WEBVTT

00:00:01.040 --> 00:00:01.520
Dave P.

00:00:02.960 --> 00:00:04.160
David Pinsoff.

00:00:04.640 --> 00:00:07.120
Podcast about Evolutionary Psychology.

00:00:07.360 --> 00:00:09.920
Tagline is Actual Evolutionary Psychology.

00:00:10.240 --> 00:00:12.480
By actual evolutionary psychologists.

00:00:12.640 --> 00:00:14.560
And that this is what it is.

00:00:14.880 --> 00:00:16.480
Yep, let's continue listening.

00:00:17.280 --> 00:00:17.600
Okay.

00:00:25.440 --> 00:00:31.390
David, um Paul Smaldino is a I don't know how would you characterize

00:00:31.470 --> 00:00:32.670
Paul Smaldino?

00:00:33.390 --> 00:00:39.390
Um hmm polymathic, um good word, rigorous,

00:00:39.630 --> 00:00:43.870
insightful, um musical genius.

00:00:43.950 --> 00:00:46.190
He has a a band called Small Dinosaurs.

00:00:46.590 --> 00:00:47.310
Oh nice.

00:00:47.550 --> 00:00:49.390
Is this is it ongoing?

00:00:49.950 --> 00:00:50.750
I I don't know.

00:00:50.910 --> 00:00:52.910
I I we talked about it a bit.

00:00:53.310 --> 00:00:55.950
I think did we I think we talked about it a bit at the end of the episode.

00:00:56.260 --> 00:00:56.580
Yeah.

00:00:56.660 --> 00:00:57.460
Um, I don't know.

00:00:57.540 --> 00:00:58.500
I didn't listen yet.

00:00:58.820 --> 00:01:00.260
I haven't listened to the music.

00:01:00.820 --> 00:01:02.740
No, I haven't listened to the episode.

00:01:04.100 --> 00:01:05.700
Can I just you were there during the episode?

00:01:06.180 --> 00:01:10.900
But I honestly I do I sometimes re- I often re-listen to our episodes because I'm so I've

00:01:10.900 --> 00:01:16.020
heard other podcast hosts say this that they're so paying attention to the

00:01:16.020 --> 00:01:19.220
conversation that they don't hear everything.

00:01:19.380 --> 00:01:24.470
The first the the the the real time, but it's a different way of processing.

00:01:24.790 --> 00:01:27.350
Yeah, it's a different mindset because like when you're in the middle of a conversation,

00:01:27.430 --> 00:01:31.430
you're thinking about what you're gonna say next and what question you're gonna ask, and

00:01:31.510 --> 00:01:33.590
like you know, how to interact with the person.

00:01:33.670 --> 00:01:38.150
But when you're listening as a bystander or as a third party, you get to just take in the

00:01:38.150 --> 00:01:39.910
whole gestalt in a very different way.

00:01:40.390 --> 00:01:43.590
So what you're saying is we don't listen to the guest when we're interviewing.

00:01:43.830 --> 00:01:44.310
No, kidding.

00:01:44.390 --> 00:01:45.750
No, no, no, it's the opposite, actually.

00:01:45.830 --> 00:01:50.440
It's it's I I for me, it's like it's so much that I am so carefully listening to what

00:01:50.440 --> 00:01:54.440
they're saying and then trying to figure out where to go.

00:01:54.840 --> 00:01:55.160
Yeah.

00:01:55.400 --> 00:01:58.040
Um, yeah, this is probably very boring for listeners.

00:01:58.120 --> 00:02:02.360
But anyway, the point is I don't remember that from the Paul Smaldino episode.

00:02:02.440 --> 00:02:08.200
But what I do remember is the Lego um um solar system.

00:02:08.850 --> 00:02:11.330
Yes, we did talk about a Lego solar system.

00:02:11.570 --> 00:02:11.730
Yeah.

00:02:11.890 --> 00:02:15.090
So by now, you can't wait to hear this episode.

00:02:15.170 --> 00:02:20.770
So I mean we all what whilst we talked about mathematical models, yeah.

00:02:20.850 --> 00:02:24.210
The main thing is like why they're why math models models are stupid, we need more of them,

00:02:24.290 --> 00:02:25.170
which is you know a thing.

00:02:25.330 --> 00:02:27.730
We talked about the natural selection of bad science.

00:02:27.810 --> 00:02:29.970
Uh oh yeah, big yeah, we talked about that a lot.

00:02:30.290 --> 00:02:35.490
I mean, I think the the bottom line thing with Paul is that he is embodying and is a voice

00:02:35.650 --> 00:02:41.560
about it doesn't matter what people intend, whether it's social policy or good science or

00:02:41.560 --> 00:02:46.120
bad science, what matters is the selection criteria of the institutions.

00:02:46.360 --> 00:02:49.960
And he's doing modeling and empirical work, all about that.

00:02:50.200 --> 00:02:55.240
And I think the more that everybody sort of gets on board with whatever people intend, what

00:02:55.240 --> 00:02:57.000
we select for is what we get.

00:02:57.240 --> 00:02:59.720
I think that's a really uh important thing.

00:02:59.800 --> 00:03:02.760
And I think that's that's sort of what the conversation was about.

00:03:03.090 --> 00:03:04.450
Yeah, it was a it was a fun one.

00:03:04.530 --> 00:03:08.610
We also get into punk rock and covert signaling, sending signals that some people might get

00:03:08.610 --> 00:03:10.130
and some people might not get.

00:03:10.370 --> 00:03:16.050
So hopefully um this podcast will be full of signals that you get and uh no one else

00:03:16.050 --> 00:03:16.530
gets.

00:03:17.090 --> 00:03:18.930
Or maybe we want everyone to get it.

00:03:19.090 --> 00:03:19.810
I don't know.

00:03:20.290 --> 00:03:22.690
Uh you be the judge.

00:03:22.930 --> 00:03:23.650
You be the judge.

00:03:24.320 --> 00:03:24.560
All right.

00:03:24.640 --> 00:03:26.240
Enjoy enjoying the enjoyment.

00:03:31.440 --> 00:03:33.840
Um so yeah, let's get started.

00:03:34.000 --> 00:03:39.840
Um Paul is we're we we always do just an intro beforehand, so we can just get right

00:03:39.840 --> 00:03:40.320
into it.

00:03:40.560 --> 00:03:44.000
David, do you wanna do you wanna start um with this?

00:03:44.320 --> 00:03:44.640
Sure.

00:03:44.770 --> 00:03:45.010
Yeah.

00:03:45.250 --> 00:03:47.410
Well, I I am David Pinsoff.

00:03:47.490 --> 00:03:48.770
That is an identity.

00:03:48.930 --> 00:03:52.770
Uh I know I I I have a prediction, but I'm not I'll shut up.

00:03:52.850 --> 00:03:54.290
I'll let you ask the question, yeah.

00:03:54.850 --> 00:03:59.490
Well, it's not necessarily the type of identity that Paul Smaldino is interested in,

00:03:59.730 --> 00:04:02.930
because there's only one David Pinsov so far as I'm aware.

00:04:03.170 --> 00:04:08.930
Um but Paul, you're interested in identities that many people share and what functions they

00:04:08.930 --> 00:04:09.330
serve.

00:04:09.490 --> 00:04:15.340
Uh this is a huge topic in uh life as we know it in the United States and in many

00:04:15.340 --> 00:04:16.220
other places.

00:04:16.460 --> 00:04:21.820
Um there's lots of talk about identity, uh who I am, who we are.

00:04:22.140 --> 00:04:24.620
Uh you have a very interesting way of thinking about it.

00:04:24.700 --> 00:04:26.700
So, what is your way of thinking about it?

00:04:28.140 --> 00:04:32.140
Yeah, uh I mean I think identity is just sort of wrapped up.

00:04:32.380 --> 00:04:38.220
It's hard to be a person in the world and not have identity be something that

00:04:38.220 --> 00:04:39.660
you think about a lot.

00:04:39.900 --> 00:04:45.890
Uh and I think the way that a lot of people think

00:04:46.050 --> 00:04:51.650
about it is very sort of phenomenologically, meaning like the the experience of

00:04:52.530 --> 00:04:55.970
who I am in the world and what that means to me.

00:04:56.450 --> 00:05:01.010
Um, and I I think about things uh that that's true.

00:05:01.090 --> 00:05:04.930
Like that is a true thing about identity and a real thing and an important thing about

00:05:04.930 --> 00:05:05.410
identity.

00:05:05.570 --> 00:05:11.220
But I'm much more interested in its aspect as you know, the functional role of

00:05:11.220 --> 00:05:11.700
identity.

00:05:11.940 --> 00:05:13.140
What's identity for?

00:05:13.300 --> 00:05:15.700
And how do we use it purposefully?

00:05:15.860 --> 00:05:21.300
And I um this is where I think, you know, sort of evolutionary thinking, be it uh

00:05:21.620 --> 00:05:27.380
genetic evolution or uh sort of cultural evolution, which I tend to be a bit more

00:05:27.380 --> 00:05:32.530
uh interested in, but You know,

00:05:34.370 --> 00:05:40.370
the way we identify ourselves and the way we categorize others based on

00:05:41.410 --> 00:05:42.610
our identity categories.

00:05:42.690 --> 00:05:47.890
And identity is just a sort of a kind of category that we impose on other people.

00:05:49.090 --> 00:05:51.890
We use that for all kinds of adaptive decision making.

00:05:51.970 --> 00:05:57.970
Like how do we find partners for cooperation or finding a romantic

00:05:57.970 --> 00:05:58.850
partner or friends?

00:05:59.010 --> 00:06:04.700
How do we use what we perceive about something in order to gauge what kind of react

00:06:04.860 --> 00:06:09.660
uh interaction we can have with them, whether it's at the store or a stranger or at a party

00:06:09.740 --> 00:06:11.100
or in a classroom?

00:06:11.340 --> 00:06:17.340
Um how do we use identity to think about how we evaluate a

00:06:17.340 --> 00:06:18.540
source of information?

00:06:18.780 --> 00:06:19.740
Do we believe it?

00:06:19.900 --> 00:06:20.620
Do we accept it?

00:06:20.780 --> 00:06:24.170
Do we say this is the kind of thing that I'm definitely going to accept?

00:06:24.330 --> 00:06:26.890
Are we much more skeptical or rejecting of it?

00:06:27.050 --> 00:06:31.210
Do we use this to attack somebody or dismiss somebody?

00:06:31.690 --> 00:06:37.370
Um what kind of person am I also affects

00:06:37.370 --> 00:06:39.210
adaptive decision making.

00:06:39.370 --> 00:06:45.370
Like, do I am I bound by these rules because I am this kind of person and

00:06:45.370 --> 00:06:49.770
therefore tied to this kind of community and will judge be judged by this set of norms

00:06:49.770 --> 00:06:50.730
within this group?

00:06:50.970 --> 00:06:56.900
I think identity is really a kind of category that is at the core

00:06:56.900 --> 00:06:59.460
of so much of our behavior and decision making.

00:07:00.100 --> 00:07:05.860
And just to say, like all categories in general, like categorization is

00:07:05.860 --> 00:07:09.780
kind of uh one of the fundamental things that we do, right?

00:07:09.860 --> 00:07:14.500
That's foundational to how we behave and how we think, how we perceive the world.

00:07:14.660 --> 00:07:20.020
And I just think that identity is a really important social category that is

00:07:20.450 --> 00:07:25.090
more fleshed out and developed and nuanced in humans than in other species.

00:07:26.210 --> 00:07:31.410
Do you view identity as identical to categories in that it serves the same function as a

00:07:31.410 --> 00:07:32.050
category?

00:07:32.210 --> 00:07:37.730
Like if I were to categorize, you know, an animal, say birds, um, that's going to help me

00:07:37.730 --> 00:07:41.410
make predictions about um uh a new bird that I haven't seen before.

00:07:41.490 --> 00:07:44.610
I'm gonna assume that it flies, that it builds nests or whatever.

00:07:44.770 --> 00:07:50.260
And that's a useful category for me because it helps me make predictions uh about new cases

00:07:50.340 --> 00:07:51.140
I haven't seen.

00:07:51.540 --> 00:07:55.860
Similar similarly, if I if I see a new fruit, I'm gonna assume that it's sweet.

00:07:56.020 --> 00:07:59.380
Um and so these categories help structure our understanding of the world.

00:07:59.540 --> 00:08:01.940
Do you view identities as just the social version of that?

00:08:02.100 --> 00:08:07.380
It's just about uh helping us make better predictions about uh people and and and in what

00:08:07.380 --> 00:08:10.260
ways they're likely to behave or is it something more than that?

00:08:12.830 --> 00:08:18.190
yeah I be basically yes uh I think of identity as a a kind of

00:08:18.190 --> 00:08:23.710
category and you know uh being a a sort of special kind it has uh

00:08:24.430 --> 00:08:26.190
the features that it has.

00:08:26.510 --> 00:08:32.030
Um like the category identity is a kind of category that uh

00:08:32.420 --> 00:08:34.180
has certain features.

00:08:34.420 --> 00:08:40.340
But yeah, I do think that the main to me, the most helpful way to think about identity is

00:08:40.340 --> 00:08:41.780
as a kind of category.

00:08:42.180 --> 00:08:42.420
Okay.

00:08:42.740 --> 00:08:47.540
And David, can I well let me just poke on on that too a little bit, which is that, but

00:08:47.540 --> 00:08:50.500
there's also the part, and Paul, this is also like what you study.

00:08:50.580 --> 00:08:55.940
So I'm just saying what you've said is that it would be also as if like the fruit was also

00:08:55.940 --> 00:08:59.170
in a strategic uh stance towards the world.

00:08:59.330 --> 00:09:02.770
So unlike the fruit case where we're talking about the third-party perspective of well,

00:09:02.930 --> 00:09:07.010
what can I make an inference about for this fruit, there's also this sort of

00:09:07.010 --> 00:09:12.290
entrepreneurial or creative part of identity or strategic part of identity, negotiating

00:09:12.290 --> 00:09:15.890
identities or creating them or embodying them or broadcasting them.

00:09:16.370 --> 00:09:16.690
Right.

00:09:16.850 --> 00:09:19.800
And I I think that that's exactly right.

00:09:19.960 --> 00:09:24.200
And beyond that, it's not just uh that the fruit is, you know, behaving, right?

00:09:24.360 --> 00:09:25.720
Other people are behaving.

00:09:25.880 --> 00:09:31.720
Um, and so we have to think about strategic uh views on on what because due to their

00:09:31.720 --> 00:09:34.120
someone else's identity, how will they behave?

00:09:34.280 --> 00:09:38.520
But we also have to think about that they are another person with theory of mind who has,

00:09:38.760 --> 00:09:41.560
you know, categories about identity of their own.

00:09:41.720 --> 00:09:45.400
And we're all making decisions with limited information.

00:09:45.720 --> 00:09:51.490
And so understanding that another person may have their own categories of identity,

00:09:51.650 --> 00:09:56.690
they may have categorized me in a way that I wouldn't categorize myself.

00:09:56.930 --> 00:10:02.450
Or maybe I hope that they categorize me that is not necessarily entirely accurate, but it's

00:10:02.450 --> 00:10:05.890
useful strategically for them to categorize me in that way.

00:10:06.130 --> 00:10:10.210
And so I think the fact that you have lots of people categorizing each other in an

00:10:10.210 --> 00:10:14.530
interacting population makes one of the things that makes identity so interesting.

00:10:15.280 --> 00:10:21.120
And that would also explain then for what I'm hearing you say is why, and this is like a

00:10:21.120 --> 00:10:24.560
very David Pinsafian and Paul Smalldinian Smalldin.

00:10:27.120 --> 00:10:29.520
Small small small Dinonian.

00:10:30.080 --> 00:10:31.920
Whatever you go with, just don't it.

00:10:33.360 --> 00:10:34.160
I like it.

00:10:34.240 --> 00:10:37.280
Um is that uh there's a lot of negotiation.

00:10:37.470 --> 00:10:43.070
Like one of the one of the jobs in in social status competition

00:10:43.310 --> 00:10:49.150
is to say, here are the identities, or let's try out if this this

00:10:49.470 --> 00:10:53.870
we attribute this identity to you and act on behalf of this identity.

00:10:54.190 --> 00:10:55.950
Um That's right.

00:10:56.110 --> 00:11:01.550
And I we we often sort of put on, we try on roles, we try on identities in different

00:11:01.550 --> 00:11:03.630
contexts, and we sort of see how they fit.

00:11:03.790 --> 00:11:09.300
And we we modify how we view other people and we modify how we present ourselves,

00:11:09.540 --> 00:11:14.180
um, you know, at various times in our lives and also in different contexts.

00:11:14.420 --> 00:11:17.540
And so there are also like, you know, yeah.

00:11:17.700 --> 00:11:21.300
So this is one of the things that I think is so interesting about identity because it's

00:11:21.300 --> 00:11:25.540
actually extreme, it's not one thing, it's it's actually very multifaceted.

00:11:25.620 --> 00:11:29.540
And we have aspects of our identity that are pretty permanent.

00:11:29.780 --> 00:11:35.570
We have aspects that are maybe, you know, more malleable but over longer

00:11:35.570 --> 00:11:36.210
time scales.

00:11:36.370 --> 00:11:39.890
I mean, part of our identity is like what age category we are in, right?

00:11:40.210 --> 00:11:46.050
Um, and then there is the ones that are are very fluid and that change by context,

00:11:46.530 --> 00:11:51.810
um, like who we're with or what purpose we're in, um, what role we take.

00:11:51.890 --> 00:11:57.240
Uh, you maybe you're you're in a leadership role in like your you know

00:11:57.560 --> 00:12:03.000
garage band that you go practice with but you're actually in a subordinate role at work uh

00:12:04.200 --> 00:12:09.400
and I think yeah so we're really good at

00:12:09.640 --> 00:12:12.920
knowing what our role is in different contexts.

