problems spread and very much becomes a behavioral
3:00
Speaker 2: problem as well, because of course it's not just these
3:02
Speaker 2: isolated houses. It's actually, well, how do people perceive them?
3:05
Speaker 2: How do people behave around them? And so that's a
3:08
Speaker 2: lot of my work over the years is how do
3:10
Speaker 2: we set up studies.
3:11
Speaker 3: To understand health and behavior.
3:13
Speaker 2: And How do we get closer to a lot of
3:15
Speaker 2: the things we care about, which often we can't observe?
3:17
Speaker 2: You know, we can observe these, the maybe later outcomes
3:20
Speaker 2: of a problem, but we actually can't understand what's influencing this,
3:23
Speaker 2: what's causing it. And then alongside that, the academic research
3:27
Speaker 2: has done a lot of writing about this kind of
3:30
Speaker 2: broad area, I'd say. How do you make sense of
3:32
Speaker 2: this quite uncertain, noisy, hard to measure world that we
3:36
Speaker 2: have to deal with? particularly working with organisations, governments and
3:41
Speaker 2: others in terms of how does that inform decision-making?
3:44
Speaker 3: What's the level of evidence you need to actually take action?
3:47
Speaker 2: And I think your background is obviously a lot in sports,
3:51
Speaker 2: both in terms of policy, government side, sports teams, organisations.
3:55
Speaker 2: There's a lot of demand often for having to make
3:58
Speaker 2: sense of a situation where you don't have all the information.
4:01
Speaker 2: You're going to have to make a decision. You're going
4:02
Speaker 2: to have to take some risk. How do you balance that?
4:05
Speaker 2: And then more recently, a lot of the work I've
4:06
Speaker 2: been doing is moving into AI. But I think fundamentally,
4:11
Speaker 2: for me, it's how do we get a lot of
4:13
Speaker 2: these problems historically, we haven't been able to get a
4:16
Speaker 2: lot of it is, is more about the why and
4:18
Speaker 2: relying less on just these kind of easy measurements that
4:21
Speaker 2: don't maybe tell us that much, and getting much closer,
4:24
Speaker 2: like what's actually the source and the driver of what's
4:28
Speaker 2: really happening. So I think that's the kind of the
4:30
Speaker 2: common frustration driving through a lot of my work and
4:33
Speaker 2: kind of manifesting in different ways and different industries over
4:35
Speaker 2: the years.
4:37
Speaker 1: When I first saw your background, I'm like, what does
4:42
Speaker 1: a professor of mathematics do anyway? I mean, other than maths.
4:47
Speaker 1: And I was thinking, how does maths, other than the obvious,
4:55
Speaker 1: but the less obvious ways that maths spills into or
4:58
Speaker 1: connects with or relates to day-to-day life for us, and
5:02
Speaker 1: research in other areas and How does it intersect with
5:09
Speaker 1: day-to-day life for people?
5:13
Speaker 2: I think a lot of, certainly at the applied end,
5:16
Speaker 2: it's really trying to understand and effectively write down the
5:22
Speaker 2: rules that drive life. the processes we're caring about. I mean,
5:25
Speaker 2: the kind of very simple level, you know, if you've
5:28
Speaker 2: put some money in an investment and you want to
5:30
Speaker 2: know what's going to happen in the future, you're not
5:32
Speaker 2: just going to say, oh, it's just going to go
5:34
Speaker 2: up by a fixed amount. We know that you get
5:36
Speaker 2: compounding and you get some growth and you get growth
5:39
Speaker 2: upon growth. And that same principle, whether you're working with
5:42
Speaker 2: a disease epidemic, whether you're working on an online trend,
5:45
Speaker 2: a lot of those ideas translate over. There's a lot
5:47
Speaker 2: of work, for example, on things like spread of behavior,
5:49
Speaker 2: spread of crime, And if you just look at the
5:52
Speaker 2: data on the surface of it, kind of your mental
5:54
Speaker 2: model of what's happening might be quite distant from actually
5:57
Speaker 2: what's really going on and lead you to some incorrect conclusions.
6:00
Speaker 2: So a lot of the work we do is to
6:02
Speaker 2: try and get actually closer to what's genuinely happening so
6:06
Speaker 2: you can make a better, whether it's trying to predict
6:08
Speaker 2: growth based on interest rates.
6:11
Speaker 3: So you don't get hammered on a mortgage.
6:12
Speaker 2: You're not expecting, or whether some of the work we've
6:15
Speaker 2: done over the years, you know, you get an online
6:16
Speaker 2: trend that takes off and a lot of people treat
6:19
Speaker 2: it as like, this is just this random occurrence.
6:21
Speaker 3: You know, what do we do?
6:22
Speaker 2: I mean, a week or so ago, you had a
6:24
Speaker 2: lot of the coverage of some of the AI agents
6:26
Speaker 2: kind of running rampage in computer systems. And some of
6:29
Speaker 2: the things is, oh, this is just a random bad
6:31
Speaker 2: thing that's happened. But actually there's a lot of just
6:33
Speaker 2: underlying kind of predictable principles. And so I see a
6:36
Speaker 2: lot of what we do is, you know, Everyone sort
6:39
Speaker 2: of thought the planets was kind of a bit luck
6:41
Speaker 2: and just, oh, here's some stuff happening. And then we
6:43
Speaker 2: tried to map out, well, no, there's a bunch of rules.
6:46
Speaker 2: And then Newton came up with his rules. And then
6:48
Speaker 2: Einstein was like, oh, actually, that doesn't work for Mercury.
6:50
Speaker 2: So we're going to have to come up with some
6:51
Speaker 2: slightly better rules. I think we're now moving into territory
6:54
Speaker 2: where not everything, but there's certain features of behavior.
6:58
Speaker 3: There's certain features of health where.
7:01
Speaker 2: it's actually much more predictable than we would have thought
7:03
Speaker 2: it was say 50 years ago. And if you, if
7:05
Speaker 2: you understand the system crucially, that gives you a better
7:08
Speaker 2: sense of what you can do about it. And I
7:10
Speaker 2: think what we, I kind of view like anyone who
7:12
Speaker 2: has an opinion on how to change the world is
7:15
Speaker 2: basically running a little mathematical kind of thing in that
7:18
Speaker 2: they don't realize they're doing it because it's often quite
7:19
Speaker 2: top level. But what you're saying is if I cause this,
7:22
Speaker 2: if I change this, I think this is going to
7:23
Speaker 2: happen and this is going to happen and this is
7:25
Speaker 2: going to happen.
7:26
Speaker 4: Yes.
7:26
Speaker 2: And you're stringing together a bunch of assumptions, each with
7:29
Speaker 2: a bit of uncertainty in a very kind of just
7:32
Speaker 2: ad hoc way.
7:33
Speaker 3: We do it all the time.
7:33
Speaker 2: It's kind of how we navigate the world. But that's
7:36
Speaker 2: a lot of my job is to, can we actually
7:37
Speaker 2: write that down and can we work out, okay, that
7:39
Speaker 2: little assumption you're making, is that actually supported by data?
7:43
Speaker 3: Where's the gaps? Where do we need more information to
7:45
Speaker 3: know what might happen when we intervene?