00:12:13.000 --> 00:12:16.680
And so I think this is one of the reasons it makes sense to think of identity as a kind of

00:12:16.680 --> 00:12:17.400
category.

00:12:17.650 --> 00:12:17.970
Yeah.

00:12:18.450 --> 00:12:22.770
One of the uh ways you study, or can you talk a little bit about how you study social

00:12:22.770 --> 00:12:23.170
identity?

00:12:23.330 --> 00:12:27.570
Because this is also a transition to sort of how you do your work, because you know, many

00:12:27.570 --> 00:12:30.770
listeners they might not know how you approach topics.

00:12:30.930 --> 00:12:34.850
And so you could imagine somebody just sort of approaching the the idea of identity and

00:12:34.850 --> 00:12:38.770
what you've just described with sort of paper and pencil surveys, like a social

00:12:38.770 --> 00:12:40.770
psychologist, no offense to that.

00:12:40.930 --> 00:12:46.840
But you know, the way you approach top any topic you study is with a bunch of cool

00:12:47.000 --> 00:12:52.200
tools that are, you know, you're pretty you're you're one of the the people doing these

00:12:52.200 --> 00:12:53.640
things and writing books about them.

00:12:53.720 --> 00:12:57.480
So can you talk a little bit about like how do you approach these and why do you use that

00:12:57.480 --> 00:12:57.960
tool?

00:12:58.440 --> 00:12:58.920
Sure, yeah.

00:12:59.080 --> 00:13:01.640
So I mean what you're talking about is is modeling, right?

00:13:01.720 --> 00:13:04.360
Is is mathematical and computational modeling.

00:13:04.600 --> 00:13:07.880
And and I do do those other things too a bit.

00:13:08.040 --> 00:13:13.950
Like the the you know behavioral experiments, because uh and it's important to

00:13:13.950 --> 00:13:18.590
actually you know see what people really do in like a lab setting in a behavioral setting

00:13:18.750 --> 00:13:22.670
it's also important to see what people do not in a lab setting right which is why sort of

00:13:22.670 --> 00:13:28.350
field work and naturalistic observation are important um but i i

00:13:28.510 --> 00:13:34.270
so i mostly do modeling i i the the bulk of my work i would say 80 to 90

00:13:34.270 --> 00:13:39.480
percent of the work i do is building mathematical or

00:13:39.480 --> 00:13:41.400
computational models.

00:13:41.800 --> 00:13:47.080
Um and I can I can talk more about how that works and what that is and uh etc.

00:13:47.480 --> 00:13:51.880
But you know, part of it is like I think the world is really complicated.

00:13:52.120 --> 00:13:54.280
I I mean I I know that the world is really complicated.

00:13:54.520 --> 00:13:55.560
Controversial, controversial take.

00:13:55.800 --> 00:13:56.040
Yeah, right.

00:13:56.120 --> 00:13:57.000
It's hot take.

00:13:57.160 --> 00:14:03.070
Uh it's just a few simple gizmos, just you know, you got your your atoms

00:14:03.150 --> 00:14:08.510
and you know they're all just uh yeah atoms and time atoms and time put that on a

00:14:08.510 --> 00:14:14.510
t-shirt uh yeah so I think it's really

00:14:14.750 --> 00:14:20.750
it's if it's really hard to think about all the things and we all have to make

00:14:20.750 --> 00:14:26.600
simplifications and we like every single person makes simplifications in our uh

00:14:26.680 --> 00:14:28.440
just going about the world, right?

00:14:28.600 --> 00:14:29.640
Going about our day.

00:14:29.880 --> 00:14:34.920
You don't think about every single molecule of dirt or sand or pavement on the street.

00:14:35.000 --> 00:14:40.040
You don't worry about the wind or the well, you know, unless unless you do, right?

00:14:40.200 --> 00:14:43.000
Unless the thing is important for your decision making.

00:14:43.160 --> 00:14:48.840
So most of the things that you know, when we're walking around doing our thing, we ignore

00:14:48.840 --> 00:14:54.770
tons of stuff if they're not likely to be important about the decisions we need to make.

00:14:55.010 --> 00:15:00.930
And when we build experiments, we come up with some paradigm

00:15:00.930 --> 00:15:04.530
where people behave in some way and we try to get them to do something.

00:15:04.850 --> 00:15:10.450
And we're usually not like, oh, what I'm really mostly interested in is like how people

00:15:10.450 --> 00:15:15.580
sitting at a laptop will click, you know, which one of these things at this time, like

00:15:15.660 --> 00:15:17.580
that's the main thing I'm interested in.

00:15:17.740 --> 00:15:23.660
You know, behavioral experiments these days are increasingly online, um, or like fill out a

00:15:23.660 --> 00:15:24.780
survey or whatever.

00:15:24.940 --> 00:15:30.460
We're interested in in what we try to do when we build experiments

00:15:30.700 --> 00:15:36.620
is capture some essence of the real question that we're interested in, whether it's about

00:15:36.700 --> 00:15:41.420
you know, identity or cognition or behavior or whatever it is.

00:15:41.900 --> 00:15:44.310
And that's limited.

00:15:44.470 --> 00:15:46.550
There's only so much you can do there.

00:15:47.110 --> 00:15:52.230
And it also leaves a lot of unexamined assumptions.

00:15:53.030 --> 00:15:55.350
And that's just the nature of the thing, right?

00:15:55.430 --> 00:16:01.430
When you deal with a very complex system, you have to ignore a lot of it in order to

00:16:01.430 --> 00:16:04.710
get any tractability and do science.

00:16:05.360 --> 00:16:11.040
and actually like isolate a signal that is strong enough to give you a sort of this is

00:16:11.360 --> 00:16:13.520
going this way and not that way.

00:16:13.920 --> 00:16:19.840
Um and as a supplement

00:16:19.840 --> 00:16:25.600
to that, which is sort of where I come in mostly, is taking trying to

00:16:25.600 --> 00:16:31.520
build sort of artificial systems, fake systems that capture the sort of essence

00:16:31.520 --> 00:16:37.410
of uh whatever dynamic that we're talking about, whatever the sort of components of the

00:16:37.410 --> 00:16:42.530
dynamic we're talking about, and think about the kinds of interactions, the properties that

00:16:42.530 --> 00:16:46.690
they have, and the way they interact and evolve over time, change over time.

00:16:47.970 --> 00:16:53.330
And sort of run that through mathematics or through computer simulation

00:16:53.880 --> 00:16:59.560
and sort of examine, okay, uh, based on these assumptions I made, have the

00:16:59.560 --> 00:17:02.120
outcomes that I predicted really happened?

00:17:02.360 --> 00:17:04.920
Have they is there something that I didn't expect?

00:17:05.080 --> 00:17:06.760
Uh, can I explain that?

00:17:07.000 --> 00:17:11.720
Um, that's a really powerful thing about modeling is when something I love it when

00:17:11.720 --> 00:17:13.400
something happens that you don't expect.

00:17:13.800 --> 00:17:18.310
And then because when something happens you don't expect in the real world, you're like, it

00:17:18.310 --> 00:17:19.350
could be a million things.

00:17:19.510 --> 00:17:19.910
I don't know.

00:17:20.070 --> 00:17:23.670
Maybe I don't have enough data to figure out what it is.

00:17:23.830 --> 00:17:26.390
But with the model, you actually've built the thing.

00:17:26.550 --> 00:17:28.630
You know everything that went into it.

00:17:28.950 --> 00:17:34.630
So if something happens that you don't expect, the main tool you have now is your brain to

00:17:34.630 --> 00:17:40.440
think, okay, well, you know, based on what I've put

00:17:40.440 --> 00:17:43.160
in, why did this thing come out?

00:17:43.320 --> 00:17:48.440
And sometimes it's because you made a mistake, and sometimes it's because you assumed

00:17:48.440 --> 00:17:52.040
something that you thought didn't matter, it turns out it mattered a lot.

00:17:52.600 --> 00:17:58.600
And sometimes it's because there were just unexpected consequences of what you

00:17:58.600 --> 00:17:58.920
did.

00:17:59.930 --> 00:18:05.450
The other really cool thing for me about doing these kinds of models are just the

00:18:05.450 --> 00:18:09.850
tremendous power to run variations and run lots of things.

00:18:10.090 --> 00:18:15.450
Like when we do experiments, it's you're so constrained by like how much money you have to

00:18:15.450 --> 00:18:21.370
pay participants or to go collect data or how much time a person is

00:18:21.370 --> 00:18:24.410
willing to sit and answer questions for, right?

00:18:24.730 --> 00:18:30.680
Or um, you know how many uh research assistance you can get to like run all

00:18:30.680 --> 00:18:32.200
the variants of your study.

00:18:32.440 --> 00:18:37.400
Or, you know, oh, it would actually be unethical to like kill a bunch of people or like

00:18:37.400 --> 00:18:40.760
poison them or like, you know, release a disease into the world.

00:18:40.920 --> 00:18:45.080
Like we definitely can't do that experimentally and we wouldn't want to, but we would kind

00:18:45.080 --> 00:18:49.480
of sometimes want to know what would happen on let's say a really long time scale that's

00:18:49.480 --> 00:18:53.290
really hard to study, especially when we're doing evolutionary thinking, right?

00:18:53.370 --> 00:18:59.050
It's like if you all you have is experiments, especially with humans, you mostly have

00:18:59.210 --> 00:19:05.130
at your tools, you have archaeology to look at the past and maybe some paleontology and

00:19:05.130 --> 00:19:05.770
history.

00:19:06.490 --> 00:19:10.490
And then you have you know this snapshot of what's happening now.

00:19:10.570 --> 00:19:15.370
And you try to make these really complicated inferences about like, okay, well, how did

00:19:15.370 --> 00:19:20.340
what we see now come out of what we think we know about the past.

00:19:20.980 --> 00:19:26.340
And using simulations, we can sort of say, well, okay, well, if the past was like that, and

00:19:26.340 --> 00:19:32.340
if these sort of general principles of change and behavior apply the way we think they do,

00:19:32.820 --> 00:19:37.460
then we can kind of simulate forward in time and look at different variations.

00:19:37.540 --> 00:19:43.230
We can run these simulations thousands of times With basically the same things.

00:19:43.310 --> 00:19:48.030
We can randomize a bunch of um starting conditions to try to get them to start over and

00:19:48.030 --> 00:19:48.910
over again.

00:19:49.070 --> 00:19:52.350
Um, so I I do this a lot, I really like it.

00:19:52.590 --> 00:19:58.350
Um I I one of the things uh and I I'm happy to talk a lot about this.

00:19:58.430 --> 00:20:04.270
Um I I wrote a textbook on doing this kind of modeling, sort of a general social

00:20:04.270 --> 00:20:10.000
science um approach, although um it definitely has sort of an evolutionary

00:20:10.000 --> 00:20:10.880
flavor.

00:20:11.280 --> 00:20:16.320
And uh I don't know I'm working actually on a second textbook uh that is more evolutionary

00:20:16.320 --> 00:20:22.160
in its uh approach uh explicitly um but um

00:20:22.480 --> 00:20:27.200
one of the things that I really like about this approach in addition to I think math is

00:20:27.200 --> 00:20:32.320
cool, I think computers are cool, I think these help me learn um is I'm also just very kind

00:20:32.320 --> 00:20:35.040
of ADD in my interests.

00:20:35.430 --> 00:20:41.110
and it takes like a really long time to set up a lab or a field site and I know this you

00:20:41.110 --> 00:20:45.430
know I have friends and colleagues who who do this kind of work right you you yourself Dave

00:20:45.590 --> 00:20:51.190
right like you know it takes so long to to get good experiments done

00:20:51.510 --> 00:20:57.030
yeah it's like making a movie as opposed to switching channels or channels but like

00:20:57.030 --> 00:21:01.990
watching it watching movies versus making movies it's like making making the setting up the

00:21:01.990 --> 00:21:03.910
lab is like making a movie it's really slow.

00:21:04.520 --> 00:21:05.560
That's a good analogy.

00:21:05.720 --> 00:21:11.240
And I think, like, you know, uh, instead of you know,

00:21:11.720 --> 00:21:17.560
if I just wrote the movies, yeah, and that's more like what I do, which is

00:21:17.640 --> 00:21:23.320
like I I know how the you know it works in theory, and I can just kind of like imagine it,

00:21:23.480 --> 00:21:28.890
except I can do more than imagine it and I can simulate it, but the once you sort of learn

00:21:28.890 --> 00:21:34.330
how to code and and do these kinds of analyses right like the startup time to getting

00:21:34.330 --> 00:21:40.170
simulations um the the thing that takes the longest honestly is the design part is

00:21:40.250 --> 00:21:44.090
just making sure that the model makes sense and is a good model and I think that's the

00:21:44.090 --> 00:21:49.930
hardest thing to do and uh but and once you can do that right running

00:21:51.320 --> 00:21:54.040
it running simulations is pretty low cost.

00:21:54.440 --> 00:21:59.160
Um, especially like if you're like me and are at a major university and you have access to

00:21:59.160 --> 00:22:04.600
a high performance computing cluster, you can run tens of thousands of

00:22:04.600 --> 00:22:06.440
simulations in a day.

00:22:06.680 --> 00:22:12.450
Uh I it's it's so funny to me when I talk to people who do simulations and they'll

00:22:12.450 --> 00:22:15.330
just say, Well, I don't want to do that.

00:22:15.410 --> 00:22:17.650
Like that's too many simulations to run.

00:22:17.730 --> 00:22:20.290
It's going to take like a week to get all that data.

00:22:20.530 --> 00:22:22.050
And I'm like, Are you kidding me?

00:22:22.210 --> 00:22:25.410
Like talk to an infant researcher or an anthropologist.

00:22:25.570 --> 00:22:26.930
Yeah, or field anthropologist.

00:22:27.010 --> 00:22:27.250
Yeah.

00:22:27.650 --> 00:22:30.690
Um, my wife has a PhD in developmental psychology.

00:22:30.770 --> 00:22:35.650
And so like she did this like you know longitudinal work that uh for her dissertation that

00:22:35.650 --> 00:22:39.330
took uh you know a year and a half just to collect all the data.

00:22:39.840 --> 00:22:44.240
And yeah, a field anthropologist who's spending, you know, months out of the air every

00:22:44.240 --> 00:22:49.680
every year going back, you know, over years to collect data for a single study.

00:22:49.840 --> 00:22:54.400
Um, so I find it sort of laughable when people complain about it taking like a week to do

00:22:54.400 --> 00:22:54.560
data.

00:22:54.800 --> 00:22:57.600
And it's not even like you have to be there for that week.

00:22:57.760 --> 00:23:01.040
You just like hit run and then walk away.

00:23:01.360 --> 00:23:07.070
Um so the the challenge really is uh, but you know, the bar

00:23:07.230 --> 00:23:12.590
is is is then the standard becomes like do you do something meaningful with it?

00:23:12.750 --> 00:23:16.270
Do you do something that makes sense that is develops insight, right?

00:23:16.430 --> 00:23:18.510
Because otherwise it's like what the fuck am I doing?

00:23:18.750 --> 00:23:19.070
Yeah.

00:23:19.230 --> 00:23:19.470
Yeah.

00:23:21.630 --> 00:23:22.270
Yeah.

00:23:22.670 --> 00:23:28.280
So it might be useful to like um take a concrete example of how this

00:23:28.280 --> 00:23:29.800
approach is used.

00:23:30.040 --> 00:23:36.040
Um so have you run simulations or developed models relating to identity

00:23:36.040 --> 00:23:40.200
that that you could walk us through like how you break apart the com the components?

00:23:40.440 --> 00:23:44.520
Or is there another model that you think is a good illustrative example of how this

00:23:44.520 --> 00:23:45.800
approach works?

00:23:46.200 --> 00:23:46.840
Yeah, sure.

00:23:46.920 --> 00:23:49.800
I'm I I think I mean you know the answer is yes to that.

00:23:50.040 --> 00:23:55.850
Uh uh uh let's talk about uh we can talk about the covert signaling stuff, because a lot a

00:23:55.850 --> 00:23:56.330
lot on that.

00:23:56.490 --> 00:23:57.450
Um okay.

00:23:57.770 --> 00:24:03.770
So uh this came out of

00:24:03.770 --> 00:24:09.770
an idea uh basically first of all that context would

00:24:10.330 --> 00:24:15.450
influence the the nature of identity as is if you think of identity not only as a category

00:24:15.530 --> 00:24:21.360
but as a communicated piece of information that I know, you know, I

00:24:21.360 --> 00:24:23.200
have I know things about myself.

00:24:23.520 --> 00:24:25.520
I, you know, am a certain way.

00:24:25.680 --> 00:24:30.080
I am affiliated with certain groups, I have certain commitments, and I meet you, a

00:24:30.080 --> 00:24:30.720
stranger.

00:24:31.040 --> 00:24:33.840
And I'm gonna communicate certain things to you.

00:24:34.080 --> 00:24:39.520
And there's an old literature uh from anthropology mostly and sociology, social psychology

00:24:39.680 --> 00:24:44.370
have sort of worked on this too, on uh like ethnic markers for identity tags.

00:24:44.530 --> 00:24:50.290
It's like, well, if we live in a world where people are different from each other and they

00:24:50.290 --> 00:24:56.050
meet strangers, it would be good to have a way sort of from a distance, at a safe distance,

00:24:56.210 --> 00:25:02.050
to be able to know very quickly some things about, you know, what this person is likely to

00:25:02.050 --> 00:25:04.420
be, uh, what this person is likely to be like.

00:25:04.660 --> 00:25:10.420
And so the way they dress, the way they, you know, decorate themselves, that's

00:25:10.420 --> 00:25:11.140
obvious, right?

00:25:11.300 --> 00:25:13.300
Their fashion, their, their, their clothing.

00:25:13.460 --> 00:25:15.380
You can see that really far away.

00:25:15.540 --> 00:25:19.860
Um then there are certain things like the way they move, certainly the way they talk, like

00:25:19.940 --> 00:25:24.500
obviously the way they they use language, the language they speak, but also the the accents

00:25:24.500 --> 00:25:26.980
they have, the the mannerisms of their speech.

00:25:27.140 --> 00:25:30.660
These are things we can really easily use to categorize people.

00:25:30.820 --> 00:25:32.330
And this really helps.

00:25:32.490 --> 00:25:38.170
Um, and it helps uh those kinds of ideas are really were really influential

00:25:38.730 --> 00:25:44.250
to me, but they're really kind of based on this idea of fairly

00:25:44.250 --> 00:25:50.010
small insular groups that where everyone in the group is an in group member and works with

00:25:50.010 --> 00:25:53.770
each other, and everyone from another group is an out group member and works with each

00:25:53.770 --> 00:25:53.930
other.

00:25:54.090 --> 00:25:58.410
And so we just want to know like if I meet somebody in the forest, are they a member of my

00:25:58.410 --> 00:26:03.720
sort of larger collective, you know, let's say tribal group or identity group, or are they

00:26:03.720 --> 00:26:09.560
from you know a different place or from a different and if so, what kind of person are

00:26:09.560 --> 00:26:09.720
they?

00:26:09.880 --> 00:26:13.000
Are they a a group that like we have relationships with?

00:26:13.160 --> 00:26:14.600
Are they someone that we're enemies with?

00:26:14.760 --> 00:26:15.880
I need to know.

00:26:16.680 --> 00:26:22.570
But now we live in really diverse societies where we're interacting constantly with people

00:26:22.890 --> 00:26:27.930
that we don't know, we've never met before, that nobody we know knows

00:26:28.090 --> 00:26:29.050
necessarily.

00:26:29.370 --> 00:26:35.210
Um, and we need to be able to make sort of rapid decisions about them.

00:26:35.290 --> 00:26:39.130
Because sometimes, like often, one of the things that really characterizes our, you know,

00:26:39.210 --> 00:26:44.330
modern societies is we sort of behave cooperatively with strangers every day.

00:26:44.820 --> 00:26:50.580
Just like you go to the store, you you know, buy things from the person to the store, you

00:26:50.580 --> 00:26:52.820
don't try to steal from the other customers, right?

00:26:52.900 --> 00:26:56.020
You like ride the bus or the train or get on a plane, right?

00:26:56.100 --> 00:26:57.620
Where we're all sort of calmly, right?

00:26:57.780 --> 00:27:03.300
Sarah Hurdy has this famous example of like, if you took like 200 chimpanzees and put them

00:27:03.300 --> 00:27:06.790
on an airplane in like a locked, you know, cabin in the sky.

00:27:06.870 --> 00:27:11.270
Like you would by the time that plane landed, you'd have like a bunch of dead chimpanzees

00:27:11.350 --> 00:27:12.950
because they'd all have killed each other.

00:27:13.190 --> 00:27:15.110
Um movie option idea.

00:27:15.350 --> 00:27:15.510
Yeah.

00:27:15.750 --> 00:27:15.830
Yeah.

00:27:15.990 --> 00:27:16.870
Chimps on a plane.

00:27:17.030 --> 00:27:18.150
Yeah, chimps on a plane.

00:27:18.950 --> 00:27:19.270
Yeah.

00:27:19.430 --> 00:27:20.310
I don't know about you guys.

00:27:20.390 --> 00:27:24.070
Every time I go to the grocery store, I form a coalition of males and I uh raid someone

00:27:24.070 --> 00:27:25.270
else's uh shopping cart.

00:27:25.670 --> 00:27:28.790
I was gonna say d I think David might be the exception to the what was that?

00:27:29.030 --> 00:27:30.390
So I I'm an exception to those rules.

00:27:30.630 --> 00:27:31.350
Yeah, exactly.

00:27:31.880 --> 00:27:32.680
that's fair.

00:27:32.840 --> 00:27:36.120
Uh I mean that's why I don't go to Ralph's.

00:27:36.280 --> 00:27:36.840
But yeah.

00:27:37.240 --> 00:27:43.160
Um the part of the the the sort of um

00:27:43.560 --> 00:27:49.400
so I I had this initial idea that like the way that society was structured,

00:27:49.640 --> 00:27:53.320
about the way the society was structured, and then the the sort of largeness and the

00:27:53.320 --> 00:27:59.090
diversity of the society would influence the sort of breadth versus depth trade-off in the

00:27:59.090 --> 00:28:04.290
way people signal their identity to each other and the sort of nature of the context

00:28:04.450 --> 00:28:05.730
specificity of that.

00:28:05.970 --> 00:28:09.410
And that's an idea that I've held on to for quite a long time.

00:28:09.570 --> 00:28:15.250
I think I I first started really thinking hard about that about maybe 10, 11 years ago, 12

00:28:15.250 --> 00:28:16.050
years ago.

00:28:17.660 --> 00:28:23.580
I really sort of I published like a sort of conceptual idea paper about it, you know,

00:28:23.740 --> 00:28:25.100
shortly thereafter.

00:28:25.500 --> 00:28:31.420
And uh just have like the first kind of major modeling paper on this

00:28:31.420 --> 00:28:33.900
idea coming out, hopefully coming out.

00:28:33.980 --> 00:28:37.100
Um we're just about to submit the final revisions uh soon.

00:28:37.260 --> 00:28:43.250
But um and i i'll talk more about that but but um one idea

00:28:43.250 --> 00:28:49.010
that kind of came along with that i i uh developed was this

00:28:49.010 --> 00:28:54.610
idea that we actually aren't just trying to honestly signal about

00:28:54.610 --> 00:29:00.610
ourselves um it's actually a lot of the time we we hide information we're not only

00:29:00.610 --> 00:29:06.480
dishonest but just sort of um uh sort of sneaky about, you know, or

00:29:06.560 --> 00:29:12.480
or um selective, let's say is the word um about how we talk about eat ourselves

00:29:12.480 --> 00:29:13.120
to other people.

00:29:13.200 --> 00:29:14.880
And that obviously we do that, right?

00:29:14.960 --> 00:29:19.280
We we just like you don't walk into a store and just start like yelling your political

00:29:19.280 --> 00:29:20.080
beliefs, right?

00:29:20.240 --> 00:29:24.800
You use, you know, you use uh uh judgment, right?

00:29:24.960 --> 00:29:25.680
Prudence, right?

00:29:25.760 --> 00:29:27.360
You're you're you're being savvy.

00:29:27.760 --> 00:29:32.570
Um But one of the things that's really interesting is what happens when you have a diverse

00:29:32.570 --> 00:29:38.570
society, and like often you find ourselves in positions, both like in person and

00:29:38.570 --> 00:29:43.290
certainly now online, uh, where there's a large audience, multiple listeners.

00:29:43.770 --> 00:29:49.530
And the thing that occurs then is every person, if you have people from

00:29:49.530 --> 00:29:55.530
diverse backgrounds with diverse experiences and assumptions and categories, if

00:29:55.530 --> 00:30:01.380
I say something that potentially gives information away about who

00:30:01.380 --> 00:30:05.540
I am, that's going to be interpreted differently by different people.

00:30:05.780 --> 00:30:11.540
I mean, the most simple version of that is like one person is like, great, they're like me.

00:30:11.620 --> 00:30:12.740
I might want to be friends with them.

00:30:12.820 --> 00:30:16.900
And another person is like, oh, they're part of that group that I hate.

00:30:17.140 --> 00:30:22.980
Like, I want to fight them or rob them or just like not help them or

00:30:22.980 --> 00:30:23.860
whatever.

00:30:24.500 --> 00:30:26.730
And what happens?

00:30:26.890 --> 00:30:28.250
How do I talk about myself?

00:30:28.410 --> 00:30:33.850
Or, you know, uh well, first a further version of that is like, what if, you know,

00:30:34.410 --> 00:30:40.250
if I'm a member, let's say, of an oppressed minority, I might want to talk about myself in

00:30:40.250 --> 00:30:44.170
a way that I need to find friends, I need to find people who will help me, I might need to

00:30:44.170 --> 00:30:46.490
find, you know, romantic partners.

00:30:47.540 --> 00:30:53.140
But I don't I don't want to risk outing myself in public and be

00:30:53.140 --> 00:30:57.940
potentially ostracized or hurt or imprisoned or killed.

00:30:58.180 --> 00:31:03.460
And so I need to be more selective about the way I speak.

00:31:03.620 --> 00:31:08.500
And one of the things that's really cool about human language, and not just human language

00:31:08.580 --> 00:31:13.380
in terms of spoken words, but also in terms of like the way that we communicate more

00:31:13.380 --> 00:31:18.690
broadly, is how nuanced, ambiguous, and

00:31:18.690 --> 00:31:20.130
multifaceted it is.

00:31:20.610 --> 00:31:24.850
There's the linguists have this term called implicature, which I really like, right?

00:31:24.930 --> 00:31:29.970
Which is me like the thing that the meaning that is implied by this the statement, right?

00:31:30.050 --> 00:31:31.490
This comes from pragmatics.

00:31:31.730 --> 00:31:37.410
So uh if I glance at an open window and say, it's a little chilly in here,

00:31:38.340 --> 00:31:43.940
I don't have to say like the implicature is maybe you should close the window, even though

00:31:43.940 --> 00:31:45.700
I didn't say it explicitly.

00:31:46.500 --> 00:31:52.260
And uh what's neat is that because we because

00:31:52.420 --> 00:31:58.420
language is complicated, because we have enough life experience to

00:31:58.420 --> 00:32:02.900
draw on previous categories, we are smart enough to think about lots of things in different

00:32:02.900 --> 00:32:04.420
ways, contexts, etc.

00:32:04.890 --> 00:32:09.290
I can say one thing that will have different implicatures to different people.

00:32:10.330 --> 00:32:16.250
And when this is used strategically for uh identity

00:32:16.250 --> 00:32:20.330
signaling, this is what we uh call covert signaling.

00:32:21.130 --> 00:32:27.130
So um, this is work that I originally developed with Tom Flampson and uh Richard McElreath,

00:32:27.450 --> 00:32:30.490
and have have continued to develop it with other colleagues.

00:32:30.730 --> 00:32:36.640
Um the this idea that um you

00:32:36.640 --> 00:32:42.320
know I could signal overtly and just that's like hey this is who I am let me tell you very

00:32:42.320 --> 00:32:47.040
clearly, unambiguously, this is something that you want to know about me, right?

00:32:47.440 --> 00:32:52.880
Like I'm a gay socialist who loves football and you know,

00:32:53.040 --> 00:32:55.600
whatever, hates square dancing.

00:32:56.320 --> 00:33:02.290
Um and that's gonna be

00:33:02.290 --> 00:33:06.290
really effective at making sure nobody is mistaken about who I am.

00:33:06.370 --> 00:33:11.010
And therefore, like it's gonna really you know maximize the fact that the the the chance

00:33:11.010 --> 00:33:15.330
that people who are similar to me are gonna know that they're similar to me and therefore,

00:33:15.410 --> 00:33:18.370
you know, might wanna extend relationships with me.

00:33:18.610 --> 00:33:24.210
On the other hand, though, if anyone is gonna be turned off by that or enraged by that,

00:33:24.460 --> 00:33:26.940
that's gonna be very bad for me.

00:33:27.580 --> 00:33:33.420
On the other hand, right, I could say something like, I really love this movie, or

00:33:33.500 --> 00:33:39.100
you know, I come in with a t-shirt that has an obscure reference, or just

00:33:39.420 --> 00:33:45.100
use jokes or fashion or other kinds of sometimes we call them dog whistles,

00:33:46.250 --> 00:33:51.850
Aspects of speech and communication that if you know what I'm talking about,

00:33:52.090 --> 00:33:57.850
if you are also aware of, you know, that these clothes or that movie

00:33:57.850 --> 00:34:02.810
reference or whatever, if you know those things, that the people who are likely to know

00:34:02.810 --> 00:34:07.930
those things are also likely to be similar to me and less similar to be members of other

00:34:07.930 --> 00:34:08.410
groups.

00:34:08.490 --> 00:34:13.130
And therefore, people who are similar to me will catch those references, whereas the people

00:34:13.130 --> 00:34:15.050
who are different from me will miss them.

00:34:15.420 --> 00:34:17.660
Because language can have multiple implicatures.

00:34:17.740 --> 00:34:19.020
I could say something.

00:34:19.180 --> 00:34:24.940
Um, this is one of the reasons why, like, you know, for most people who like who like

00:34:24.940 --> 00:34:26.540
humor, which is most people, right?

00:34:26.620 --> 00:34:30.460
Like, there's this old trope that like the funniest people you know are actually like your

00:34:30.460 --> 00:34:33.020
friends are the funniest people you know, always.

00:34:33.260 --> 00:34:36.380
Like the people who can make you laugh the most consistently are your friends.

00:34:36.460 --> 00:34:38.700
And there's like a very good reason for that, right?

00:34:38.860 --> 00:34:44.070
Like humor is an incredible tool For

00:34:44.230 --> 00:34:50.230
establishing that like we have shared values, shared references, and so I can make

00:34:50.230 --> 00:34:55.990
a joke and it will land really well with my friends who like were there

00:34:56.310 --> 00:35:00.950
at the thing that I'm referencing, or you know, understand that that thing is funny in the

00:35:00.950 --> 00:35:01.990
way that I meant it.

00:35:02.150 --> 00:35:05.670
Whereas other people might just say, you know, they might hear what I say.

00:35:05.830 --> 00:35:10.940
And it's not like I'm talking another language, they can hear the literal words I mean, but

00:35:10.940 --> 00:35:13.420
there is there's the this there's the text, right?

00:35:13.500 --> 00:35:14.780
And there's subtext.

00:35:15.500 --> 00:35:18.940
And subtext can act as this kind of covert signal.

00:35:19.260 --> 00:35:21.180
So that's the basic idea.

00:35:21.340 --> 00:35:25.100
And when we build these models, it's really helpful to start with a basic idea.

00:35:25.260 --> 00:35:28.460
This is what we call a verbal theory, what I've just talked about.

00:35:28.540 --> 00:35:34.380
It's it's an idea, it's talked about in words, um, it it seems to make

00:35:34.380 --> 00:35:40.010
sense, and then we build models in order to do a few things.

00:35:40.250 --> 00:35:46.010
One is like develop a proof of concept, like, okay, well, here's a way it

00:35:46.010 --> 00:35:50.330
could really work or might really work under certain assumptions.

00:35:50.570 --> 00:35:52.650
And here are also the boundary conditions.

00:35:52.810 --> 00:35:55.610
Here are sort of when it would or wouldn't work.

00:35:55.690 --> 00:35:58.970
And we can show how it would or wouldn't work.

00:35:59.050 --> 00:36:02.490
And we can even use this to make explicit predictions.

00:36:04.020 --> 00:36:09.460
At some case, if you have very good data and very good theory, you can even make um

00:36:09.700 --> 00:36:11.940
explicit quantitative predictions.

00:36:12.180 --> 00:36:18.180
But a lot of these models uh are used at least initially to develop more qualitative

00:36:18.180 --> 00:36:19.060
predictions.

00:36:19.620 --> 00:36:25.620
Um so uh some of the some of the the first modeling we did of this was to just sort

00:36:25.620 --> 00:36:30.980
of establish uh uh basically you create an artificial world

00:36:31.460 --> 00:36:36.430
where you you say, okay, well, the the world is going to be populated by people.

00:36:37.150 --> 00:36:39.550
What are the properties of those people?

00:36:39.630 --> 00:36:40.750
What do they do?

00:36:41.230 --> 00:36:46.750
Well, everyone has uh in these

00:36:46.750 --> 00:36:49.310
models some things about them.

00:36:49.870 --> 00:36:55.710
And to anyone else, to make things easier, we just said, well, let's just say any two pairs

00:36:55.710 --> 00:36:59.020
of people are either similar or not similar.

00:37:00.060 --> 00:37:05.260
And in a later model, we say, well, we can make explicit things, we give people like a

00:37:05.260 --> 00:37:09.820
vector of things about them, and we establish a cutoff where if you share enough things in

00:37:09.820 --> 00:37:12.220
common, you're similar, otherwise, you're dissimilar.

00:37:12.460 --> 00:37:16.060
And you could do this if you think that's a bad assumption, you could do it and you want

00:37:16.220 --> 00:37:17.340
you should do it another way.

00:37:17.500 --> 00:37:23.470
That's you've just identified one of the great things about modeling, which is that if if I

00:37:23.470 --> 00:37:28.750
do modeling and I say, look, I've made I've made this discovery with my model, you can say,

00:37:28.910 --> 00:37:33.950
well, you've made an assumption that doesn't hold in the systems that I've I'm interested

00:37:33.950 --> 00:37:34.190
in.

00:37:34.350 --> 00:37:35.310
And that's great.

00:37:35.550 --> 00:37:39.790
That means that we can you can take my model as a jumping off point and build on it and

00:37:39.790 --> 00:37:43.230
extend it and say, look, if you make these other assumptions, it doesn't hold.

00:37:43.390 --> 00:37:49.160
And that I think all goes towards building more robust, more falsifiable

00:37:49.480 --> 00:37:50.760
scientific theory.

00:37:51.320 --> 00:37:56.040
So we build these models and we assume that

00:37:57.000 --> 00:38:01.880
basically people are in a population where there's other people in the population that are

00:38:01.880 --> 00:38:03.800
similar or dissimilar to them.

00:38:04.440 --> 00:38:09.830
And in these first models, we say, well, let's not even worry about what signals they're

00:38:09.830 --> 00:38:11.270
using or how they're signaling.

00:38:11.430 --> 00:38:15.910
Let's just assume that what they're doing is signaling either overtly or covertly.

00:38:16.070 --> 00:38:21.990
And overt signaling has a probability of you know being identified by

00:38:22.630 --> 00:38:24.710
anyone who's around.

00:38:25.350 --> 00:38:27.830
And regardless if they're similar or dissimilar.