7:49
Speaker 1: Yeah, that's fascinating because I feel like we're always doing
7:52
Speaker 1: in the moment calculations and it might not be a
7:55
Speaker 1: mathematical per se, but it might be if I do this,
7:58
Speaker 1: what's the outcome going to be? I think most likely
8:00
Speaker 1: it would be that outcome or it could be this
8:03
Speaker 1: or it could be that, but I want that. So
8:06
Speaker 1: what do I need? What do I need? How do
8:08
Speaker 1: I need to manipulate myself or this situation or these
8:11
Speaker 1: factors or variables to create the outcome that I want
8:15
Speaker 1: or to make it more likely than not? which is
8:18
Speaker 1: kind of an equation. It's kind of a problem solving,
8:21
Speaker 1: almost mathematical internal process as well.
8:26
Speaker 3: And I think a lot of.
8:27
Speaker 2: It is just, you're trying to move beyond, and humans
8:30
Speaker 2: are like really, really good at that. We're really good
8:33
Speaker 2: at pattern matching often what we, what we'd call kind
8:36
Speaker 2: of intuition or, you know, kind of experience in difficult situations.
8:41
Speaker 2: You see it a lot in industries that rely very
8:42
Speaker 2: heavily on performance and people who've got a lot of
8:45
Speaker 2: experience can just spot this is something we need to try,
8:48
Speaker 2: this works, this doesn't. But I think even if you
8:50
Speaker 2: look over the years, 30, 40 years ago in medicine,
8:54
Speaker 2: it was a bit too kind of vibes based that,
8:56
Speaker 2: you know, you'd have someone say, I think this patient,
8:59
Speaker 2: I reckon this will work.
9:01
Speaker 3: And actually you just didn't have that.
9:03
Speaker 2: Kind of baseline level of evidence of, okay, if you
9:07
Speaker 2: do this and it works, will it work next time?
9:09
Speaker 2: Or was that luck? Or was there another factor that
9:11
Speaker 2: you're not seeing kind of getting involved there? I think
9:13
Speaker 2: it's been certainly in health, a lot of, effort to
9:16
Speaker 2: try and just build up that a bit more systematically
9:19
Speaker 2: to say this isn't just you know we're going to
9:21
Speaker 2: try things as we go but actually can we say
9:23
Speaker 2: something a bit more consistently i mean again in a
9:26
Speaker 2: lot of performance strategies of not just oh i tried
9:29
Speaker 2: this in a team and i happen to get a
9:30
Speaker 2: good outcome i think as we all know from things
9:32
Speaker 2: that involve risk you can make good decisions and get
9:35
Speaker 2: bad outcomes or make bad decisions and you know luck
9:37
Speaker 2: out and get a good outcome And it's really just
9:40
Speaker 2: trying to put more in place to say, okay, you know,
9:43
Speaker 2: over season after season or over player after player kind
9:46
Speaker 2: of recovery, or, you know, whether it's dealing with a
9:49
Speaker 2: sort of policy threat you need to react to, how
9:51
Speaker 2: do we have something.
9:52
Speaker 3: Where we can just be more confident we're going to
9:54
Speaker 3: get what we we want out.
9:55
Speaker 1: Of it yeah uh i apologize to my listeners for
9:59
Speaker 1: what i'm about to say because i literally spoke about
10:02
Speaker 1: it yesterday on a podcast so for the next three
10:05
Speaker 1: or four minutes make yourself a cup of tea put
10:06
Speaker 1: your feet in the bean bag light a candle uh
10:09
Speaker 1: whatever so there's a i think you'll just find this
10:14
Speaker 1: interesting prof um So in Australia there's the AFL, which
10:18
Speaker 1: is the Australian Football League, which I'm sure you don't
10:21
Speaker 1: follow but you've heard of maybe?
10:23
Speaker 3: Yeah, I've heard.
10:24
Speaker 2: I've seen odd bits when I've tuned into the wrong
10:28
Speaker 2: channel and sort of very quickly struggled to follow.
10:31
Speaker 1: Yeah, yeah. What on earth is that? Correct. So what's
10:35
Speaker 1: been really interesting this year in the 2026 season is
10:40
Speaker 1: so there's a team. Anyway, the team's called Carlton. But
10:43
Speaker 1: the team had a coach who was the– the head
10:47
Speaker 1: coach and in the first nine games of the season,
10:52
Speaker 1: it's a 23 game season and then finals. So nine
10:56
Speaker 1: of the first 23, they lost eight games. So they
10:59
Speaker 1: won one, then they lost eight games. Uh, and then
11:04
Speaker 1: they replaced the coach at the end of game nine
11:09
Speaker 1: after yet another loss, they replaced the coach with, uh,
11:15
Speaker 1: a kind of support staff coach, an assistant coach. So
11:19
Speaker 1: he became the temporary head coach with no real plan
11:23
Speaker 1: for him to carry on post-season, right? Now, what was
11:27
Speaker 1: amazing was the same team that had gone 1-8, they
11:32
Speaker 1: then lost for the rest of the season. Now, they'd
11:34
Speaker 1: lost 8. They lost for the rest of the season 2.
11:37
Speaker 4: Oh, okay.
11:40
Speaker 3: Right?
11:41
Speaker 1: So of the remaining... 14 games. They won 11, they
11:48
Speaker 1: drew one and they lost two. And I was just thinking,
11:52
Speaker 1: I was trying to do the maths on that myself
11:54
Speaker 1: in terms of percentages, but I'm thinking like this as
11:58
Speaker 1: a statistical analysis of performance.
12:01
Speaker 3: Okay.
12:01
Speaker 1: They played nine, they lost eight. Then the same team
12:04
Speaker 1: with the same bodies, the same skill, the same resources
12:07
Speaker 1: playing the same Absolute same sport with the same rules
12:11
Speaker 1: and dudes with the same genetics, like there's no new players.
12:15
Speaker 1: They didn't have some magic formula. They then went on
12:18
Speaker 1: this long winning streak, which changed, obviously, the data, the outcomes,
12:23
Speaker 1: the statistics, the percentages for the whole season. And what
12:29
Speaker 1: isn't talked about much over here, which fascinates me, is why.
12:33
Speaker 1: Why was there, apart from the obvious performance, But what's
12:39
Speaker 1: the psychology and what's the mechanism that created that incredible...
12:45
Speaker 1: I don't know. What's one in nine? What percentage is that?
12:48
Speaker 1: You'll tell me off the top of your head. Is
12:49
Speaker 1: it like 89 or 91? So they were on a 91%
12:55
Speaker 1: bloody losing record, and then they went up to like.
13:00
Speaker 4: 84% winning record. And I'm thinking... That made me curious.
13:08
Speaker 4: But what else made me curious is nobody's talking about
13:12
Speaker 4: what was the mechanism. Because what didn't change was the players,
13:17
Speaker 4: the personnel, the talent, the genetics, the resources, the game,
13:21
Speaker 4: the rules.
13:22
Speaker 1: Nothing changed. But the same humans doing the same thing
13:26
Speaker 1: that they do every week were now, it's almost like
13:30
Speaker 1: it was hard to believe. Like, Statistical anomalies like that
13:36
Speaker 1: freak me out. I'm fascinated with how that happens.
13:40
Speaker 3: I think that's really nice.
13:42
Speaker 2: We see it a lot in football as well and
13:45
Speaker 2: other sports where it is just, as you said, the
13:47
Speaker 2: only thing you've changed is what feels like just kind
13:50
Speaker 2: of one piece and it's transformed this. I think my
13:53
Speaker 2: first book actually was a lot about risk and prediction
13:56
Speaker 2: and sport I think is a great example just because A,
13:58
Speaker 2: there's lots of data. you have a lot of structure
14:00
Speaker 2: to it as well. So unlike, you know, a policy or,
14:04
Speaker 2: you know, just how life works, there's so many variables
14:07
Speaker 2: in sport. You know, there is some structure to, there's
14:09
Speaker 2: a game, there's players, there's rules. But one of the
14:12
Speaker 2: things that's kind of really interesting is that a lot
14:14
Speaker 2: of the early analysis of sport was in very structured.
14:17
Speaker 2: It was in the baseball, cricket, you know, where it's
14:20
Speaker 2: essentially kind of one-on-one battles to some extent.
14:22
Speaker 4: Yes.
14:22
Speaker 2: But once you start getting into team dynamics, the amount
14:25
Speaker 2: of combinations of stuff that could happen just goes really
14:28
Speaker 2: much further up. I think there's some really nice results
14:32
Speaker 2: from actually study of ecological systems, like how does bits
14:36
Speaker 2: of nature and animal populations work. You can actually get
14:39
Speaker 2: very fragile transitions in those kinds of complex sets of interaction.
14:45
Speaker 2: A lot of it, I think we have a sense
14:47
Speaker 2: of it if we worked with a team over the year,
14:50
Speaker 2: that basically if you have quite a strong performing or
14:52
Speaker 2: strong mindset, team it can pull up weak links very
14:56
Speaker 2: easily but you often get these these really quite sensitive
14:59
Speaker 2: transitions and i think we've all been in a situation
15:01
Speaker 2: like that where you've got a team that's struggling for
15:05
Speaker 2: some reason and then suddenly the attempts to be strong
15:08
Speaker 2: get pulled down by the the kind of weakness of
15:10
Speaker 2: the situation and it's something we've seen um explains a
15:14
Speaker 2: lot of what's happened with like financial crises again similarly
15:16
Speaker 2: like what looks on the face of it similar little
15:18
Speaker 2: small change and and drops off um And I think
15:23
Speaker 2: it creates, again, it's trying to understand those underlying principles.
15:27
Speaker 2: It's so hard to explain in sport, but I think
15:29
Speaker 2: you're right. You can get those things that seem quite innocuous,
15:32
Speaker 2: but actually because you're dealing with this quite, in some cases,
15:36
Speaker 2: quite fragile system. Certainly if a team's losing eight games
15:38
Speaker 2: on a go, there's something that's quite fragile about what's
15:41
Speaker 2: going on there. But it might actually be that there's
15:44
Speaker 2: a much smaller change than perhaps is obvious. I mean,
15:46
Speaker 2: I think a lot of us intuitively would think, you know,
15:48
Speaker 2: if you're losing that many games, you just need 11 or...
15:50
Speaker 2: 11 players in the NFL?
15:52
Speaker 3: No, less than that.
15:54
Speaker 1: 18 on the field.
15:55
Speaker 2: Yeah, 18 on the field. So, yeah, it's not necessarily
16:00
Speaker 2: that you have to replace the whole team. And I
16:02
Speaker 2: think this is one of the things that.
16:05
Speaker 2: Working with sports teams a bit over the years and
16:08
Speaker 2: writing about it is really interesting is the measurement problem
16:10
Speaker 2: that I think I mean soccer like was a classic
16:14
Speaker 2: example where it was just really hard to measure the
16:17
Speaker 2: most valuable players because stuff like scoring goals making tackles
16:20
Speaker 2: you know good defenders like Maldini legendary Italian defenders you
16:24
Speaker 2: know made a tackle once every two games and so
16:28
Speaker 2: on paper it's like you're not doing anything but he
16:30
Speaker 2: was remarkably good at controlling space and all these things
16:32
Speaker 2: are very hard to measure so I think I think
16:34
Speaker 2: we're only starting to scratch the surface actually in terms
16:36
Speaker 2: of what's driving a lot of these things that you're
16:38
Speaker 2: talking about. Because we're so reliant on, yeah, okay, what's
16:42
Speaker 2: goals scored? What's, you know, recovery and training? What's player availability?
16:46
Speaker 2: What's this kind of stuff? And actually a lot of it,
16:49
Speaker 2: because you can get these dramatic transitions, isn't just about
16:51
Speaker 2: those top level metrics.
16:53
Speaker 1: Yeah, love it. So I want to jump back to
16:56
Speaker 1: something you spoke about, which your new book, which we'll
16:58
Speaker 1: talk about, explores, which is certainty and uncertainty. It feels
17:05
Speaker 1: like this is, I don't know, so you tell me,
17:08
Speaker 1: feels like not many people love uncertainty and a lot
17:12
Speaker 1: of people love certainty, predictability, familiarity, you know, what is known,
17:18
Speaker 1: not what is unknown. And I think that's because you know,
17:24
Speaker 1: we're less fearful. Like if I know what's coming, even
17:28
Speaker 1: if it's not great, but I know what's coming, I'm
17:30
Speaker 1: more prepared, da, da, da, da. But it also feels
17:33
Speaker 1: to me from a psychological, sociological development point of view,
17:40
Speaker 1: we seem to grow and evolve and learn and adapt
17:43
Speaker 1: more in uncertainty. So when I'm working with people, coaching them,
17:49
Speaker 1: which I do quite a lot, I always talk about
17:52
Speaker 1: the value of doing the hard thing, not the easy
17:54
Speaker 1: thing or the unknown thing, not the known thing or
17:56
Speaker 1: the uncertain thing, not the certain thing, not in a
18:00
Speaker 1: reckless way, but in a strategic way. Like, you know,
18:03
Speaker 1: you're going to have to do, if you want to
18:05
Speaker 1: be a corporate speaker, you got to go do your
18:07
Speaker 1: first one at some stage and you're not going to
18:09
Speaker 1: be very good. And your skill won't be great. And
18:12
Speaker 1: your connection with the audience probably won't be great. But
18:15
Speaker 1: you can't get good at the thing that you're not doing. Right.
18:19
Speaker 1: So it seems a lot of people make their decisions
18:22
Speaker 1: around what is the safest, what is the most certain,
18:25
Speaker 1: what is the most familiar. Talk about that for a bit.
18:30
Speaker 3: Yeah, I think that's a really good point.
18:31
Speaker 2: I think it partly just reminds me of a lot
18:35
Speaker 2: of the work and where learning comes from.
18:38
Speaker 3: I think that's kind of what you're getting at.
18:40
Speaker 2: Even just from just understanding problems, understand the world, to
18:44
Speaker 2: some extent, variance and variability in outcomes is our friend, right? Because,
18:48
Speaker 2: you know, if you only ever do one thing and
18:50
Speaker 2: you only ever get one outcome, you learn absolutely nothing
18:52
Speaker 2: about if you change something and.
18:55
Speaker 3: What that realm of possibility looks like.
18:56
Speaker 2: And when we're doing research, for example, trying to evaluate
19:01
Speaker 2: whether things work, um and we work a lot with
19:04
Speaker 2: organizations at the moment on changes they've done and they
19:06
Speaker 2: want to know is this is this having an effect
19:08
Speaker 2: on performance is this not and you need variability you
19:11
Speaker 2: need different people to react in different ways you need
19:14
Speaker 2: different things to have been tried otherwise you just don't
19:16
Speaker 2: have um that underlying information there um i think also
19:20
Speaker 2: what you said about just just what people like having
19:23
Speaker 2: in front of them um i think we've seen actually
19:26
Speaker 2: some quite interesting evidence of that emerging around ai use
19:29
Speaker 2: so a lot of these um The models, I think
19:31
Speaker 2: GPT-6 has just come out, but the model two ago,
19:34
Speaker 2: there was some data they had when they were testing it.