00:38:27.920 --> 00:38:33.360
And covert signaling is noisier, so even similar people are less likely to pick up on it.

00:38:33.440 --> 00:38:35.760
Sometimes you miss a joke even when your friend makes it.

00:38:36.000 --> 00:38:40.080
But dissimilar people are really unlikely to get it.

00:38:41.120 --> 00:38:46.560
And then it's this trade-off, right, between the increased

00:38:46.560 --> 00:38:52.480
fidelity of the covert signaling versus the uh avoiding sort of burning bridges with

00:38:52.480 --> 00:38:53.840
dissimilar individuals.

00:38:54.000 --> 00:38:59.890
And then we assume, well, okay, now people pair up to actually work on something

00:38:59.890 --> 00:39:00.610
together.

00:39:00.850 --> 00:39:06.370
And there are factors that influence the payoff that they get, how good that is, how good

00:39:06.370 --> 00:39:07.970
that that working together is.

00:39:08.290 --> 00:39:13.650
Say, well, it's better to work with someone who's similar to you, but it's if you have to

00:39:13.650 --> 00:39:17.570
work with someone who's dissimilar to you, it's better not to be aware of all your

00:39:17.570 --> 00:39:18.210
differences.

00:39:19.170 --> 00:39:19.490
Right?

00:39:19.680 --> 00:39:21.040
So it's better, you know.

00:39:21.200 --> 00:39:23.760
We we all like, you know, you can work with someone.

00:39:24.000 --> 00:39:28.320
Uh it's why, like, you know, we have these these heuristics of like don't talk about

00:39:28.320 --> 00:39:30.880
politics and religion at Thanksgiving or whatever, right?

00:39:31.040 --> 00:39:35.680
It's sometimes better not to not to highlight the differences in order to have a good time.

00:39:35.840 --> 00:39:38.480
And especially if you're trying to work with somebody.

00:39:38.720 --> 00:39:43.290
Um and we can vary things like, well, how bad is it?

00:39:43.450 --> 00:39:48.170
Like, is it a little bad to know you're dissimilar, or is it like really bad, like they

00:39:48.170 --> 00:39:49.050
kill you?

00:39:49.290 --> 00:39:55.290
Um and how often do you have to work with people who are different

00:39:55.290 --> 00:39:55.850
from you?

00:39:56.170 --> 00:40:01.930
Um is it really easy to like if you're if you can always use information that you have,

00:40:02.090 --> 00:40:07.770
your preferences about other people, and only choose people who you like and never have to

00:40:07.770 --> 00:40:12.560
work with people you don't like, yeah, maybe you should always you should signal overtly

00:40:12.560 --> 00:40:13.520
all the time.

00:40:13.920 --> 00:40:18.880
Uh but if sometimes you have to work with people, a fair amount of time you have to work

00:40:18.880 --> 00:40:24.880
with people uh that you don't like, um then yeah, the

00:40:24.880 --> 00:40:28.560
the model sort of predicts that we should expect more covert signaling.

00:40:28.800 --> 00:40:33.730
Now I'll I'll take a pause here, but I'll just say like we've that was sort of the

00:40:33.730 --> 00:40:38.450
groundwork, this kind of modeling with these kinds of qualitative predictions from the

00:40:38.450 --> 00:40:39.010
math.

00:40:39.170 --> 00:40:44.850
And we did simulations later, both where we did an agent-based sort of

00:40:44.850 --> 00:40:49.890
computational model where we could have minority and majority groups to make more explicit

00:40:49.890 --> 00:40:50.930
predictions there.

00:40:51.570 --> 00:40:57.410
We did uh an experiment using um sort of pre-Elon Musk Twitter

00:40:57.650 --> 00:41:03.100
to um to look at how people were actually signaling about their political identities online

00:41:03.260 --> 00:41:08.140
in different contexts, and how people we did in a behavioral experiment where to look at

00:41:08.140 --> 00:41:13.900
like what signals people chose to see if like the predictions of the model actually aligned

00:41:13.900 --> 00:41:17.420
with how people act in the world and and they do.

00:41:18.460 --> 00:41:24.290
And then, you know, later this, you know, the the work was sort of like

00:41:24.290 --> 00:41:30.210
I said, we in these models we said, well, we don't know, let's be agnostic about what

00:41:30.210 --> 00:41:34.850
people what signaling, what signals people are using, how they're using that, and even like

00:41:34.930 --> 00:41:39.250
you know, how they choose their signals, all it is is a strategy of covert signaling or

00:41:39.250 --> 00:41:40.530
overt signaling.

00:41:40.850 --> 00:41:46.720
Um but that led to sort of new ideas about the

00:41:46.720 --> 00:41:52.160
sort of lifespan of signals um and the idea that like

00:41:53.920 --> 00:41:59.680
well, if I'm using, let's say, a uh a r a reference, uh fashion

00:41:59.680 --> 00:42:05.270
choice or uh, you know, uh uh some words that I'm using, a joke that I'm using, and I'm

00:42:05.270 --> 00:42:09.350
doing this in public, like the whole idea is like there are third party listeners, and I'm

00:42:09.350 --> 00:42:12.870
using this because the th you know the people who are not like me don't know this

00:42:12.870 --> 00:42:14.870
reference, and the people who are like me do.

00:42:15.030 --> 00:42:17.110
Well, like people aren't stupid.

00:42:17.430 --> 00:42:23.110
Eventually, if everybody in my in-group is using these covert signals, people are

00:42:23.110 --> 00:42:26.710
gonna people from the out groups are gonna wise up, they're gonna notice.

00:42:26.950 --> 00:42:28.390
Like we notice all the time, right?

00:42:28.550 --> 00:42:31.720
We this is something that's like we call out the dog whistles.

00:42:31.880 --> 00:42:36.440
We call out the dog whistles, usually often years after they were effective.

00:42:36.680 --> 00:42:37.080
Yeah.

00:42:37.240 --> 00:42:37.800
Yep.

00:42:38.360 --> 00:42:40.760
And that's sort of the nature of it.

00:42:40.840 --> 00:42:43.720
So something that used to be covert is now overt.

00:42:44.360 --> 00:42:49.640
And uh we became interested in what

00:42:50.280 --> 00:42:56.110
are the dynamics of like does that imply that covert signals will have like a a

00:42:56.110 --> 00:42:57.790
natural lifespan.

00:42:58.030 --> 00:42:59.390
How long is that lifespan?

00:42:59.550 --> 00:43:01.710
What would determine that lifespan?

00:43:01.870 --> 00:43:07.710
Um and how exactly how long the the lifespan is, um we

00:43:07.710 --> 00:43:08.430
don't know.

00:43:08.750 --> 00:43:13.550
But the things that determine the lifespan now we do know from from you know, we we built a

00:43:13.550 --> 00:43:17.950
completely different kind of computational model and also a math model um to do this.

00:43:18.110 --> 00:43:22.110
Covert signaling, I think also this is one of the I'll I'll pause for a second.

00:43:22.270 --> 00:43:28.040
I I keep saying I'll pause and then not, but um one of the this is one of the

00:43:28.040 --> 00:43:29.480
additional benefits of modeling.

00:43:29.640 --> 00:43:34.360
It's not just that these things are hard to study because we're talking about timelines

00:43:34.520 --> 00:43:39.800
that are hard to study or and naturalistic interactions that are hard to study, but also

00:43:39.800 --> 00:43:45.080
because by definition, I'm talking about a a signal, a piece of

00:43:45.080 --> 00:43:51.000
communication the that you have to already be in the in-group to be able

00:43:51.000 --> 00:43:52.280
to identify.

00:43:53.640 --> 00:43:58.190
And that's really hard to study empirically.

00:43:58.430 --> 00:44:04.030
Um we we came up with some some workarounds uh for a couple of different um

00:44:04.270 --> 00:44:09.150
empirical studies, but uh they're limited, right?

00:44:09.470 --> 00:44:15.310
Um and and the the modeling really drove what we were looking for and helped us sort of see

00:44:15.310 --> 00:44:17.230
what kind of thing we were looking for.

00:44:18.240 --> 00:44:23.920
Yeah, if it's designed to not be detected uh by the experimenter, then that makes the

00:44:23.920 --> 00:44:26.880
experimenter's job uh extremely difficult.

00:44:27.680 --> 00:44:28.640
It does, yeah.

00:44:28.800 --> 00:44:34.720
So when we did this um uh with uh with Twitter data, and this is

00:44:34.720 --> 00:44:40.480
uh data collected in 2020, um we we specifically

00:44:40.480 --> 00:44:46.310
used you know tweets collected in the like weeks right before the 2020

00:44:46.310 --> 00:44:51.910
presidential election in the US we only used US participants and US based accounts

00:44:52.230 --> 00:44:58.230
and we figured well you know political identity was really salient I don't know if you guys

00:44:58.230 --> 00:45:01.350
remember 2020 is a really fun year.

00:45:01.590 --> 00:45:07.270
Um the uh it was like you know um

00:45:10.100 --> 00:45:13.300
It was uh wait, was this 2020 or 2016?

00:45:13.380 --> 00:45:14.660
Now I can't remember.

00:45:14.900 --> 00:45:20.820
Um either way, it was like, you know, this we're talking about like right

00:45:20.820 --> 00:45:24.740
before, you know, the the election, which was very heated.

00:45:24.900 --> 00:45:30.740
It was right um around when Amy Coney Barrett had been uh approved to the United States

00:45:30.740 --> 00:45:31.540
Supreme Court.

00:45:31.700 --> 00:45:34.420
It was just months after George Floyd was murdered.

00:45:34.580 --> 00:45:37.700
So like political identity was super salient.

00:45:38.180 --> 00:45:43.810
And what we did was collect a bunch of tweets based on

00:45:44.770 --> 00:45:50.130
accounts that were following a fair number of either far right or far-left news sites.

00:45:50.370 --> 00:45:54.290
So we're like, these people are probably going to be tweeting about politics.

00:45:55.010 --> 00:46:00.130
And then what we did was we got

00:46:00.370 --> 00:46:06.060
participants recruited through MTurk at the time, And we gave them a

00:46:06.060 --> 00:46:10.780
survey to determine their sort of where they fell on the political spectrum.

00:46:10.860 --> 00:46:14.060
Not just left and right, but how far left and how far right.

00:46:14.140 --> 00:46:19.820
So we could we sort of group people into four categories: radical right,

00:46:20.060 --> 00:46:23.340
moderate right, radical left, moderate left.

00:46:23.820 --> 00:46:29.470
And We then sort of gave people just lots and lots of tweets.

00:46:29.630 --> 00:46:33.710
And we said, all right, like, what do you think about this tweet?

00:46:33.790 --> 00:46:37.710
And, you know, more importantly, like, what can you tell about the likely political

00:46:37.710 --> 00:46:41.310
identity of this of the person who tweeted this just based on the text?

00:46:41.390 --> 00:46:45.070
And we we, you know, blanked out the account name and the picture.

00:46:45.150 --> 00:46:47.470
So all they saw was the text of the tweet.

00:46:48.110 --> 00:46:52.300
And uh, you know, sometimes it's really obvious, right?

00:46:52.460 --> 00:46:57.260
Like if someone was just like, you know, fuck Trump, like that's probably not someone on

00:46:57.260 --> 00:46:57.900
the right.

00:46:58.220 --> 00:47:03.340
Uh whereas, you know, there are things, other things that were a lot more subtle, like

00:47:03.340 --> 00:47:09.340
references to laws or, you know, sort of dog whistles that people use internally

00:47:09.340 --> 00:47:13.020
with an in-group, but don't really use with the out-group.

00:47:13.830 --> 00:47:19.590
And we found that there were a number of tweets that reliably

00:47:19.910 --> 00:47:25.910
people on the same side of the spectrum as the the tweeter on one side of the spectrum

00:47:26.070 --> 00:47:27.990
were like, Yeah, yeah, yeah, I know what this is.

00:47:28.070 --> 00:47:28.470
I like this.

00:47:28.550 --> 00:47:29.990
I'm I'm in favor of this.

00:47:30.230 --> 00:47:34.950
Whereas people on the other side, especially the moderates who are the majority, didn't

00:47:34.950 --> 00:47:35.190
get.

00:47:35.780 --> 00:47:40.260
So we had to sort of uh adapt the theory a little bit that we hadn't really thought about

00:47:40.260 --> 00:47:42.580
radicals and moderates in the modeling.

00:47:42.980 --> 00:47:45.220
But empirically, like it makes a lot of sense.

00:47:45.300 --> 00:47:50.900
Like the people who are really, you know, on like very left or very right,

00:47:51.220 --> 00:47:56.740
they they tend to also be sort of higher in their aggression and enmity for the other side.

00:47:56.820 --> 00:48:00.500
And they tend to be more savvy about the signals used by the other side.

00:48:02.350 --> 00:48:03.310
And yeah.

00:48:03.630 --> 00:48:07.470
They have more of an incentive to detect the dog whistles than say a moderate would.

00:48:07.790 --> 00:48:08.430
Exactly.

00:48:08.670 --> 00:48:12.510
Um and I and I wanna point out like they are in the minority though.

00:48:12.590 --> 00:48:17.070
Like it way more people are moderates than are radical.

00:48:19.310 --> 00:48:24.920
Is it that um the m the the radical were were better at detecting their

00:48:25.720 --> 00:48:31.000
same side dog whistles and the other side dog whistles, or what was the result?

00:48:31.720 --> 00:48:37.480
So basically, if somebody is signaling, if somebody is on the left and

00:48:37.640 --> 00:48:43.640
sending a signal that like communicates a left sort of you know biased message,

00:48:43.960 --> 00:48:47.720
everybody on the left understands that signal.

00:48:48.440 --> 00:48:54.310
If they're doing it covertly, The radicals on the right will also

00:48:54.630 --> 00:48:56.870
be fairly likely to notice it.

00:48:57.030 --> 00:48:58.310
But the moderates on the right won't.

00:48:58.630 --> 00:48:59.190
Yeah, got it.

00:48:59.270 --> 00:49:03.110
They're like higher information and even higher information about the other side in other

00:49:03.190 --> 00:49:03.350
words.

00:49:03.750 --> 00:49:03.990
That's right.

00:49:04.070 --> 00:49:04.230
Cool.

00:49:04.310 --> 00:49:04.790
Okay, yeah.

00:49:05.190 --> 00:49:08.470
So we basically use this difference between the moderates.

00:49:08.630 --> 00:49:13.350
So anything where like moderates on both sides could immediately identify as a political

00:49:13.350 --> 00:49:16.340
signal, we were like, that's an overt message.

00:49:16.500 --> 00:49:21.940
And things that everyone on one side, but only the radicals on the other side recognized

00:49:22.260 --> 00:49:24.180
were like, that's probably covert.

00:49:25.460 --> 00:49:31.380
And um in the, you know, so we did two things there, right?

00:49:31.460 --> 00:49:35.780
In the and this is all for sort of predictions that the whole design of this came out of

00:49:35.780 --> 00:49:37.300
the modeling that we had done.

00:49:38.100 --> 00:49:43.970
Um but we first just looked at like what does it look like on

00:49:43.970 --> 00:49:44.610
Twitter?

00:49:44.850 --> 00:49:48.690
And again, this is before Musk took it over, so it was Twitter, not X.

00:49:49.170 --> 00:49:54.610
Um, we just looked at the follower networks and sort of like who tweeted this?

00:49:55.570 --> 00:50:00.690
Who's what are let's we looked at all their followers and the kinds of news sites that they

00:50:00.850 --> 00:50:01.490
followed?

00:50:01.700 --> 00:50:07.540
And we looked at, you know, basically how partisan they were based on how mixed the the

00:50:07.540 --> 00:50:09.060
partisan sources were.

00:50:09.220 --> 00:50:11.780
So if you only followed right wing sources, you were right wing.

00:50:11.860 --> 00:50:14.900
If you only follow left wing sources, you were right left wing.

00:50:15.460 --> 00:50:20.740
If you followed both, you were more in the middle, and or at least more um

00:50:22.740 --> 00:50:24.180
harder to tell.

00:50:24.500 --> 00:50:30.350
And um and what we found is

00:50:30.350 --> 00:50:36.270
like the you know, the people who tweeted more overtly, most of their followers

00:50:36.270 --> 00:50:39.630
tended to be partisans on the same side of the spectrum.

00:50:39.710 --> 00:50:45.230
The people whose tweets were more covert tended to have more followers from both

00:50:45.230 --> 00:50:45.790
sides.

00:50:46.270 --> 00:50:52.240
And this is so this is um so r say radicals uh on either side.

00:50:52.400 --> 00:50:58.080
Um radicals with politically homogenous audiences um

00:50:58.720 --> 00:51:04.160
were more likely to signal overtly and radicals with uh more

00:51:04.240 --> 00:51:09.280
distributed or politically distributed audiences were more likely to signal covertly is

00:51:09.280 --> 00:51:15.120
that not exactly so we didn't we didn't re- we so the radical versus moderate

00:51:15.250 --> 00:51:19.250
those were our raiders, the people who were raiding tweets after we'd collected them,

00:51:19.330 --> 00:51:22.690
because we could actually ask them all these questions about their political identity.

00:51:22.850 --> 00:51:23.090
Got it.

00:51:23.570 --> 00:51:27.410
People who are actually producing the tweets, the only things we know about them are what

00:51:27.410 --> 00:51:30.050
news news accounts they follow.

00:51:30.370 --> 00:51:30.690
Got it.

00:51:31.010 --> 00:51:35.090
We're not really able to say that much about how radical they are, but we do know that

00:51:35.490 --> 00:51:40.370
they're probably at least fairly politically engaged because they're following lots of news

00:51:40.370 --> 00:51:43.540
sites and are tweeting about political things.

00:51:43.860 --> 00:51:44.100
Got it.

00:51:44.260 --> 00:51:47.940
And were the participants rating the radicalness or just the left-rightness?

00:51:48.740 --> 00:51:51.700
Um, so let me let me back up.