19:37
Speaker 2: So what they do is they typically develop these AI
19:40
Speaker 2: tools in two phases. So one is they train.
19:42
Speaker 3: It on loads of data from.
19:43
Speaker 2: The Internet and try and get it to kind of
19:46
Speaker 2: you know essentially encode and memorize a lot of these
19:50
Speaker 2: these kind of facts and behaviors um so that's the
19:52
Speaker 2: first one is the kind of the raw data crunching
19:54
Speaker 2: element the second one is they put it in front
19:56
Speaker 2: of users and they you know whenever you get the
19:59
Speaker 2: kind of which response do you like or what yeah
20:00
Speaker 2: and they get essentially user feedback of making it more.
20:04
Speaker 3: Um yeah user friendly and palatable.
20:07
Speaker 2: For people and there's some really kind of striking data
20:11
Speaker 2: sets where if you give some of these models kind
20:14
Speaker 2: of multiple choice questions and ask it to report its
20:16
Speaker 2: confidence in an answer, you know, Ideally, you'd hope that
20:21
Speaker 2: the ones it's confident in, it gets right more, the
20:23
Speaker 2: ones it's less confident in, it gets right less.
20:25
Speaker 3: And the raw models for simple questionnaires are pretty good
20:28
Speaker 3: at that now.
20:29
Speaker 2: But once it goes through this user training, the model
20:32
Speaker 2: tends to be overconfident across the bar. So it's almost
20:35
Speaker 2: like interacting with people and saying like, what do you like?
20:37
Speaker 2: What kind of style of answer? People kind of hammer
20:41
Speaker 2: it out, any of that ability to be truthful. And
20:43
Speaker 2: I think there's this trade-off between that actually we see
20:47
Speaker 2: appearing in a few other places. And you see it
20:48
Speaker 2: a lot in life that if you're ever in a
20:51
Speaker 2: meeting or something, there's this trade-off between being right and
20:54
Speaker 2: keeping everyone happy. And you can see it actually just
20:59
Speaker 2: in kind of the costiness, if you think of it,
21:01
Speaker 2: of information. So if I ask you, what's the capital
21:03
Speaker 2: of Paris? Sorry, what's the capital of France? That's information
21:07
Speaker 2: that everyone knows is fairly uncontentious. It's quite a high
21:11
Speaker 2: cost to you getting it wrong, essentially. And if I say, oh,
21:14
Speaker 2: I think it's London, there's quite a low cost to
21:16
Speaker 2: you correcting me because everyone knows what the.
21:19
Speaker 3: Correct answer is.
21:20
Speaker 2: But if it's something that's maybe a bit contentious, you know,
21:22
Speaker 2: a food preference or something where people do debate it,
21:26
Speaker 2: you pushing for your way of, oh, I want this
21:28
Speaker 2: on the order. there's a big advantage perhaps for you
21:31
Speaker 2: and not so much of a cost. I think we're
21:33
Speaker 2: seeing that in AI is almost kind of surfacing a
21:38
Speaker 2: lot of in reality, the trade-offs that people make, but
21:41
Speaker 2: perhaps don't realize they're making it between, you know, wanting
21:44
Speaker 2: uncertainty and wanting accuracy perhaps, but actually their behavior is
21:48
Speaker 2: realizing that sometimes they just like a confident kind of
21:51
Speaker 2: people pleasing type response.
21:56
Speaker 1: People don't love being wrong, do they? But the practical
21:59
Speaker 1: real world reality is we're wrong all the time about
22:04
Speaker 1: many things. Like I get, I've been doing this forever
22:09
Speaker 1: and if I'm really blunt and honest and a little
22:12
Speaker 1: bit self-aware, I probably get 20 things wrong a day.
22:15
Speaker 1: It could be big, could be little. But to assume
22:18
Speaker 1: that you're always right, what is that need that people
22:22
Speaker 1: have for that where they just, even if they're, It's not,
22:28
Speaker 1: they don't have certainty, but it's like they believe they,
22:32
Speaker 1: like it's like they believe they know, whereas no, you
22:35
Speaker 1: don't know, you just think. And in reality, your version
22:39
Speaker 1: of what is absolute truth isn't, you know, what is
22:42
Speaker 1: that about?
22:43
Speaker 2: I think there's a couple of things at play there.
22:46
Speaker 2: I think one is as you alluded to, it's just,
22:49
Speaker 2: it's much more comfortable to have that. So, you know,
22:51
Speaker 2: whether it's just, it's just views on the, you know,
22:54
Speaker 2: quite complex issues about the world. It's very easy to
22:56
Speaker 2: be like, look, these people are wrong. These people are right.
22:58
Speaker 2: I like what they're doing. I don't like what they're doing.
23:02
Speaker 2: And to some extent it's, it's like almost a form
23:06
Speaker 2: of simplification that makes things very easy to navigate. Like
23:10
Speaker 2: if you have to sort of put things in boxes,
23:13
Speaker 2: this is a good thing, this is a bad thing,
23:15
Speaker 2: that's much easier than being like, actually this good thing
23:17
Speaker 2: has a bunch of bad stuff about it and uncertain
23:19
Speaker 2: stuff about it. You have to kind of, to have
23:22
Speaker 2: a view on that, you have to maintain a lot
23:24
Speaker 2: more complexity. So it's kind of just from a pure
23:26
Speaker 2: perspective sort of processing demand on our brains, it simplifies
23:33
Speaker 2: and it reduces the exertion on us to have something
23:36
Speaker 2: which is just much easier to work with. And I
23:39
Speaker 2: think there's a lot.
23:42
Speaker 2: I think there's also a practical element for decision-making. Lyndon Johnson,
23:46
Speaker 2: the US president, had this line that someone tried to
23:48
Speaker 2: give him a range of outcomes.
23:49
Speaker 3: He said, ranges are for cattle. Give me a number.
23:52
Speaker 3: Tell me what to do. I think there is a
23:55
Speaker 3: tension in a lot of the work we've done where.
23:58
Speaker 2: You have situations sometimes with a lot of uncertainty, particularly
24:01
Speaker 2: early in a problem or if it's a new problem
24:03
Speaker 2: you're dealing with. But people also need to take action.
24:06
Speaker 2: And how do you translate between that? And I think
24:09
Speaker 2: one of the things I spend a lot of my
24:10
Speaker 2: time thinking about is how do you convert a very
24:13
Speaker 2: uncertain situation into something, as you say, either where you.
24:17
Speaker 3: Can take a step perhaps with lower risk. You know,
24:19
Speaker 3: I think certainly.
24:21
Speaker 2: Anything where there's a chance of failure and the stakes
24:23
Speaker 2: are quite low, the best way to learn, as you said,
24:26
Speaker 2: is just do it. If you don't have enough information
24:29
Speaker 2: and it's cheap to make mistakes, make a mistake. Buy
24:32
Speaker 2: yourself some information cheaply. That's just the optimal kind of
24:35
Speaker 2: way to do it. I think where it becomes harder
24:37
Speaker 2: is where you've got something where the stakes are quite high,
24:40
Speaker 2: you have uncertainty. And then at some point you need
24:43
Speaker 2: to do something about it. And that's where sometimes converting
24:47
Speaker 2: even into something quite narrow. So we do a lot
24:49
Speaker 2: of work, for example, where I can't tell you exactly
24:52
Speaker 2: what the benefit of something is going to be, but
24:54
Speaker 2: I can tell you with confidence, you're not going to
24:57
Speaker 2: get it horrendously wrong. So yeah, I said, okay, there's uncertainty,
25:00
Speaker 2: there's a range here, but I can rule out these possibilities.
25:03
Speaker 2: And that, because you're narrowing that space of possibility, that's
25:06
Speaker 2: still quite useful for decisions.
25:07
Speaker 3: I think the big danger I often see with.
25:11
Speaker 2: people at the top of trees, whether it's on the
25:14
Speaker 2: political side, whether it's an organization, is there in that
25:17
Speaker 2: situation and their solution is to demand more confidence. So
25:20
Speaker 2: you're in a situation where you've got uncertainty, you have
25:22
Speaker 2: to make a decision and you say, I'll wait for
25:24
Speaker 2: some more data.
25:25
Speaker 3: And implicitly what you're doing there is avoiding making a decision.
25:28
Speaker 2: And I think that's, is that kind of like people
25:31
Speaker 2: freeze and they just, Oh, I'll wait, I'll wait. And
25:33
Speaker 2: you actually, you're deciding to not decide. And that in
25:36
Speaker 2: itself is taking action. And so that, That sort of
25:39
Speaker 2: balance between those two things I think is something that people.
25:42
Speaker 3: Often fall into. Yeah.
25:44
Speaker 1: Yeah, choosing not to make a decision is making a decision.
25:47
Speaker 1: You know, that's the irony of that. I think also
25:51
Speaker 1: like when people have certain ideas or theories or conspiracy
25:56
Speaker 1: theories or beliefs, they often get their, if not all,
26:02
Speaker 1: part of their identity through that. So if you question
26:06
Speaker 1: my truth, you question my identity, so fuck you. You know,
26:10
Speaker 1: it's a little bit that. It's like my sense of
26:14
Speaker 1: who I am ties into this thing that I believe
26:17
Speaker 1: being correct or unequivocal truth. And so it's very hard.
26:25
Speaker 1: I mean, the whole idea of objectivity and open-mindedness is
26:29
Speaker 1: fascinating because we all experience the world subjectively and we
26:33
Speaker 1: all interpret all the data's well, let's talk about global data.
26:38
Speaker 1: I'm not talking so much about actual data, but all
26:41
Speaker 1: that inbound kind of information, we turn it into a
26:45
Speaker 1: story and our truth becomes a version of what we've
26:48
Speaker 1: seen and observed and interpreted anyway. Like we're all in
26:52
Speaker 1: the same chat, but we're not in the same experience
26:56
Speaker 1: because your experience right now is not determined by what's
27:00
Speaker 1: going on. It's determined by how you respond to or
27:04
Speaker 1: think about or react to what's going on. Same with me,
27:07
Speaker 1: same with Tiff. So there's that whole kind of the,
27:11
Speaker 1: the idea of being truly open-minded, I think is almost
27:15
Speaker 1: impossible depending on the context.
27:20
Speaker 3: Yeah.
27:20
Speaker 2: I think that's a really important point of kind of how, um,
27:25
Speaker 2: how we think about things and how it kind of
27:28
Speaker 2: interacts with just a lot of how people are in
27:30
Speaker 2: terms of social structure. So a great talk, um, early,
27:33
Speaker 2: but yeah, it was a, it was a comedy writer
27:35
Speaker 2: for a lot of kind of quite well known comedy
27:38
Speaker 2: shows in the UK. And he made this point about
27:41
Speaker 2: a good joke or that the best jokes are often
27:43
Speaker 2: quite exclusive. Yeah. So he said like almost like the
27:46
Speaker 2: most powerful meaningful joke is like a joke between family
27:50
Speaker 2: members that nobody else gets and it's kind of that
27:53
Speaker 2: in joke and it's that bonding around and you often
27:55
Speaker 2: see that around a lot of comedy that you know
27:57
Speaker 2: you're there because you you kind of have this view
27:59
Speaker 2: on stuff that other people perhaps in other ways don't
28:03
Speaker 2: you know like you know the jokes you have amongst
28:05
Speaker 2: people who support the same sports team or kind of
28:06
Speaker 2: you know come from the same town or this kind
28:09
Speaker 2: of thing And I think it's when that kind of
28:12
Speaker 2: demand for community then starts to layer on to beliefs
28:15
Speaker 2: and other things, you can start to get some attention
28:18
Speaker 2: as you kind of talked about where actually it's belonging
28:22
Speaker 2: to that belief matters more than perhaps whether that's true
28:27
Speaker 2: or not, or what the implications of that thing are.
28:30
Speaker 2: And I think a lot of it you see is
28:33
Speaker 2: kind of where the stakes land. You know, if, if, if, if,
28:37
Speaker 2: I don't know, if you say your background colour is
28:39
Speaker 2: this and I say, no, it's not.
28:40
Speaker 3: This is quite a kind of low stakes.
28:42
Speaker 2: But I've definitely noticed, I mean, even amongst people, you know,
28:48
Speaker 2: scientists or, you know, kind of experts or people who
28:51
Speaker 2: kind of work a lot.
28:51
Speaker 3: And sometimes you find people kind of go a bit
28:53
Speaker 3: off with a view. And how do you end up there?
28:55
Speaker 2: And I think a lot of it is that just
28:58
Speaker 2: putting more and more kind of social capital and more
29:00
Speaker 2: on the line. And I've seen this in academic conferences
29:02
Speaker 2: where you get people who just defend ideas to the
29:05
Speaker 2: death because they've got paper on it. And if that's,
29:08
Speaker 2: you know, removed... there's no interest in like, Oh, this
29:12
Speaker 2: is great. Someone's discovered something new. This is like, well,
29:14
Speaker 2: this is my thing that I need to protect. And
29:17
Speaker 2: if someone else is, I kind of get defensive. I
29:20
Speaker 2: think it's a very natural response. And even if you
29:22
Speaker 2: look back through history, there's these, these kind of competing
29:25
Speaker 2: views for how research works and the term paradigm shift
29:29
Speaker 2: that a lot of people are familiar with, like Thomas
29:30
Speaker 2: Kuhn in the sixties kind of came up with the
29:33
Speaker 2: idea that a lot of ideas don't evolve because everyone
29:38
Speaker 2: kind of piles in and tries to just stuff. It basically,
29:41
Speaker 2: you get a bunch of people coalesce and defend it
29:42
Speaker 2: to the hilt. And then eventually you get enough momentum
29:46
Speaker 2: behind something new that the old one, I mean, Max
29:48
Speaker 2: Planck had this line, the physicist, that science advances one
29:52
Speaker 2: funeral at a time.
29:53
Speaker 3: That basically the.
29:55
Speaker 2: Old generation was just too powerful to let new ideas through,
29:58
Speaker 2: eventually start to kind of fall by the wayside and
30:01
Speaker 2: that bubbles up. So over time, you've got this actually
30:05
Speaker 2: quite kind of sociological outlook on how a lot of
30:08
Speaker 2: even kind of scientific discovery works beyond just everyone's trying
30:12
Speaker 2: to get, you know, get the truth every moment in time.
30:16
Speaker 1: Tiff, can you look up something for me? I'll start
30:18
Speaker 1: telling a story. I just, I don't want to mess
30:21
Speaker 1: it up too much. So the guy, he was a surgeon.