00:51:51.860 --> 00:51:52.020
Right.

00:51:52.180 --> 00:51:57.780
So we actually the first thing I did, right, or that we did was uh

00:52:00.980 --> 00:52:05.780
do some just ratings of tweets and ratings of so we we got people

00:52:06.790 --> 00:52:10.550
to ask, you know, answer questions just about their own political views, and that's how we

00:52:10.550 --> 00:52:11.830
could tell who they were.

00:52:11.990 --> 00:52:15.830
And then we asked them just to sort of rate these tweets that we had we had gotten.

00:52:15.990 --> 00:52:21.910
And we use that to sort of as a basically a metric, as an instrument to

00:52:21.910 --> 00:52:26.630
determine if the messages were being sent covertly or overtly.

00:52:29.180 --> 00:52:35.180
And when we just looked at who's sending what kinds of messages online and who's

00:52:35.180 --> 00:52:41.100
following them, we looked at not so the people who are sending, let's say,

00:52:41.340 --> 00:52:45.980
right wing tweets that are right wing, they're following lots of right-wing news sources.

00:52:46.140 --> 00:52:47.340
Who's following them?

00:52:47.580 --> 00:52:51.180
And what news sites are they of those people following?

00:52:53.070 --> 00:52:58.990
If that person who's following lots of right wing sites is tweeting out very overtly right

00:52:59.070 --> 00:53:04.430
wing messages, that correlates with most of their followers also

00:53:04.750 --> 00:53:08.270
being people who mostly follow right wing sites or sources.

00:53:09.070 --> 00:53:14.030
If they're tweeting more covertly, that means that a larger proportion of their audience,

00:53:14.190 --> 00:53:19.880
their followers, are following uh left wing sites, whether they're

00:53:19.880 --> 00:53:22.280
following both right or left or just left.

00:53:22.520 --> 00:53:23.160
Got it.

00:53:23.560 --> 00:53:29.400
Um, that signal was that effect was you know statistically significant and it was as

00:53:29.400 --> 00:53:34.440
they say, but it wasn't huge because Twitter is already fairly homophilus, like there's

00:53:34.440 --> 00:53:36.440
already a fair amount of assortment on Twitter.

00:53:36.520 --> 00:53:36.760
Yeah.

00:53:37.000 --> 00:53:42.440
So to boost the signal for this effect to test this, we did a behavioral experiment

00:53:43.080 --> 00:53:48.890
where we collected sets of tweets that are categorized as either left

00:53:48.890 --> 00:53:50.810
or right, overt or covert.

00:53:51.130 --> 00:53:53.210
So we have four groups of tweets.

00:53:53.930 --> 00:53:59.770
And then we we bring a participant in, now it's a new set of people, we say, hey,

00:54:00.730 --> 00:54:06.650
um you're gonna rate a bunch of these tweets, just like you know we

00:54:06.650 --> 00:54:07.370
did before.

00:54:07.450 --> 00:54:13.060
Uh uh and mostly we're just looking like, You know, we asked a few questions about them,

00:54:13.140 --> 00:54:17.540
like are they political, whatever, but also like once we determined that they were

00:54:17.540 --> 00:54:19.780
political, do you like it or dislike it?

00:54:19.940 --> 00:54:24.980
Do you have positive or negative feelings toward the person who sent this?

00:54:25.860 --> 00:54:28.500
And so everyone did that.

00:54:28.580 --> 00:54:34.070
And then we said, now, now you're gonna do an experiment, you're gonna have a

00:54:34.070 --> 00:54:40.070
task where we're gonna show you 40 tweets that are all, and

00:54:40.070 --> 00:54:45.830
we didn't tell them any of anything about the tweets, but actually they were

00:54:45.830 --> 00:54:51.670
all from people from the same side of the political spectrum as them, and half of them were

00:54:51.670 --> 00:54:53.190
covert and half of them were overt.

00:54:53.430 --> 00:54:54.870
They didn't know which was which.

00:54:55.110 --> 00:54:59.830
We just said, you're just gonna pick this, and we're just gonna and um

00:55:00.560 --> 00:55:06.560
there's gonna be a random set of 10 other people who rated who

00:55:06.560 --> 00:55:09.280
are gonna rate these tweets just like you rated them.

00:55:10.080 --> 00:55:15.920
And for every tweet that you select that most of your audience liked

00:55:16.240 --> 00:55:21.360
more than they disliked, you're gonna get more money than we say we were originally gonna

00:55:21.360 --> 00:55:21.840
give you.

00:55:22.720 --> 00:55:28.710
But for every tweet you pick that in this audience that you if most of them

00:55:28.790 --> 00:55:34.070
or or a majority of them dislike it, you're gonna lose money.

00:55:34.950 --> 00:55:36.790
And so you be careful.

00:55:36.870 --> 00:55:39.190
But you can select as many or as few as you want.

00:55:39.510 --> 00:55:43.590
And we're like, there's a minimum amount you can get, but let's say, you know, I don't

00:55:43.590 --> 00:55:44.150
remember what it was.

00:55:44.230 --> 00:55:45.670
It was like five dollars.

00:55:45.940 --> 00:55:50.100
If all of them hate it, if you pick all the tweets and all of your audience hates it, you

00:55:50.100 --> 00:55:51.380
might only get two dollars.

00:55:51.540 --> 00:55:54.500
But maybe if you go up, you can go like as many as eight dollars.

00:55:54.580 --> 00:55:56.500
I don't remember the exact numbers.

00:55:57.300 --> 00:56:03.060
And uh and all we told them about their audience was how many of the ten

00:56:03.380 --> 00:56:06.420
were on the same side of the political spectrum as them.

00:56:07.940 --> 00:56:10.660
And you know, unsurprisingly, right?

00:56:10.740 --> 00:56:15.830
People are real this is like in in hindsight fairly obvious, but like, you know, the it was

00:56:15.830 --> 00:56:17.030
sort of a proof of concept.

00:56:17.110 --> 00:56:18.710
People are really good at this.

00:56:19.030 --> 00:56:24.870
Um when people when when people knew that the majority of their

00:56:24.870 --> 00:56:29.350
audience was on the same side as them, they picked almost everything.

00:56:29.510 --> 00:56:33.030
They were like, all these, I agree with all these these messages.

00:56:33.190 --> 00:56:35.510
I want to send them all to this audience.

00:56:35.750 --> 00:56:41.740
When people knew that the audience was made up of a fair number, like a mix, they

00:56:41.740 --> 00:56:42.940
were more careful.

00:56:43.180 --> 00:56:47.580
And when they knew that a lot of the audience, a majority of their audience was different

00:56:47.580 --> 00:56:50.380
from them, they were even more careful in two ways.

00:56:50.540 --> 00:56:54.140
They just selected way fewer tweets overall.

00:56:54.300 --> 00:56:57.900
And of the ones they selected, a higher percentage of them were covert.

00:57:00.620 --> 00:57:04.810
And so like, it's a fairly

00:57:05.610 --> 00:57:11.450
artificial construct construction, but the idea was to sort of show that we could

00:57:11.610 --> 00:57:15.530
that people actually do this in the real world, make these strategic considerations in the

00:57:15.690 --> 00:57:18.490
real world that align with the kinds of modeling that we did.

00:57:18.570 --> 00:57:23.930
And I think the conclusions that we make from the model are broader than the ones we can

00:57:23.930 --> 00:57:24.890
make from the experiment.

00:57:25.050 --> 00:57:30.900
But the experiment is just an illustration that, yeah, people can behave in

00:57:30.900 --> 00:57:33.540
this way that we predicted that they be would behave.

00:57:34.660 --> 00:57:37.780
I see an interesting uh extension of this.

00:57:38.020 --> 00:57:42.980
Um so it it would be interesting to look at people who are on the same side politically,

00:57:43.060 --> 00:57:48.500
say say left-wing people, um, and have them evaluate um

00:57:48.980 --> 00:57:54.790
the political uh allegiances of of the tweeter uh who is either tweeting a

00:57:54.790 --> 00:57:59.510
covert signal or an overt signal, and whether they think the people who are tweeting uh

00:57:59.670 --> 00:58:05.270
overt signals are more radical or more extreme uh than the people tweeting covert

00:58:05.270 --> 00:58:10.310
signals, even though they know e even in cases where they know the the the the meaning of

00:58:10.310 --> 00:58:11.830
the covert signal.

00:58:12.140 --> 00:58:17.580
And so the idea there would be that if I'm sending an overt signal, I'm sending a costly

00:58:17.580 --> 00:58:23.340
signal of my loyalty to uh to leftists because I'm I am burning bridges

00:58:23.660 --> 00:58:26.860
with uh moderates and with right wing people.

00:58:27.100 --> 00:58:31.900
Whereas if I have a more mixed social network, I can't afford to burn those bridges.

00:58:32.060 --> 00:58:37.340
Uh and so I'm going to be uh uh unlikely to be as committed

00:58:37.340 --> 00:58:38.140
politically.

00:58:38.790 --> 00:58:40.150
I think that's exactly right.

00:58:40.230 --> 00:58:45.270
And I this is actually, yeah, what you're talking about were uh is related to a lot of the

00:58:45.270 --> 00:58:49.110
conversations we were having as we first you know started doing this kind of work.

00:58:49.270 --> 00:58:54.870
And w one of the the things that really kind of inspired uh, especially the this experiment

00:58:54.870 --> 00:59:00.790
that I just talked about, was um uh I, you know, I had done the modeling and and and I

00:59:00.790 --> 00:59:06.660
just kept thinking about, you know, uh I think the first paper was published in 2018,

00:59:06.820 --> 00:59:08.660
but it was it was written.

00:59:08.900 --> 00:59:14.180
The first draft of that paper was was finished, you know, I think uh sometime in

00:59:14.180 --> 00:59:15.140
2015.

00:59:15.860 --> 00:59:21.540
And, you know, something happened around 2015, 2016 in the

00:59:21.540 --> 00:59:22.820
US, especially, right?

00:59:22.900 --> 00:59:27.380
All of a sudden it just felt like, and you know, we've been seeing it happen more and more.

00:59:27.460 --> 00:59:32.850
Like all these kind of Extreme white

00:59:33.010 --> 00:59:38.930
national racist, you know, fascist sympathizers just

00:59:38.930 --> 00:59:40.530
started coming out of the woodwork.

00:59:40.690 --> 00:59:44.610
Like all of a sudden, they just were there were more and more of them.

00:59:44.690 --> 00:59:49.970
Like people were on Stormfront, uh, which is like a right-wing kind of fascist uh

00:59:50.130 --> 00:59:51.970
Nazi website.

00:59:52.130 --> 00:59:57.250
And they were just we had that march in um Charlottesville.

00:59:57.810 --> 01:00:03.780
Um and it became very clear, like, you

01:00:03.780 --> 01:00:05.300
know, why is this happening?

01:00:05.460 --> 01:00:07.620
Or the you know, question arose, like, why is this happening?

01:00:07.780 --> 01:00:11.620
And and two possibilities sort of immediately present themselves.

01:00:11.780 --> 01:00:17.300
Like, one, people have been recruited to this cause that weren't previously of this

01:00:17.300 --> 01:00:17.780
mindset.

01:00:17.940 --> 01:00:22.500
But all of a sudden, this has become much more attractive uh for people, and people uh it's

01:00:22.500 --> 01:00:26.500
much more compelling, and a lot of people who previously didn't think this way now think

01:00:26.500 --> 01:00:27.140
this way.

01:00:28.070 --> 01:00:33.910
Another hypothesis, though, is that people there were there were there

01:00:33.910 --> 01:00:38.870
had been for quite some time a lot of people who felt this way, but were much more quiet

01:00:38.870 --> 01:00:39.510
about it.

01:00:39.910 --> 01:00:44.550
That maybe they talked this way in their insular social networks, but they would never talk

01:00:44.550 --> 01:00:49.670
this way among you know in public because of the negative consequences.

01:00:50.500 --> 01:00:56.420
And between the kind of assortment that the internet and social media facilitate,

01:00:56.740 --> 01:01:02.340
so it becomes much more easy to find networks of like-minded people online,

01:01:02.740 --> 01:01:06.100
and that that has good things and bad things about it.

01:01:06.900 --> 01:01:12.810
And you know, the rise of a political leader, specifically

01:01:12.810 --> 01:01:18.730
Donald Trump and others, who seemed ideologically aligned

01:01:18.730 --> 01:01:23.210
with these, sort of emboldened people and said, I think their perceptions of both their

01:01:23.210 --> 01:01:28.330
social networks and the potential payoffs, costs, and benefits of signaling

01:01:28.650 --> 01:01:29.450
changed.

01:01:29.690 --> 01:01:35.530
And so they went from signaling either only in private or covertly in public,

01:01:36.740 --> 01:01:40.100
to being much more emboldened to signal c uh overtly.

01:01:40.180 --> 01:01:42.820
And that was really where where the idea came from.

01:01:42.980 --> 01:01:48.580
And and the initial uh you know uh plan to do this study was much more ambitious, but it

01:01:48.580 --> 01:01:49.540
turned out to be really hard.

01:01:49.620 --> 01:01:52.180
So we we did the smaller version of it.

01:01:52.660 --> 01:01:53.860
Um right.

01:01:54.180 --> 01:01:58.370
I mean it's occurring to me that another interesting extension of this would be to look at

01:01:58.370 --> 01:02:00.930
perceptions of group formidability.

01:02:01.090 --> 01:02:05.250
And you might expect that groups that are signaling overtly will be perceived as more

01:02:05.250 --> 01:02:10.450
formidable or more of a political threat, uh, and perhaps even you know larger in number

01:02:10.610 --> 01:02:12.690
than groups that are signaling covertly.

01:02:12.850 --> 01:02:17.730
Because if I have a huge social network of like-minded peers, I can afford to burn bridges

01:02:17.730 --> 01:02:18.690
with everyone else.

01:02:18.850 --> 01:02:24.280
Uh, I can afford to signal overtly, but if I'm a tiny minority, um, let's say you know I'm

01:02:24.280 --> 01:02:28.760
a I'm a minority of dissidents in a totalitarian regime, I have to signal very, very, very

01:02:28.760 --> 01:02:33.560
covertly because if I get detected, you know, off to the gulags I go, right?

01:02:33.800 --> 01:02:39.640
But if I'm in the majority, uh, I can afford to signal as overtly as I want uh without care

01:02:39.640 --> 01:02:43.880
for you know burning bridges because there are so few bridges to burn or they're not very

01:02:43.880 --> 01:02:45.560
valuable bridges in the first place.

01:02:45.800 --> 01:02:49.720
And so it might be that if I'm looking, even if I know with certainty the political

01:02:49.720 --> 01:02:55.010
allegiances of the of the person, um, You know, if they're s if they're signaling overtly,

01:02:55.090 --> 01:02:59.810
I'm going to be more threatened and estimate that they uh have a more homogenous social

01:02:59.810 --> 01:03:03.970
network, they have uh a larger social network of like-minded peers than if they're

01:03:03.970 --> 01:03:05.570
signaling covertly.

01:03:06.130 --> 01:03:06.770
That's right.

01:03:06.930 --> 01:03:08.530
Uh yeah, that's exactly right.

01:03:08.850 --> 01:03:12.210
And would it also, I mean, the other question would be the uh same side.

01:03:12.370 --> 01:03:18.370
So if I'm uh a persecuted minority, for example, and I see somebody else explicitly,

01:03:18.610 --> 01:03:24.400
overtly signaling, that should be minimally interesting to me and

01:03:24.400 --> 01:03:29.200
maybe a little bit positive, if not really positive, because it's doing two things.

01:03:29.360 --> 01:03:32.080
One is it's sort of they're the canary in the coal mine.

01:03:32.320 --> 01:03:34.000
So are they going to go to the gulag or not?

01:03:34.160 --> 01:03:38.800
Because it tells me, you know, how many of their how many other people like me are there.

01:03:38.960 --> 01:03:44.770
Um, but then it also is maybe my group is more formidable than I thought.

01:03:44.850 --> 01:03:47.810
And maybe I don't have to uh signal covertly.

01:03:48.050 --> 01:03:53.890
So you can imagine that there's also an appetite for overt signaling of your own side

01:03:54.130 --> 01:03:58.770
when you're experiencing at some level in your brain that you're having to do it covertly

01:03:58.770 --> 01:03:59.650
for the time being.

01:04:00.130 --> 01:04:01.010
Yeah, that's right.

01:04:01.090 --> 01:04:05.330
And this is actually something I'm super interested in because I think it's really hard to

01:04:05.330 --> 01:04:10.690
explain um from a kind of strategic perspective.

01:04:10.930 --> 01:04:14.300
But I think uh which is the sort of like

01:04:16.940 --> 01:04:22.300
you know, especially in political situations when dissidents kind of go above and beyond,

01:04:22.460 --> 01:04:23.900
like they're oppressed.

01:04:23.980 --> 01:04:28.620
And if they signal in a particular way, they're gonna be oppressed more, but they choose to

01:04:28.620 --> 01:04:30.860
do it anyway in a kind of act of science.

01:04:30.940 --> 01:04:35.580
I find this really interesting um and inspiring in a lot of ways.

01:04:35.900 --> 01:04:41.890
And, you know, I think I can we can make kind of gesso stories about

01:04:41.970 --> 01:04:47.810
or you know uh uh speculate about that a lot of these people, and I think the evidence at

01:04:47.810 --> 01:04:53.570
least partially supports this, that like a lot of these people are like the ones who

01:04:53.570 --> 01:04:57.970
feel like there is no hope in the mainstream system.

01:04:58.130 --> 01:05:04.100
There's just um uh Benwat de Corson and I think is

01:05:04.260 --> 01:05:10.180
Dan Nettle the other author, they have these papers uh on um you

01:05:10.180 --> 01:05:15.700
know the the on on risk and uh

01:05:16.580 --> 01:05:20.660
and there are other people who've who've written about this too, but basically like risk

01:05:20.660 --> 01:05:21.460
and thresholds.