30:24
Speaker 1: I think he was Polish doctor. And he was the
30:27
Speaker 1: guy who came up with the idea of washing hands
30:31
Speaker 1: between surgeries, right? And this was before germ theory. This
30:35
Speaker 1: was before Pasteur. And it was before, I think it's
30:38
Speaker 1: Joseph Lister. But anyway, Prof, so... So that was Samuel Weiss, yeah?
30:45
Speaker 1: Was that? Yes, yes. And like 1847 or something, was it, Tiff? Anyway,
30:52
Speaker 1: so he thought... They had this, I forget the condition,
30:56
Speaker 1: but up to 18% of the women who gave birth
31:01
Speaker 1: were dying in the hospital. So almost one in five.
31:07
Speaker 1: And his theory was, he didn't understand the mechanism, but
31:10
Speaker 1: his theory was if we wash in this, it was
31:13
Speaker 1: like a chlorine-based cleaner where they washed and basically, you know,
31:18
Speaker 1: they sanitized their hands clean. And at that time, sorry, everybody,
31:23
Speaker 1: if this puts you off a little bit, but they
31:24
Speaker 1: literally were having doctors who would perform an autopsy and
31:28
Speaker 1: then go and deliver a baby or do a vaginal examination, right?
31:35
Speaker 1: Excuse me. It's just what was going on. I bet
31:37
Speaker 1: you're glad you didn't live then, Tiff. So they were
31:40
Speaker 1: going from one. And he kind of tried to encourage
31:45
Speaker 1: everybody in the clinic that he worked at. He had
31:48
Speaker 1: his own ward. to do this, and they all renounced him,
31:51
Speaker 1: scoffed at him, told him he was an idiot, blah, blah, blah.
31:55
Speaker 1: So he started this protocol, and in the rest of
31:59
Speaker 1: the hospital, it was up to 18%. I think the
32:02
Speaker 1: figure was 10% to 18% were dying. And he started this,
32:08
Speaker 1: and his patients, it was 1%, right? So he had–
32:13
Speaker 1: and it was all recorded, and so he– shouted it
32:17
Speaker 1: from the rooftops, and he was still, you probably know
32:19
Speaker 1: all this, anyway, he was still rejected, still made fun of,
32:23
Speaker 1: still criticized, and they carried on not washing their hands
32:28
Speaker 1: despite all this. I mean, that's a classic example of
32:32
Speaker 1: people not actually having any awareness or capacity to say, oh,
32:40
Speaker 1: it seems we're wrong and he's right, so let's do
32:44
Speaker 1: something better. doesn't sound very scientific at all.
32:49
Speaker 2: And I think there's, there's a term actually that came
32:51
Speaker 2: out of that called the Semmelweis reflex, which is when
32:54
Speaker 2: you hold some of the challenges, you know, it's this
32:57
Speaker 2: people came up with all these, you know, Oh, well
32:59
Speaker 2: it's this and it's this and something else and something else.
33:01
Speaker 2: And all these things in hindsight, you're like, well, obviously
33:04
Speaker 2: that's just a terrible idea to do it that way. But,
33:07
Speaker 2: but I think we, we still see that, that, you know,
33:09
Speaker 2: people's beliefs are being kind of challenged. I think there's also, um,
33:14
Speaker 2: An aspect of that, which you see pop up a
33:16
Speaker 2: lot in how we think about doing things better in,
33:20
Speaker 2: as you said, he didn't have an explanation for why,
33:23
Speaker 2: because we didn't have germ theory. We didn't actually know
33:25
Speaker 2: that kind of underlying what was going on. And I
33:27
Speaker 2: think you see that quite quite a lot of this
33:29
Speaker 2: tension between, you know, I've got data, I've done it differently.
33:33
Speaker 3: This is a better outcome.
33:35
Speaker 2: Like, you know, the manager taking over a team and
33:37
Speaker 2: getting better performance and just, you know, off the scale,
33:41
Speaker 2: this is just a turnaround. But if it's quite hard
33:44
Speaker 2: to explain why exactly is that, I think that can
33:47
Speaker 2: sometimes leave this kind of space for skepticism. And I
33:50
Speaker 2: think we see it even sometimes on the sort of
33:54
Speaker 2: negative side that when people kind of come up with
33:56
Speaker 2: these quite kind of extreme explanations for, you know, the
34:01
Speaker 2: shadowy forces behind everything. I think sometimes that's actually building
34:05
Speaker 2: on quite an innate human desire to explain things, that
34:10
Speaker 2: we want a why, we want something a bit deeper.
34:13
Speaker 2: So I completely agree. It's insane that that wasn't acted
34:17
Speaker 2: on sooner. I think that we do see parallels kind
34:21
Speaker 2: of coming into the modern world where people don't have
34:24
Speaker 2: a clear why, so it won't act. I think the
34:26
Speaker 2: other kind of sad footnote of that story is you
34:30
Speaker 2: probably came across the what happened to him later in life.
34:34
Speaker 2: So I think he became extremely bitter and extremely angry.
34:37
Speaker 2: So I think there's also this, this challenge that if
34:40
Speaker 2: you're not believed, what do you do about it? And
34:42
Speaker 2: I think he kind of just ended up in, in
34:44
Speaker 2: quite a bad place. I think there was a, there
34:46
Speaker 2: was a lot of kind of, reasons just just health-wise
34:50
Speaker 2: for him uh in the background anyway but i think
34:53
Speaker 2: yeah it's a really interesting case study both of what
34:56
Speaker 2: evidence people need to take action but also that at
34:59
Speaker 2: the individual level what do you do when you're not
35:01
Speaker 2: believed and also yeah yeah what convinces people do we
35:04
Speaker 2: need the why or is it enough just to have
35:06
Speaker 2: overwhelming evidence.
35:08
Speaker 1: I think i heard you say i might get this
35:10
Speaker 1: wrong but in one of something i watched of yours
35:13
Speaker 1: might have been the ted talk Hi, everybody. Really good
35:16
Speaker 1: TED Talk too. It was done last year, I think,
35:19
Speaker 1: was it?
35:19
Speaker 4: 2025?
35:19
Speaker 3: Yeah.
35:20
Speaker 1: Yeah, great talk, man. Fantastic. I was riveted from start
35:25
Speaker 1: to finish. I wanted to keep going. I'm like, this
35:27
Speaker 1: is not long enough. I want more. Anyway, you were
35:31
Speaker 1: talking about how, correct me if I get this wrong,
35:36
Speaker 1: but I'm pretty sure you said we know general anaesthetic works,
35:40
Speaker 1: but we don't really know how it works exactly. But nonetheless,
35:45
Speaker 1: people will go, people will put up their hand to
35:48
Speaker 1: have whatever operation, which clearly requires a general anesthetic. And
35:52
Speaker 1: even the person, the anesthetist doesn't necessarily know exactly how.
35:56
Speaker 1: They know if we give this much of whatever, this
35:58
Speaker 1: drug at this dose for this body weight for this,
36:01
Speaker 1: then that creates this outcome. But we still don't know why.