01:05:21.700 --> 01:05:22.180
Yeah, yeah.

01:05:22.500 --> 01:05:27.860
And uh, you know, how like people who are sort of just above the threshold where like you

01:05:27.860 --> 01:05:31.070
have everything things are going okay for you as long as you don't rock the boat.

01:05:31.150 --> 01:05:33.550
Like you'll be fine as long as you don't rock the boat.

01:05:33.790 --> 01:05:38.110
Those people are incentivized to not take any risks, to be very careful.

01:05:38.350 --> 01:05:42.990
But the people who've fallen below that threshold who are basically fucked either way.

01:05:43.230 --> 01:05:47.870
Like if they keep going on the way they've always been going, they've got nothing anyway.

01:05:47.950 --> 01:05:50.910
So they might as well try and take a chance.

01:05:51.470 --> 01:05:56.760
And I think one of the things that's also underappreciated is like those people can form

01:05:56.760 --> 01:06:02.680
coalitions and raise themselves up, rise them uh raise themselves up out of that

01:06:02.760 --> 01:06:08.520
that gulf explicitly by taking these risks and by by sort

01:06:08.520 --> 01:06:10.280
of going against the mainstream.

01:06:10.360 --> 01:06:14.760
Um one of my favorite books that I read in the last uh number of years was this book called

01:06:14.760 --> 01:06:16.680
um burning down the house.

01:06:16.920 --> 01:06:18.050
Do you know this Dave?

01:06:18.370 --> 01:06:24.290
I know the song, but uh yeah, not not the um not the talking head song, which is great, but

01:06:24.290 --> 01:06:27.410
it's uh house spelled in the German way, H-A-U-S.

01:06:27.730 --> 01:06:28.290
Oh nice.

01:06:28.450 --> 01:06:32.690
Uh and the the subtitle is Punk Rock Revolution and the Fall of the Berlin Wall.

01:06:32.930 --> 01:06:33.410
Oh, nice.

01:06:33.490 --> 01:06:34.290
I shouldn't read that.

01:06:34.370 --> 01:06:34.450
Yeah.

01:06:34.770 --> 01:06:40.160
It's written by this journalist, Tim Moore, and it's about basically punk rockers in East

01:06:40.160 --> 01:06:46.000
Germany in the nineteen eighties and kind of highlighting these people like it just

01:06:46.000 --> 01:06:50.960
makes like the sort of American punk scene around the same time feel just completely uh

01:06:51.120 --> 01:06:53.600
like a bunch of posers because suburban.

01:06:53.680 --> 01:06:53.760
Yeah.

01:06:54.000 --> 01:06:55.520
Yeah, it's totally suburban, right?

01:06:55.600 --> 01:06:57.920
It was uh like no risk.

01:06:58.000 --> 01:06:58.320
No risk.

01:06:58.640 --> 01:07:03.520
German punks were being like arrested uh by the Stasi and thrown in jail.

01:07:03.680 --> 01:07:08.730
And then and yeah, getting getting beaten up, getting literally thrown in prison, not being

01:07:08.730 --> 01:07:09.690
able to get jobs.

01:07:09.850 --> 01:07:15.690
They the bands weren't actually allowed to play in public because you needed a permit given

01:07:15.690 --> 01:07:17.610
by the government to play punk rock.

01:07:17.690 --> 01:07:20.970
So you had all the shows were underground or in churches.

01:07:21.370 --> 01:07:27.290
Um but uh m Tim Moore argues that basically, you know, there was an effect of

01:07:27.450 --> 01:07:32.740
of these people sort of actively resisting and you can't help but notice them on the

01:07:32.740 --> 01:07:33.140
street.

01:07:33.220 --> 01:07:38.180
And when you see them continuously getting harassed and oppressed, uh despite not having

01:07:38.180 --> 01:07:44.180
done anything, uh that this he argues that uh this helped sort of turn the tide

01:07:44.180 --> 01:07:49.780
against the GDP uh or the DDR, sorry, um in uh

01:07:50.020 --> 01:07:51.070
in East Germany.

01:07:51.310 --> 01:07:52.030
Yeah, that's awesome.

01:07:52.190 --> 01:07:56.190
It also reminds me of Michael Monkrief we had on, and he was talking about um uh drag

01:07:56.190 --> 01:08:01.790
queens, and it was a similar thing where it was, you know, 1950s, and it was these guys who

01:08:01.790 --> 01:08:03.630
were down and out, low status.

01:08:03.790 --> 01:08:09.230
So it was exactly this, they're kind of screwed either way, and they um they they they're

01:08:09.230 --> 01:08:14.910
using this high variance strategy um to help their own status, but then they're

01:08:14.910 --> 01:08:20.900
doing this work to push on and provoke, you know, for for you know,

01:08:21.140 --> 01:08:22.020
reaction basically.

01:08:22.100 --> 01:08:22.660
And it was David.

01:08:22.900 --> 01:08:23.460
No, that's right.

01:08:23.780 --> 01:08:28.020
If there's um do you ever read there's uh Hunter Thompson's first book, Hell's Angels.

01:08:28.580 --> 01:08:29.380
Yeah, that's a great book.

01:08:29.460 --> 01:08:31.460
I love great author.

01:08:31.620 --> 01:08:36.420
Good if you want to be a better uh author or read writer, um, read read good right writers,

01:08:36.500 --> 01:08:37.060
and Hunter S.

01:08:37.220 --> 01:08:37.860
Thompson's great.

01:08:37.940 --> 01:08:38.100
Yeah.

01:08:38.340 --> 01:08:39.300
Yeah, agreed.

01:08:39.540 --> 01:08:42.740
Um, yeah, Hell's Angels uh is is really great.

01:08:42.820 --> 01:08:47.950
I mean, he talks about them too in that book, like you know, not ro in this kind of

01:08:47.950 --> 01:08:50.270
romantic way, but he's like, these are the the dregs.

01:08:50.510 --> 01:08:56.110
This is the love the people who have nothing in society who are uh and they've banded

01:08:56.110 --> 01:08:58.670
together to to sort of form this outlaw group.

01:08:58.910 --> 01:09:04.030
And um this actually this kind of thinking has has gotten I've become more and more

01:09:04.030 --> 01:09:10.030
interested in this kind of dynamic between mainstream society and

01:09:11.150 --> 01:09:13.630
groups, you know, kind of at the edges.

01:09:14.110 --> 01:09:17.900
Um Yeah, it's super cool.

01:09:18.140 --> 01:09:22.620
I think one of the one of the things I want to ask, speaking of groups at the edges, is

01:09:22.620 --> 01:09:23.180
science.

01:09:23.340 --> 01:09:28.460
Um and um you have this paper with Richard uh

01:09:28.540 --> 01:09:31.980
McAlbright about the natural selection of bad science.

01:09:32.140 --> 01:09:35.820
Um, and I and there's two things I just want to preface for the audience to say.

01:09:35.980 --> 01:09:40.970
One is that everything you've been talking about, I think it's really important, and I want

01:09:40.970 --> 01:09:46.570
to be explicit about is that I think the work you're doing is part of a

01:09:46.570 --> 01:09:51.530
movement to recognize that there are unintended consequences.

01:09:51.770 --> 01:09:56.410
So I think like a simple model of the world is sort of like what people intend is what

01:09:56.410 --> 01:09:57.050
happens.

01:09:57.290 --> 01:10:02.730
But I think the reason to do modeling or social study social dynamics and

01:10:03.370 --> 01:10:08.440
is to is to say, look, selection, what you select for

01:10:09.240 --> 01:10:10.520
is what you get.

01:10:11.480 --> 01:10:13.240
Intentions be damned.

01:10:13.480 --> 01:10:16.520
And so I think that's really important.

01:10:16.680 --> 01:10:21.320
And I think that in the same way this applies to social identities, you've done work, kind

01:10:21.320 --> 01:10:27.240
of meta science work, about how do intentions versus selection,

01:10:27.400 --> 01:10:30.760
so like what are we selecting for, for example, when we do science?

01:10:31.000 --> 01:10:36.690
And might it be that there's unintended consequences that even things that seem good

01:10:37.010 --> 01:10:42.850
or nobody's with malicious intent or everybody's just sort of trying to get

01:10:42.850 --> 01:10:47.730
by in the in the aggregate produce these bad

01:10:48.050 --> 01:10:49.170
dynamics?

01:10:49.410 --> 01:10:51.330
I think that idea is really powerful.

01:10:51.410 --> 01:10:57.400
And I don't think outside of maybe the modeling work or the modeling world, I

01:10:57.400 --> 01:10:59.960
think they're not everybody fully appreciates that.

01:11:00.040 --> 01:11:02.440
And so I want to just say, like, I think what you're doing is really important.

01:11:02.680 --> 01:11:07.880
Can you talk a little bit about um the natural selection of

01:11:08.200 --> 01:11:09.640
bad science?

01:11:10.360 --> 01:11:10.840
Yeah.

01:11:11.080 --> 01:11:13.160
Uh uh thanks thanks for that.

01:11:13.320 --> 01:11:19.320
Um so I mean, for uh first of all, I'll just say like I I think what you said is is

01:11:19.320 --> 01:11:21.000
spot on and and really important.

01:11:21.240 --> 01:11:26.010
And I think that you know It's a naive view, right?

01:11:26.170 --> 01:11:32.090
If people think that they can understand social behavior, the way that people

01:11:32.090 --> 01:11:37.610
behave, and almost all behavior is social behavior to some extent, um when you're talking

01:11:37.610 --> 01:11:43.050
about humans, uh, by only studying how individuals behave in kind of isolation, right?

01:11:43.210 --> 01:11:48.700
Like the lots of people interacting in a, you know,

01:11:48.860 --> 01:11:54.620
cultural and social context over time produces behaviors that are direct

01:11:54.620 --> 01:11:58.780
consequences of the organization, the way those things, which is you know, more like sort

01:11:58.780 --> 01:12:03.260
of sociological thinking, but um and I I think that's really important.

01:12:03.580 --> 01:12:06.140
Um so okay.

01:12:06.620 --> 01:12:12.230
Um So the natural selection of bad science was this paper,

01:12:12.470 --> 01:12:17.030
this project that kind of came out a little bit about, you know, uh Richard and I, this was

01:12:17.030 --> 01:12:17.750
a long time ago.

01:12:17.830 --> 01:12:22.070
We started working on this, um 2014, 2015.

01:12:22.230 --> 01:12:24.230
The paper came out in 2016.

01:12:24.390 --> 01:12:29.910
Um, and I think the context is important then because um

01:12:30.390 --> 01:12:36.360
you guys may recall, like this was kind of right in the aftermath of

01:12:36.360 --> 01:12:41.240
uh what's sometimes called the replication crisis, uh, where which was really kind of

01:12:41.240 --> 01:12:47.000
started in psychology as being recognized, but gradually people noticed it in other fields,

01:12:47.160 --> 01:12:52.440
biomedicine and you know, lots of, lots of fields, um, which was basically that

01:12:53.400 --> 01:12:59.210
uh a lot of the standard methods that had uh become standard in uh a

01:12:59.210 --> 01:13:02.570
bunch of these fields were producing results that didn't replicate.

01:13:02.810 --> 01:13:07.210
And it all like the the one of the main catalysts, it wasn't the only one, but one of the

01:13:07.210 --> 01:13:12.410
main catalysts that really kind of launched this investigation was this publication in this

01:13:12.410 --> 01:13:16.730
top journal in in social psychology by uh um by Daryl Bem.

01:13:17.130 --> 01:13:18.090
Daryl Bem, yeah, yeah.

01:13:18.330 --> 01:13:24.240
Yeah, he claimed to have uh evidence for um uh seeing the future,

01:13:24.400 --> 01:13:30.320
predicting the future, that uh uh undergraduates could choose which uh

01:13:31.120 --> 01:13:35.520
uh which curtain had a uh pornographic image behind it.

01:13:35.680 --> 01:13:39.600
Uh and it wasn't even a real curtain, it was uh like a computer screen.

01:13:39.760 --> 01:13:44.960
And the point was like the computer chose which curtain to put the image behind

01:13:45.250 --> 01:13:48.050
after the students had made their guess.

01:13:48.290 --> 01:13:51.330
So there was literally no way for the students to guess.

01:13:51.410 --> 01:13:57.170
And yet he's like, well, it was actually they they guessed correctly like 53% of the

01:13:57.170 --> 01:13:57.490
time.

01:13:57.890 --> 01:14:03.010
And you know, we did a t-test and found that that was statistically significant.

01:14:03.490 --> 01:14:09.380
And it like it's it's just ludicrously terrible methodology,

01:14:09.540 --> 01:14:14.740
just the thinking and I got really upset uh for another reason too about it, or kind of was

01:14:14.740 --> 01:14:17.780
up in arms about it, because I was like, this is a physics result.

01:14:18.340 --> 01:14:24.260
If you're claiming to like have evidence that everything we know

01:14:24.260 --> 01:14:27.890
about causality Is wrong.

01:14:28.050 --> 01:14:31.730
Like that we there's another substance in the other thing.

01:14:32.050 --> 01:14:34.290
This is like the foundation of everything we do.

01:14:34.450 --> 01:14:36.370
This is a massive claim.

01:14:36.610 --> 01:14:42.530
And your evidence is just like you know, uh some like a like

01:14:42.530 --> 01:14:48.210
20 undergraduates clicked a button uh in in slightly above chance.

01:14:49.000 --> 01:14:53.080
And you know, when people started investigating, they found that like actually he had

01:14:53.080 --> 01:14:57.480
showed them a whole bunch of different kinds of images, and pornographic images were the

01:14:57.480 --> 01:14:59.000
only ones that had a high enough.

01:14:59.240 --> 01:15:05.000
So it's like it's like the whole like uh there's this XKCD comic uh about

01:15:05.160 --> 01:15:10.920
where like he's joking about uh scientists giving people different color jelly beans

01:15:10.920 --> 01:15:15.950
and see if any of them will cure cancer, and then you just try all the different colors and

01:15:15.950 --> 01:15:21.310
finally one of them is like, oh yeah, like fewer people who had green jelly beans have

01:15:21.310 --> 01:15:21.870
cancer.

01:15:22.030 --> 01:15:26.990
And then like the headline is like, uh, green jelly beans cause can uh cure cancer.

01:15:27.310 --> 01:15:32.350
But it's if you try a million things, like statistically, one of them will have a

01:15:32.350 --> 01:15:32.990
correlation.

01:15:33.070 --> 01:15:38.350
But it's like there's that website, Spurious Correlations, where it's just amazing and

01:15:38.350 --> 01:15:39.390
everyone should look at it.

01:15:39.470 --> 01:15:45.420
It's like uh, you know the the number of uh like kevin bacon movies in a

01:15:45.420 --> 01:15:48.540
given decade correlates with like deaths by drowning.

01:15:48.940 --> 01:15:51.020
Um I mean tremors.

01:15:51.500 --> 01:15:53.100
That one is actually causal I think.

01:15:53.740 --> 01:15:54.140
Yeah.

01:15:55.100 --> 01:15:59.740
I can I I can understand how seeing Kevin Bacon would would make you less careful around

01:15:59.740 --> 01:16:00.540
bodies of water.

01:16:01.980 --> 01:16:02.540
I agree with that.

01:16:02.860 --> 01:16:06.550
Can I can I interject too this is like uh inside scoop information.

01:16:06.710 --> 01:16:12.630
I, as a graduate student, submitted a paper to that journal and the same action

01:16:12.630 --> 01:16:18.390
editor that handled the BEM article, and my paper was uh rejected.

01:16:18.550 --> 01:16:21.510
And it was like careful, it was sort of an obvious thing.

01:16:21.590 --> 01:16:26.550
It was like accents are categorized, they had tons of control conditions, and it was not

01:16:26.550 --> 01:16:28.950
accepted at the same time that the BEM paper was.

01:16:29.110 --> 01:16:34.950
So it's it's also for me like very egregious in that it's also it w it wasn't just sort of

01:16:34.950 --> 01:16:40.880
like people were doing collecting careful science, and then

01:16:40.880 --> 01:16:42.320
this just kind of got in.

01:16:42.400 --> 01:16:47.200
It was just from my experience, it was obviously careful science was just being like, no,

01:16:47.280 --> 01:16:50.720
this is nonsense, and then ESP, yeah, come on in.

01:16:50.800 --> 01:16:51.040
Yeah.

01:16:51.360 --> 01:16:51.520
Yeah.

01:16:51.760 --> 01:16:55.920
It seems like it's it's the opposite of what you'd want to be doing as I feel as a

01:16:55.920 --> 01:17:01.830
knowledge accumulating enterprise, which is taking into account past knowledge and trying

01:17:01.830 --> 01:17:06.630
to integrate new knowledge with the past edifice of knowledge and coming into your

01:17:06.630 --> 01:17:12.150
hypothesis with a set of well-informed priors based on prior research and prior knowledge,

01:17:12.310 --> 01:17:17.670
as opposed to just starting your investigation in a vacuum completely in ignorance of

01:17:17.670 --> 01:17:19.910
centuries of scientific research.

01:17:20.790 --> 01:17:24.870
And that's, I think, what we are all about as evolutionary social scientists is we want to

01:17:24.870 --> 01:17:29.480
integrate our knowledge of human beings with past knowledge that we've already accumulated

01:17:29.480 --> 01:17:31.320
about how evolution works, right?

01:17:31.800 --> 01:17:36.280
Uh and and and the fact that social science does not avail itself of that tremendously

01:17:36.280 --> 01:17:42.040
useful tool of using everything we know to inform our research is just mind-boggling.

01:17:42.120 --> 01:17:44.680
But anyway, Paul has a reason why people do that.

01:17:45.080 --> 01:17:46.440
Yeah, why do people do that?

01:17:46.520 --> 01:17:46.680
Yes.

01:17:47.000 --> 01:17:47.240
Yeah.

01:17:47.480 --> 01:17:50.510
So um I I agree.