36:06
Speaker 2: Yeah and there's there's quite a few things like that
36:08
Speaker 2: in medicine where if you actually have to get into
36:10
Speaker 2: like what's the the fundamental kind of biology or physics
36:13
Speaker 2: that's going on that's causing that we don't have all
36:17
Speaker 2: the answers but we know if if we do this
36:19
Speaker 2: in this way we get the outcome we want i
36:22
Speaker 2: mean heart defibrillation is another one that actually just the
36:24
Speaker 2: um the nuts and bolts of the physics. There's still
36:27
Speaker 2: some open questions of why does this shock do this,
36:30
Speaker 2: but this shock not do this, but we, it's just
36:33
Speaker 2: been refined essentially experimentally that this, this massively improves someone's chances. Um,
36:39
Speaker 2: even if, you know, we, we can't write down all
36:40
Speaker 2: the equations neatly. Um, and we're happy with that because
36:44
Speaker 2: I think there's a lot of things in medicine. I mean,
36:47
Speaker 2: you look at a lot of trial, uh, treatment trials,
36:49
Speaker 2: for example, as well. And the, the point of the
36:51
Speaker 2: treatment trial isn't to tell you why something works is
36:54
Speaker 2: to tell you, if you have it or don't have it,
36:56
Speaker 2: what's the effect on the outcome? So you still get
36:58
Speaker 2: some of these things where things work really well, but
37:02
Speaker 2: actually just the underlying, well, what's the mechanism deep down
37:06
Speaker 2: in the body? There's still some open questions there.
37:10
Speaker 1: So like it or not, we have to deal with certainty,
37:12
Speaker 1: navigated certainty and accept, sorry, uncertainty and accept uncertainty because
37:18
Speaker 1: it's an unavoidable part of the human experience, right?
37:23
Speaker 3: It is.
37:23
Speaker 2: And I think, particularly a lot of the things we
37:26
Speaker 2: care about where we interact with uncertainty you know if
37:30
Speaker 2: we're dealing with things that are new or dealing with
37:32
Speaker 2: change um or growth those are often things where it's
37:37
Speaker 2: exciting because it's you know it's putting us in new
37:38
Speaker 2: situations we're learning we're developing we're building but by definition
37:43
Speaker 2: we've got to to cope with that and i think
37:44
Speaker 2: it's um It is really striking to have worked both
37:48
Speaker 2: on things like health emergencies, trying to build a startup.
37:52
Speaker 3: That kind of feeling is extremely uncomfortable.
37:55
Speaker 2: I think it's one of the things, in hindsight, you
37:56
Speaker 2: look back on and realize that you just learned a
37:58
Speaker 2: huge amount and developed a lot. But I think there
38:01
Speaker 2: is a comfort, and I think we see it across
38:04
Speaker 2: areas of life, that desire for predictability. in many cases
38:09
Speaker 2: can be quite positive because to take the uncertainty of
38:12
Speaker 2: a problem and try and make progress on it is
38:14
Speaker 2: to some extent to reduce that uncertainty because yes you
38:17
Speaker 2: know if you're trying to get i'm trying to get
38:20
Speaker 2: an idea off the ground and there's uncertainty what you
38:22
Speaker 2: want to do is reduce that so you work out
38:24
Speaker 2: if it's going to going to succeed or not um
38:27
Speaker 2: but you need to take that on in the first
38:29
Speaker 2: place i think obviously where there's things where there's a
38:32
Speaker 2: yeah a lot of certainty no predictability often the opportunities
38:35
Speaker 2: for actually developing or learning or innovating is just much
38:38
Speaker 2: much lower.
38:39
Speaker 1: Is our need for certainty a fear-based drive.
38:43
Speaker 2: That's a really good question um i think there's There's certainly,
38:51
Speaker 2: I think, just from an evolutionary point of view, going
38:54
Speaker 2: to be elements where predictability was extremely valuable. If you're
38:59
Speaker 2: trying not to get killed by something, having confidence about
39:04
Speaker 2: whether or not that's going to be a problem is valuable.
39:06
Speaker 3: And we see that in the modern world as well.
39:08
Speaker 2: If you're facing some kind of crisis or difficult situation,
39:11
Speaker 2: you want confidence about what your options are, what the
39:15
Speaker 2: nature of the threat is going to be.
39:18
Speaker 3: And I think that.
39:20
Speaker 3: Maybe it's a fear.
39:21
Speaker 2: I think it can also be an element of curiosity.
39:25
Speaker 2: And I think certainly kind of from my background, I
39:27
Speaker 2: think a lot of people who work in science doing
39:29
Speaker 2: a PhD these kind of things is very much about
39:33
Speaker 2: trying to put more certainty into the world, you know,
39:36
Speaker 2: put more, here's the thing we don't understand, now we do.
39:40
Speaker 2: So I suspect there's probably that, like with a lot
39:42
Speaker 2: of things, there's that deeper evolutionary level, like we just
39:44
Speaker 2: don't want to get killed by something. But also the
39:48
Speaker 2: more recent, driven by curiosity, that when you've got something
39:52
Speaker 2: that's unsolved, you know, even like I've got young kids
39:56
Speaker 2: and you see with the jigsaws, they're trying to take
39:57
Speaker 2: chaos and find order in it. And I think that's
40:01
Speaker 2: quite a healthy response against that slight, kind of fear
40:04
Speaker 2: and paranoia perhaps it's a bit further back in our
40:06
Speaker 2: evolutionary history.
40:08
Speaker 1: Now if you had to go somewhere and you couldn't
40:10
Speaker 1: drive yourself and it was a half hour trip and
40:14
Speaker 1: you had the option of going with dave in the
40:17
Speaker 1: uber or a self-driving car what would you choose.
40:26
Speaker 2: I think if it was something I didn't know very well,
40:30
Speaker 2: I think having a human would probably be my vote.
40:33
Speaker 2: I mean, I think even with self-driving cars, I think
40:36
Speaker 2: we've seen them rolled out a lot in cities where
40:38
Speaker 2: essentially there's a lot more data, there's a lot more
40:41
Speaker 2: kind of structure. I think for me, some of the
40:46
Speaker 2: data is looking really good actually in these very, very
40:48
Speaker 2: structured environments.
40:49
Speaker 3: Although I do love this story.
40:50
Speaker 2: There was a great one in LA a few years
40:51
Speaker 2: ago where I think it's one of the early iterations,
40:54
Speaker 2: the self-driving car, if you've ever driven in LA, you
40:56
Speaker 2: kind of have to cut people up occasionally, basically just
40:59
Speaker 2: to kind of get around. So self-driving cars just waiting
41:01
Speaker 2: at a junction for an hour because it was just
41:03
Speaker 2: following the rules a bit too tightly.
41:05
Speaker 3: But I think, yeah, where.
41:07
Speaker 2: It's kind of like out of its data set a
41:09
Speaker 2: bit in new areas, I think that's situations where human
41:14
Speaker 2: ability to pattern match and adapt is still really valuable.
41:18
Speaker 1: Was it you that I heard talking about I won't
41:20
Speaker 1: ruin the story. You can share it with us. The
41:23
Speaker 1: yellow fire truck?
41:25
Speaker 2: Yeah, this is an example a couple of years ago
41:27
Speaker 2: now where there's a self-driving car. There were fire trucks
41:32
Speaker 2: responding to an emergency in Vegas and a self-driving car
41:34
Speaker 2: just got in the way of one of them, basically,
41:36
Speaker 2: just completely blocked it off. And it had been trained
41:38
Speaker 2: to recognize emergency vehicles, but it had been trained on
41:41
Speaker 2: data sets where the emergency vehicles are red. So it
41:44
Speaker 2: hadn't identified that the yellow fire truck was a fire
41:46
Speaker 2: truck it had to get out of the way of.