01:17:50.590 --> 01:17:55.950
And yeah, so part of this project was was to sort of maybe highlight a potential reason for

01:17:55.950 --> 01:18:00.830
why uh people are not as good scientists as they should be or could be.

01:18:01.230 --> 01:18:06.830
Um which was basically, you know, a griping about like uh various things in

01:18:06.830 --> 01:18:08.670
academia kind of came out of it.

01:18:08.830 --> 01:18:14.300
Um and seeing, you know, these headlines, exactly the kind of thing you're talking about,

01:18:14.380 --> 01:18:20.140
Dave, like, yeah, the these ridiculous studies getting uh you know

01:18:21.100 --> 01:18:26.380
highlighted, put into, you know, top journals and by people at Ivy League universities

01:18:26.540 --> 01:18:32.540
while this like very careful, methodical, creative research uh was was getting passed over

01:18:32.620 --> 01:18:37.420
or or shunted or uh into lower you know uh ranked journals.

01:18:37.820 --> 01:18:43.570
And um, you know, one of the things that's

01:18:43.570 --> 01:18:49.330
true about academics and scientists is that they're human beings and they live in a

01:18:49.330 --> 01:18:54.850
world w you know that is characterized we live in an unequal hierarchical

01:18:54.850 --> 01:19:00.530
society where having wealth and acclaim and prestige

01:19:01.200 --> 01:19:06.560
gives you hired gives you as you know people certain kinds of satisfaction.

01:19:06.640 --> 01:19:08.000
It's things that people want.

01:19:08.480 --> 01:19:12.800
And it's also the case that not everyone can get those things.

01:19:13.200 --> 01:19:18.960
Which, you know, immediately anyone who thinks about evolution to think, well, okay,

01:19:19.120 --> 01:19:22.160
scientists vary, they're different from each other.

01:19:22.400 --> 01:19:26.880
Um there's a something that you know

01:19:28.550 --> 01:19:34.550
Allows for the transmission of scientific knowledge, where you know scientists train

01:19:34.550 --> 01:19:39.030
other scientists via universities and postdoctoral fellowships and trainings.

01:19:39.110 --> 01:19:41.270
So there's inheritance in that way.

01:19:41.430 --> 01:19:43.590
And there's selection, right?

01:19:44.070 --> 01:19:50.070
So not only uh so these are all the the these are all the ingredients you need

01:19:50.070 --> 01:19:52.070
for an evolutionary dynamic.

01:19:52.390 --> 01:19:58.080
And the question then is, well, you know, what

01:19:58.080 --> 01:20:01.600
ways do they vary, and especially how are they selected?

01:20:02.080 --> 01:20:07.760
And so we built a model because the truth is they vary in lots of ways, and they're

01:20:07.760 --> 01:20:09.520
selected in lots of ways.

01:20:09.840 --> 01:20:15.760
But we wanted to focus on the quality or rigor

01:20:17.110 --> 01:20:18.630
of the methodology.

01:20:18.870 --> 01:20:24.630
So we imagined, you know, we created a model where we have um we actually did a

01:20:24.630 --> 01:20:28.230
previous model where we had a population model of hypotheses.

01:20:28.470 --> 01:20:33.030
I actually I like this paper a lot, and it's uh uh often overlooked.

01:20:33.190 --> 01:20:39.060
So that this Richard and I did this previous paper where we looked at uh just a population

01:20:39.060 --> 01:20:44.980
of hypotheses that you know where they were tested with certain amounts of error

01:20:45.140 --> 01:20:49.700
for both false negatives and false positives and there were certain kinds of communication

01:20:49.700 --> 01:20:55.620
bias like publication bias and certain kinds of replication strategies and we looked at

01:20:55.700 --> 01:21:00.510
all right well like under different assumptions what would we be able to say about the

01:21:00.510 --> 01:21:05.630
probability that a hypothesis is true given a certain record of supporting and

01:21:05.630 --> 01:21:08.270
non-supporting results in the published record?

01:21:08.910 --> 01:21:12.990
And so we so we had that model done.

01:21:13.870 --> 01:21:19.470
And we started thinking, well, you know, that model just said, like, well, there's an

01:21:19.470 --> 01:21:20.430
error rate.

01:21:20.670 --> 01:21:25.870
There's a power, the statistical power, you know, the ability to find true results and

01:21:25.870 --> 01:21:27.710
avoid false negatives.

01:21:28.800 --> 01:21:34.400
What if we created a population of artificial science agents that varied in their

01:21:34.400 --> 01:21:35.040
methods?

01:21:35.360 --> 01:21:41.360
They said, well, okay, well, you know, everyone has a uh we we just said,

01:21:41.440 --> 01:21:47.360
all right, all the agents have some sort of intrinsic methods that can vary in how rigorous

01:21:47.360 --> 01:21:50.590
they are and how powerful they are.

01:21:50.910 --> 01:21:52.510
And those things trade off.

01:21:52.670 --> 01:21:58.270
So we can avoid false positives by being extremely

01:21:58.270 --> 01:22:01.550
careful, but then we might miss a lot of true positives.

01:22:01.790 --> 01:22:06.990
We can find all the true positives by having very high power, by you know, saying

01:22:06.990 --> 01:22:11.150
everything is true, that would be good, but that's not good because then we get a lot of

01:22:11.150 --> 01:22:12.990
false false positives.

01:22:13.310 --> 01:22:17.820
So we want to be able to say the things that are true are true and the things that are

01:22:17.820 --> 01:22:19.100
false are false.

01:22:19.420 --> 01:22:22.780
And to do that takes time and effort.

01:22:22.940 --> 01:22:24.780
It's being careful.

01:22:25.580 --> 01:22:31.580
But that means that you know you're not you're less likely to submit, you know, to get

01:22:31.580 --> 01:22:35.420
false positives, which gives you fewer things to submit to journals.

01:22:35.580 --> 01:22:38.620
And it's going to take you longer between studies.

01:22:39.340 --> 01:22:43.890
And so that's actually the kind of science that we think we want, which is careful science

01:22:43.970 --> 01:22:46.450
where the things we say are true are actually true.

01:22:46.690 --> 01:22:49.090
But then there's the selection pressure.

01:22:49.250 --> 01:22:51.650
And in this case, we did a simplified version.

01:22:51.810 --> 01:22:57.650
So in reality, right, the reason why scientists get hired and promoted and get

01:22:57.650 --> 01:23:02.850
grant funding, which allows them to, you know, train more graduate students,

01:23:03.010 --> 01:23:06.930
undergraduates, allows them to get more prestigious jobs where they have more influence.

01:23:07.250 --> 01:23:13.100
It comes down to a lot of things, but some of the biggest ones will be how many papers they

01:23:13.100 --> 01:23:18.220
publish, how prestigious and influential those papers are, how much grant money they get,

01:23:18.380 --> 01:23:18.940
etc.

01:23:19.420 --> 01:23:25.020
And uh we in the model, we just distilled this down to one thing, which is just how many

01:23:25.020 --> 01:23:26.140
papers you publish.

01:23:26.380 --> 01:23:31.980
And you we you we could build a model with other things, but in this simple version, uh we

01:23:31.980 --> 01:23:32.940
did it this way.

01:23:33.180 --> 01:23:35.130
And we also allowed for replication.

01:23:35.210 --> 01:23:39.130
So we said, all right, if somebody else replicates your study and it doesn't replicate,

01:23:39.210 --> 01:23:42.090
that's bad for you and that hurts your prestige.

01:23:42.250 --> 01:23:46.010
But otherwise, the more the more papers you publish, the more prestige you have.

01:23:46.250 --> 01:23:51.210
And the more prestige you have, the more likely you are to transmit your methods to someone

01:23:51.210 --> 01:23:56.810
else in the future because they you get a good job at a prestigious university, you get

01:23:56.810 --> 01:24:01.130
lots of grant funding, your papers are influential and people apply to work with you, et

01:24:01.130 --> 01:24:01.770
cetera.

01:24:02.330 --> 01:24:08.180
And this, we just kind of simulated this dynamic over sort of generations of

01:24:08.180 --> 01:24:14.020
scientists and showed that if the selection, the major selection

01:24:14.020 --> 01:24:19.860
pressure is how many papers you publish, all things equal, that's going to

01:24:19.860 --> 01:24:25.780
select for a continual decrease in the rigor that we

01:24:25.780 --> 01:24:26.100
see.

01:24:26.260 --> 01:24:28.710
Now, you know, it's a model.

01:24:29.110 --> 01:24:31.190
And like I said, there are other things.

01:24:31.990 --> 01:24:37.670
But if it's wrong, right, then we should see the opposite happening, right?

01:24:37.750 --> 01:24:40.230
Because what we actually want is things to be true.

01:24:40.470 --> 01:24:45.430
So we should see a gradual tend uh trend for better and better methods over time.

01:24:45.590 --> 01:24:51.590
If selection is actually favoring methods that are geared at improving truth, uh,

01:24:51.670 --> 01:24:57.100
that have, you know, uh evidence that they will actually help us get

01:24:57.660 --> 01:24:59.980
epistemic value out of our studies.

01:25:01.260 --> 01:25:05.500
But you know, there's a lot of evidence from a number of fields that that hasn't happened.

01:25:05.900 --> 01:25:11.580
I mean, two of the ones I like to highlight is like, you know, there's this famous uh

01:25:11.580 --> 01:25:15.500
paper by Paul Meal, who's a philosopher of science in the 1960s.

01:25:15.740 --> 01:25:21.550
He wrote this paper in the 1960s basically saying, hey, psychologists are really not

01:25:21.550 --> 01:25:26.030
using p-values correctly, this sort of null hypothesis significance testing.

01:25:26.270 --> 01:25:27.950
You're doing it really badly.

01:25:28.110 --> 01:25:29.950
You don't understand what p-values are.

01:25:30.030 --> 01:25:35.950
A p-value is not the probability that your hypothesis is

01:25:35.950 --> 01:25:36.270
true.

01:25:36.350 --> 01:25:39.390
And it's not the probability that the null hypothesis is false.

01:25:39.470 --> 01:25:45.340
It's actually just the probability that you saw the data you saw conditional on the null

01:25:45.340 --> 01:25:47.180
hypothesis being true.

01:25:47.980 --> 01:25:51.820
But that doesn't actually support any other specific hypothesis.

01:25:53.180 --> 01:25:58.860
And you would think that, you know, in the 1960s, that was 60 years ago,

01:25:59.020 --> 01:26:00.140
that would have an effect.

01:26:00.380 --> 01:26:05.500
But just less than a decade ago, the American Statistical Association published this like

01:26:05.500 --> 01:26:09.580
statement on p-values, just reiterating all the same things that Paul Meal says.

01:26:09.740 --> 01:26:13.610
He's like, nobody, we did a survey, people don't understand p values, they're not using

01:26:13.610 --> 01:26:14.650
statistics correctly.

01:26:14.810 --> 01:26:16.170
Well, why do you think that is?

01:26:16.410 --> 01:26:18.250
Well, understanding statistics is hard.

01:26:18.330 --> 01:26:20.650
It takes a lot of time when you could be running studies.

01:26:20.810 --> 01:26:26.730
And if you really understand statistics, you're going to be a lot more qualified in how

01:26:26.730 --> 01:26:28.010
you promote the results you got.

01:26:28.090 --> 01:26:29.370
You're going to be a lot more cautious.

01:26:29.530 --> 01:26:33.210
Whereas if you don't, you might just say, look, we found this amazing thing.

01:26:33.530 --> 01:26:35.690
Green jelly beans cure cancer.

01:26:36.480 --> 01:26:38.880
And that's going to get attention.

01:26:39.120 --> 01:26:42.560
And in a world where there's not good vetting of these kinds of results, where scientists

01:26:42.640 --> 01:26:47.520
are policing themselves, you can actually get selection for misunderstanding.

01:26:49.040 --> 01:26:54.080
Is this um would you view this analogously to uh life history theory where there are

01:26:54.080 --> 01:26:57.440
trade-offs between quantity versus quality of offspring?

01:26:57.600 --> 01:27:02.560
You could either invest very intensively in a small number of offspring, um, if they're you

01:27:02.560 --> 01:27:08.070
know, if you expect them to live a long time, uh versus if you are in a high predation

01:27:08.070 --> 01:27:12.390
environment where most of your offspring are expected to die, uh, you might want to hedge

01:27:12.390 --> 01:27:16.710
your bets and have a large number of offspring lay a bunch of different eggs in hopes that

01:27:16.950 --> 01:27:21.990
and invest very little in any of them in hopes that at least a few of them uh will will

01:27:21.990 --> 01:27:22.710
survive.

01:27:22.870 --> 01:27:27.510
Um would you view this uh trade-off between quantity and quality as analogous to the

01:27:27.510 --> 01:27:30.630
trade-off between quantity and quality in scientific research?

01:27:31.510 --> 01:27:33.460
I think loosely analogous, yeah.

01:27:33.620 --> 01:27:38.180
I mean, uh to on the you know, to some extent, first of all, like I mean, all evolution,

01:27:38.260 --> 01:27:39.540
right, is about trade-offs.

01:27:39.700 --> 01:27:39.940
Right.

01:27:40.180 --> 01:27:44.740
So anything evolutionary is going to be about trade-offs, but I think even more so, right,

01:27:44.900 --> 01:27:50.900
this is an interesting analogy because um in a lot of scientific

01:27:50.900 --> 01:27:56.820
fields, people get, you know, the jobs and

01:27:56.980 --> 01:28:02.910
fellowships that junior scientists get right out of graduate school or even during

01:28:02.910 --> 01:28:04.990
graduate school matter a lot.

01:28:05.630 --> 01:28:11.390
Um there was a really neat study about a decade ago by this Dutch group, and they looked at

01:28:11.470 --> 01:28:16.990
um the people who got this um this very prestigious um funding uh award.

01:28:17.470 --> 01:28:20.110
Um so it gives you a lot of money to do your research.

01:28:20.190 --> 01:28:22.030
It's also very prestigious, helps you get jobs.

01:28:22.110 --> 01:28:27.360
And they looked at basically the people who are who are ranked just below the threshold and

01:28:27.360 --> 01:28:28.240
just above the threshold.

01:28:28.560 --> 01:28:33.520
The people who got the award, but barely, versus the people who barely missed out on the

01:28:33.520 --> 01:28:33.840
award.

01:28:33.920 --> 01:28:39.040
And then they tracked them for a number of years after, and their paths diverge wildly.

01:28:39.440 --> 01:28:43.200
There's a huge effect of these early successes.

01:28:43.680 --> 01:28:49.670
So the the value of edging past being perceived as

01:28:49.670 --> 01:28:55.270
edging past your peers early on is it ends up in our system being really high in a lot of

01:28:55.270 --> 01:28:55.830
fields.

01:28:56.070 --> 01:29:01.910
And so if you take on as a graduate student a very complicated project

01:29:01.990 --> 01:29:06.390
that's you're you're spending, you know, being very rigorous, very careful about and taking

01:29:06.390 --> 01:29:09.270
a lot of time and doesn't really lead to a lot of publications.

01:29:09.430 --> 01:29:14.760
Maybe you're maybe a couple of like smaller publications that are um you're publishing kind

01:29:14.760 --> 01:29:16.600
of on the way to something big.

01:29:16.680 --> 01:29:22.280
It's not big yet, but maybe it will be amazing later on, versus someone who's just cranking

01:29:22.280 --> 01:29:27.880
out those studies and trying and aiming for things that are the sort of the most catchy and

01:29:27.880 --> 01:29:30.760
most newsworthy and are the least careful.

01:29:31.000 --> 01:29:36.230
Well, what's gonna happen, right, is if you have a system that

01:29:36.230 --> 01:29:39.270
values that second person over the first one.

01:29:40.310 --> 01:29:43.350
And it's not every system, not every field is like this.

01:29:43.750 --> 01:29:49.030
But when it is like that, then that first person doesn't get a job.

01:29:49.190 --> 01:29:52.710
They either get a low ranked job where they're not going to have as many resources to

01:29:52.710 --> 01:29:56.550
continue doing their work, or they're going to leave academia and do something else.

01:29:56.790 --> 01:30:02.680
And so the science that gets done overall Um gets

01:30:02.680 --> 01:30:03.000
worse.

01:30:03.080 --> 01:30:07.640
And this is actually why I think it's so important to think about the the sort of cultural

01:30:07.640 --> 01:30:10.920
and institutional incentives and not just individual behavior.

01:30:11.560 --> 01:30:15.960
Like we all know about perverse incentives, but

01:30:18.520 --> 01:30:24.170
there's there are some people who think, well, look, you just have to be better than that.

01:30:24.330 --> 01:30:25.770
You just have to be better than that.

01:30:25.850 --> 01:30:28.730
You just have to do the work that you know is good.

01:30:29.690 --> 01:30:33.290
And yeah, well, that sounds really noble.

01:30:33.690 --> 01:30:38.730
Um but and let's imagine you know you do and you feel really satisfied.

01:30:38.970 --> 01:30:39.290
Yeah.

01:30:39.530 --> 01:30:45.220
Maybe you get lucky, but if you take, you know, a hundred people like that and a hundred

01:30:45.380 --> 01:30:50.740
people who are just, you know, playing the system in the using the techniques that the

01:30:50.740 --> 01:30:54.260
system selects for, well, the system is going to select for those people.

01:30:54.500 --> 01:30:54.900
Yes.

01:30:55.140 --> 01:30:57.860
And so it can't just be about individual behavior.

01:30:57.940 --> 01:31:03.700
It has to be also about, you know, the sort of cultural and institutional incentives, the

01:31:03.700 --> 01:31:07.150
mechanisms that select for certain people.

01:31:07.630 --> 01:31:10.430
I mean, I didn't we didn't think about this at the time.

01:31:10.590 --> 01:31:12.750
I completely wasn't turned on to it back then.

01:31:12.830 --> 01:31:18.670
But I mean, this is very similar to what, you know, Noam Chomsky and Edward Herman were

01:31:18.670 --> 01:31:23.950
talking about in Manufacturing Consent on their book on the media in the 1980s.