41:49
Speaker 3: Wow.
41:50
Speaker 1: Fire trucks are only red.
41:52
Speaker 3: I'm not moving.
41:55
Speaker 1: That's hilarious. All right. So the book is called Proof,
41:59
Speaker 1: the Art and Science of Certainty. Give us a snapshot
42:04
Speaker 1: of that, if you would, Adam.
42:06
Speaker 3: Yeah, I think a lot of it is.
42:08
Speaker 2: Kind of motivated partly just by my curiosity about these
42:12
Speaker 2: things over the years. I think coming from initially a
42:14
Speaker 2: field like maths, that's very, you know, you've got the
42:16
Speaker 2: answer and that's certain. And then realizing that even that
42:19
Speaker 2: field had a lot of problems, you know, in Europe,
42:22
Speaker 2: for example, negative numbers weren't really accepted until well into
42:26
Speaker 2: the 1700s, because we developed so much maths around geometry
42:30
Speaker 2: that people have kind of just ignored it. And obviously
42:32
Speaker 2: then it started running into a lot of issues. And
42:35
Speaker 2: then the book kind of, goes from what's seen as
42:38
Speaker 2: that very, you know, pure, everything's known and certain, to
42:42
Speaker 2: the modern world, whether it's looking at everything from legal
42:45
Speaker 2: systems to AI, to emergencies, to health, really trying to
42:52
Speaker 2: look fundamentally at that question of how far would you
42:54
Speaker 2: go in your search for certainty? What would convince you
42:58
Speaker 2: that something's true? What would convince you there's enough evidence?
43:00
Speaker 2: But also crucially, once you get there, how do you
43:03
Speaker 2: convince other people? And I think that's often the bit
43:05
Speaker 2: that we ignore. We focus a lot on what.
43:08
Speaker 3: Will convince us.
43:09
Speaker 2: And I think we see all these examples where somebody
43:13
Speaker 2: convinces themselves but fails to convince other people. And what
43:15
Speaker 2: are the reasons around that just in terms of how
43:17
Speaker 2: we process information, how collectively action emerges? And I think
43:23
Speaker 2: just also just coming into the modern era, we're seeing
43:25
Speaker 2: a lot with how AI is coming up with knowledge,
43:29
Speaker 2: coming up with solutions that we don't fully understand. That's
43:32
Speaker 2: quite uncomfortable. Self-driving cars, if they're much safer, but occasionally
43:36
Speaker 2: have accidents that we don't understand, How much of a
43:39
Speaker 2: problem is that? Because there's that tension between wanting the
43:42
Speaker 2: why and just wanting better performance. And the book really
43:47
Speaker 2: is an attempt to get into why people set the
43:49
Speaker 2: bar where they do. And you realize, I think, going
43:52
Speaker 2: into everything from science to law, there's all these things
43:54
Speaker 2: that have just developed over the years that we can't
43:57
Speaker 2: just take for granted. Oh, that's just how we do it.
44:00
Speaker 2: And that's how we evaluate evidence. And that's where we
44:01
Speaker 2: set the bar. And haven't necessarily thought about why that
44:05
Speaker 2: is or whether that's actually something we're comfortable with.
44:09
Speaker 1: I also think from a psychology point of view, you know,
44:14
Speaker 1: somebody like me who's been getting in cars with drivers
44:17
Speaker 1: for 62 years and driving cars for 45 years or
44:21
Speaker 1: whatever it is. Yeah, I would do it, but I
44:25
Speaker 1: would be a lot more apprehensive than if I was
44:27
Speaker 1: driving myself, even if it was statistically more likely that
44:31
Speaker 1: I would crash than the self-driving car. So I think
44:34
Speaker 1: there's that emotional and psychological component to, embracing this technology
44:41
Speaker 1: as well, or trusting it perhaps. Yeah.
44:44
Speaker 2: And I think one of the big, um, shifts we've
44:47
Speaker 2: even had, I mean, if you look back, um, there's a,
44:49
Speaker 2: there's a great story about Abraham Lincoln that when he was,
44:52
Speaker 2: so he had a run at Congress, wasn't very successful,
44:54
Speaker 2: went back to being a lawyer for a bit. And
44:56
Speaker 2: he got really concerned that he didn't really understand what
44:59
Speaker 2: it means to demonstrate something. He, you know, he would
45:02
Speaker 2: kind of give speeches and wasn't really clear about, well,
45:04
Speaker 2: what am I, what am I talking about here?
45:05
Speaker 3: So he went back and he taught, taught himself from scratch.
45:09
Speaker 3: all of Greek geometry.
45:10
Speaker 2: So all these books with all these quite fiddly geometric
45:13
Speaker 2: proofs by candlelight. But in a way, in that era,
45:17
Speaker 2: that was possible. You could be a really engaged lawyer
45:21
Speaker 2: with a bit too much time on your hands in
45:22
Speaker 2: the evenings. Teach yourself, you know, teach yourself physics. But
45:26
Speaker 2: I think where we are now, you know, you can't
45:29
Speaker 2: build yourself a self-driving car. And, you know, but like
45:32
Speaker 2: 50 years ago, people could, you know, you could pick
45:33
Speaker 2: up a car, you could work out all the pieces,
45:35
Speaker 2: you could have that control and understanding.
45:37
Speaker 3: And I think we're seeing that in a lot more.
45:39
Speaker 3: um of areas of.
45:40
Speaker 2: Life where it's almost we're having to defer that understanding
45:45
Speaker 2: to some extent to other systems and then i think
45:47
Speaker 2: that raises these questions around well what does make us
45:49
Speaker 2: trust something what is the level of certainty we're happy
45:52
Speaker 2: about and it's very it's very difficult to decouple just
45:55
Speaker 2: the you know here's the performance from actually we're moving
45:59
Speaker 2: away from things that previously we would have been able
46:01
Speaker 2: to to get our heads around if we really wanted
46:03
Speaker 2: to to something where we're just never going to be
46:05
Speaker 2: able to do that.
46:06
Speaker 1: Yeah What's funny is I just thought of this, like
46:10
Speaker 1: we all have wanted or not, like it or not,
46:14
Speaker 1: you know, most of us, or even my mum who's
46:17
Speaker 1: 87 in a minute has got an iPhone, right? Not
46:19
Speaker 1: that she really knows what to do with it, let's
46:21
Speaker 1: be honest, bless her. But we all have a relationship
46:25
Speaker 1: with technology, right? Because we have to. I don't know
46:29
Speaker 1: if relationship's the right word, but we all interact and
46:31
Speaker 1: intersect with it. So, yeah, it's... I know you've got
46:36
Speaker 1: a hard out, so I'm just going to tell everyone again,
46:38
Speaker 1: the book is called Proof, the Art and Science of Certainty.
46:42
Speaker 1: Is it on audio as well, Adam?
46:45
Speaker 3: It is, yeah.
46:46
Speaker 2: I think there was a couple of deals on various
46:49
Speaker 2: platforms in Australia recently, so hopefully people can pick up
46:51
Speaker 2: a bargain.
46:53
Speaker 3: Yep, awesome.
46:54
Speaker 1: We appreciate you, sir. You're super interesting, super smart, and
46:58
Speaker 1: that was a super relevant conversation. So thanks for agreeing
47:02
Speaker 1: to be on the show, on the You Project.
47:04
Speaker 3: Thanks for having me. It was great to chat.