01:31:24.110 --> 01:31:29.300
This is also a kind of sort of institutional selection where they're saying, you know, the

01:31:29.540 --> 01:31:35.540
the people who are in control of, you know, media organizations are hiring

01:31:35.540 --> 01:31:41.300
people and promoting people who espouse certain views and think about things in

01:31:41.300 --> 01:31:43.300
certain ways and not hiring others.

01:31:43.460 --> 01:31:48.420
And even if every single person who goes into journalism is completely noble and thinks

01:31:48.420 --> 01:31:53.620
they're doing the best they can, the people who think along the party lines are going to be

01:31:53.620 --> 01:31:58.090
the ones that end up in positions of authority and influence, and the ones who think

01:31:58.090 --> 01:31:59.370
differently are not.

01:31:59.930 --> 01:32:05.690
Yeah, it's or that it's uh the the whatever will be an audience, whatever increases the A-B

01:32:05.850 --> 01:32:09.770
testing of like what gets more eyeballs on this, so we can sell more ads.

01:32:10.090 --> 01:32:15.530
Um, so yeah, if it's um uh yeah, the idea that

01:32:15.530 --> 01:32:18.920
selection is what matters and not intention.

01:32:19.160 --> 01:32:20.840
I think politics too, right?

01:32:20.920 --> 01:32:26.760
Like you can you could say, you know, um how do we minimize crime?

01:32:27.000 --> 01:32:30.040
Well we should do things that like punish criminals, right?

01:32:30.120 --> 01:32:34.200
As opposed to like you know minimal standard wage or something, like Dan Nunnell's

01:32:34.360 --> 01:32:35.080
suggesting.

01:32:35.320 --> 01:32:40.930
Um the idea that you you can't intuit intentionally what will

01:32:40.930 --> 01:32:45.650
produce the outcome you want, and you have to just rely on the idea that there's selection,

01:32:45.810 --> 01:32:47.330
I think is a really humbling thing.

01:32:47.410 --> 01:32:51.810
And I think your work, when I've read it, every time I do something, I think to myself,

01:32:52.050 --> 01:32:56.130
like honestly, like the I often think of the the natural selection of bad science work

01:32:56.210 --> 01:32:59.490
where I'm like, what am I selecting by clicking on this?

01:32:59.730 --> 01:33:05.650
Or what are we selecting if we make a decision about who to hire based on this criteria?

01:33:06.000 --> 01:33:07.680
Um, how are we contributing?

01:33:07.840 --> 01:33:11.200
How are we like a cog in the wheel of the institution?

01:33:11.600 --> 01:33:12.080
Yeah.

01:33:12.320 --> 01:33:12.640
Yeah.

01:33:13.040 --> 01:33:13.280
Yeah.

01:33:13.440 --> 01:33:18.240
I mean, it's even I mean, I I feel like we this might even be a case of mismatch where we

01:33:18.240 --> 01:33:22.560
overestimate the effectiveness of appeals to morality and conscience.

01:33:22.720 --> 01:33:28.320
That if we just scold our way to better science or wag our finger to better science, that

01:33:28.320 --> 01:33:29.280
that's going to solve the problem.

01:33:29.360 --> 01:33:32.320
We just need to encourage people to have integrity and to do better.

01:33:32.480 --> 01:33:37.350
Well, the problem with that view is that the people who are most susceptible to those

01:33:37.350 --> 01:33:41.670
appeals to conscience are going to be selected against and under the current incentive

01:33:41.830 --> 01:33:42.390
incentive structure.

01:33:42.710 --> 01:33:46.790
And the people who are least susceptible to those appeals are going to rise to the top,

01:33:46.950 --> 01:33:47.110
right?

01:33:47.430 --> 01:33:47.590
Yeah.

01:33:47.750 --> 01:33:52.310
So you end up if you want what you want is a world where integrity and, you know, morality,

01:33:52.390 --> 01:33:58.390
let's say, are favored, you actually have to change the system so that the people who act

01:33:58.390 --> 01:34:03.700
that way are actually the ones that rise to the top and are the ones that get rewarded and

01:34:03.700 --> 01:34:04.900
the other ones get don't.

01:34:05.140 --> 01:34:09.780
And I think, I mean, with politics, I mean, obviously we're seeing the effects of uh

01:34:09.780 --> 01:34:13.060
selection every day uh right now.

01:34:13.700 --> 01:34:18.820
But I mean, you know, even just at the most basic level, uh, you know, we we have an

01:34:18.820 --> 01:34:24.180
electoral system with in the US, this like long drawn out, kind of very media

01:34:24.340 --> 01:34:30.070
based um electoral system where where people run for office over months

01:34:30.310 --> 01:34:34.070
and what what what does that do right it selects for the people who are really good at

01:34:34.070 --> 01:34:37.910
running for office but the people who are really good at running for office in that system

01:34:37.910 --> 01:34:42.150
aren't necessarily the people who are really good at governing right yeah or or even like

01:34:42.230 --> 01:34:46.230
you know gerrymandering or you know it's even worse like the selection I mean yeah yeah

01:34:46.310 --> 01:34:51.990
like it's yeah yeah you would uh uh yeah exactly the the selection or even yeah

01:34:52.070 --> 01:34:57.600
yeah tons of examples of uh if you can if you can put your thumb on the scale of the

01:34:57.600 --> 01:35:01.520
selection process, it's really leads to crazy outcomes.

01:35:02.000 --> 01:35:02.240
Yes.

01:35:02.400 --> 01:35:07.040
It's a counterintuitive, like we're we're not inclined to view things in terms of incentive

01:35:07.040 --> 01:35:11.040
structure problems as opposed to character or or ethical problems.

01:35:11.200 --> 01:35:11.360
Yeah.

01:35:11.520 --> 01:35:14.080
I I again it may be a kind of mismatch.

01:35:14.160 --> 01:35:14.640
I don't know.

01:35:14.720 --> 01:35:17.830
But it seems like our first inclination is we gotta boo the bad people.

01:35:17.990 --> 01:35:21.590
We gotta figure out who the bad people are and discourage them and we gotta praise the good

01:35:21.590 --> 01:35:25.110
people, and that's the way to solve the problem as opposed to viewing it as an

01:35:25.110 --> 01:35:30.070
incentive-based problem that we Yeah, although I do want to maybe push back on that

01:35:30.070 --> 01:35:35.830
slightly because if it becomes normative to boo the bad people

01:35:36.150 --> 01:35:41.990
and punish the bad people, and the mechanisms for bad people succeeding

01:35:42.310 --> 01:35:44.310
are not there, right?

01:35:44.540 --> 01:35:49.980
Or are shut down actively, that does change it into a selective system.

01:35:50.220 --> 01:35:50.380
Sure.

01:35:50.780 --> 01:35:55.660
So booing the bad people can be a you know a

01:35:55.660 --> 01:36:00.220
precursor or let's say an early, you know, sign of

01:36:01.500 --> 01:36:03.900
more normative, more institutionalized change.

01:36:04.140 --> 01:36:05.660
But you have to get all the way there.

01:36:06.060 --> 01:36:08.540
It just can't be yelling from the bleachers.

01:36:08.860 --> 01:36:09.180
Yeah.

01:36:09.420 --> 01:36:09.980
Totally.

01:36:10.130 --> 01:36:10.370
Yeah.

01:36:10.450 --> 01:36:10.690
Yeah.

01:36:10.930 --> 01:36:12.770
I I feel I just just went around.

01:36:12.850 --> 01:36:14.290
Like I just I can't emphasize enough.

01:36:14.370 --> 01:36:19.650
I wish like we don't have ads on this, but like the PSA of like selection and not intention

01:36:19.970 --> 01:36:21.410
is really what matters.

01:36:21.650 --> 01:36:27.090
I feel like that's such an important sort of public uh education thing.

01:36:27.250 --> 01:36:31.010
Like I wish there was something in school, even where I don't know, like just game theory

01:36:31.010 --> 01:36:31.890
models or something.

01:36:32.050 --> 01:36:35.010
I mean, I don't know, I don't know what you do with that, but just the idea.

01:36:35.170 --> 01:36:41.020
I think I just I even for myself, I still catch myself going in the wrong direction

01:36:41.100 --> 01:36:46.460
as David's talking about where I just sort of really overemphasize people's intentions and

01:36:46.460 --> 01:36:52.140
motivations or wagging a finger as opposed to people are doing what makes sense given the

01:36:52.140 --> 01:36:55.340
incentive structure in the environment and that's the thing you gotta really target.

01:36:55.420 --> 01:36:58.140
I mean it's hard because it's you know it's a bigger than one person.

01:36:58.460 --> 01:37:02.380
It's hard and I it's not it's it's not all one or all the other either.

01:37:02.780 --> 01:37:03.820
Yeah no no that's fair.

01:37:03.900 --> 01:37:04.060
Yeah.

01:37:04.300 --> 01:37:04.380
Right.

01:37:04.620 --> 01:37:09.450
Especially like the problem is there are also incentive structures for changing the

01:37:09.450 --> 01:37:10.250
incentive structure.

01:37:10.410 --> 01:37:11.770
And those can be fucked too.

01:37:12.330 --> 01:37:13.130
And that's the problem.

01:37:13.770 --> 01:37:13.850
Right.

01:37:14.890 --> 01:37:15.210
Yeah.

01:37:15.450 --> 01:37:15.610
Yeah.

01:37:15.850 --> 01:37:19.530
You know, like the people who r who rise to positions uh of power who are capable of

01:37:19.530 --> 01:37:20.730
changing the incentive structure.

01:37:20.890 --> 01:37:24.890
Well, that is itself a whole selection process that you're not getting the people who are

01:37:24.890 --> 01:37:26.650
best able to to to do that.

01:37:26.970 --> 01:37:27.210
Yeah.

01:37:27.450 --> 01:37:31.130
No one wants to change the incentive structure to be something that wouldn't have selected

01:37:31.130 --> 01:37:31.450
for them.

01:37:31.690 --> 01:37:32.010
Correct.

01:37:32.330 --> 01:37:32.970
Exactly.

01:37:33.290 --> 01:37:33.610
Yeah.

01:37:33.820 --> 01:37:34.380
Exactly.

01:37:34.620 --> 01:37:34.940
Yeah.

01:37:35.100 --> 01:37:35.900
Um Paul.

01:37:37.020 --> 01:37:37.180
Yeah.

01:37:37.420 --> 01:37:38.140
No, go go ahead.

01:37:38.220 --> 01:37:38.300
Yeah.

01:37:38.620 --> 01:37:42.780
No, I was just gonna say, I mean, we saw that in science too, with a lot of pushback uh

01:37:43.100 --> 01:37:48.300
against reform uh a decade ago, where a lot of people who had succeeded very well, you

01:37:48.300 --> 01:37:53.180
know, with the old system were very against these sort of uh increased scrutiny.

01:37:53.500 --> 01:37:53.740
Yes.

01:37:54.220 --> 01:37:54.380
Yes.

01:37:54.620 --> 01:37:55.500
Yes, I do remember.

01:37:55.580 --> 01:37:59.500
I mean, I I was in grad school during the whole when the replication crisis was like

01:37:59.500 --> 01:38:00.860
reaching its its peak.

01:38:01.020 --> 01:38:05.910
And I remember there was a lot of like political squabbles between the old guard were

01:38:05.990 --> 01:38:09.350
saying everything's fine, this is just a bunch of bunch of bullshit.

01:38:09.430 --> 01:38:12.470
These are like replication bullies, replication police, or whatever.

01:38:12.630 --> 01:38:14.790
And like the new guard was like, no, this is a serious problem.

01:38:14.870 --> 01:38:16.790
We need to really change things.

01:38:16.950 --> 01:38:22.710
Uh and it was not uh uh a coincidence that the old guard, the people who were fighting for

01:38:22.710 --> 01:38:26.310
the status quo, were the people who were the most privileged, the most prestigious, uh,

01:38:26.480 --> 01:38:28.240
with the highest positions in the current system.

01:38:28.400 --> 01:38:32.880
They were obviously the stakeholders of the status quo, and the the new guard were the

01:38:32.880 --> 01:38:37.040
people who were uh most uh likely to benefit from a a new system.

01:38:37.280 --> 01:38:42.720
Um David Pinsafian terms, yeah.

01:38:43.120 --> 01:38:43.440
Exactly.

01:38:43.520 --> 01:38:46.880
It was I was in the middle of a status game that was on the brink of collapse, and there

01:38:46.880 --> 01:38:50.560
were stakeholders in the current status game that were trying to preserve it, and uh

01:38:50.640 --> 01:38:54.320
stakeholders in a new status game that were trying to undermine it.

01:38:54.480 --> 01:38:58.710
Uh and um I think, you know, uh it uh things have improved.

01:38:58.870 --> 01:39:04.150
I mean I maybe you can correct me if I'm wrong, Paul, but um I I think um we have at least

01:39:04.150 --> 01:39:05.030
seen some progress.

01:39:05.190 --> 01:39:09.430
I I would like to see more, but I think we're certainly better than we were in the peak of

01:39:09.430 --> 01:39:10.390
the replication crisis.

01:39:10.470 --> 01:39:11.750
Would you would you disagree with that?

01:39:12.150 --> 01:39:14.470
Well yeah, yes and no.

01:39:14.630 --> 01:39:18.470
I I think that we have yes, I think there are a lot of things that are better.

01:39:18.710 --> 01:39:24.560
Um I think that a lot more the people are a lot more savvy about statistics than they were

01:39:24.560 --> 01:39:26.480
and methodological design.

01:39:27.600 --> 01:39:32.000
I think people are generally, you know, graduate students coming out of the social sciences

01:39:32.000 --> 01:39:35.120
and psychology are better at coding and better at statistics.

01:39:35.280 --> 01:39:37.840
They're increasingly thinking about causality.

01:39:39.520 --> 01:39:45.190
I still think that the sort of like Like myopic view

01:39:45.430 --> 01:39:50.790
of you know, I will do one study that changes the world, and without really thinking about

01:39:50.790 --> 01:39:56.150
how it is linked to the the larger, you know uh

01:39:58.070 --> 01:40:03.910
constellation or you know, uh blanket of human behavior, the breadth of

01:40:03.910 --> 01:40:07.990
human behavior, I think that's still missing a lot of the time, and and it leads to a lot

01:40:07.990 --> 01:40:12.920
of work that I feel like could be good if there was some more thought that went into it.

01:40:13.000 --> 01:40:16.120
But I've been I've been disappointed in a lot of the work that's come out.

01:40:16.280 --> 01:40:22.120
And you know, maybe this is just there's the old um adage that 90% of everything is crap,

01:40:22.200 --> 01:40:24.680
and maybe that's just always true.

01:40:25.720 --> 01:40:30.760
So but I think like that the good 10% uh is is is pretty good.

01:40:30.920 --> 01:40:32.360
Now there's a lot of good stuff.

01:40:32.850 --> 01:40:37.810
The the main point of progress I've seen is that you can now get more status for a failed

01:40:37.810 --> 01:40:41.570
replication attempt than you could have in the old era.

01:40:41.810 --> 01:40:45.490
Now, if I fail to replicate someone else's work, I could get that published.

01:40:45.570 --> 01:40:47.170
Uh and I could get some attention for that.

01:40:47.250 --> 01:40:50.690
Uh it used to be that if you failed to replicate someone's work, no one would pay attention

01:40:50.770 --> 01:40:51.570
and no one would care.

01:40:51.730 --> 01:40:53.250
So that I think is at least some progress.

01:40:53.490 --> 01:40:54.290
Yeah, it's an improvement.

01:40:54.530 --> 01:40:59.240
Um the the thing that I haven't seen as much progress as I would like is in terms of

01:40:59.240 --> 01:40:59.480
theory.

01:40:59.880 --> 01:41:01.480
Yeah, I was gonna say, yeah, yeah.

01:41:01.800 --> 01:41:06.840
And and and the and the uh people think of theory as like an airy, superfluous thing, but

01:41:06.840 --> 01:41:12.120
it actually feeds directly into into your your model, uh, which is that if you if you have

01:41:12.120 --> 01:41:17.080
an incentive structure where you are ignoring past knowledge and you are preferentially

01:41:17.080 --> 01:41:22.050
seeking out low prior hypotheses, And the lower prior your hypothesis

01:41:23.410 --> 01:41:28.130
if it pans out, the higher your status, well, that is going to further exacerbate the

01:41:28.130 --> 01:41:28.290
trend.

01:41:28.450 --> 01:41:32.610
So if I'm hunting for counterintuitive findings that clash with everything we have known in

01:41:32.610 --> 01:41:38.610
the past, and I get more prestige if my finding is more shocking and more titillating,

01:41:38.930 --> 01:41:44.660
well, um that is a recipe for even worse science and even less replicable

01:41:44.660 --> 01:41:44.980
work.

01:41:45.220 --> 01:41:49.460
Because if I'm pumping out lots of different studies and looking for the low prior uh

01:41:49.460 --> 01:41:52.980
hypothesis, uh, well, chances are one of them is going to turn out to be right.

01:41:53.060 --> 01:41:56.820
And if I publish that, I'm gonna get tons of status for discovering this new flashy

01:41:56.900 --> 01:41:58.500
counterintuitive thing, right?

01:41:58.740 --> 01:42:03.540
But if we're all uh in a different incentive structure where instead of uh hunting for the

01:42:03.540 --> 01:42:07.780
lowest prior hypotheses, we're hunting for the highest prior hypotheses, well then all of a

01:42:07.780 --> 01:42:09.810
sudden the dynamics are different.

01:42:09.970 --> 01:42:15.170
Uh if if we have a theoretical edifice that uh is causing us to preferentially look for uh

01:42:15.170 --> 01:42:19.890
data that is more likely to be uh confirmed and hypotheses that are more likely to bear

01:42:19.890 --> 01:42:25.410
fruit, that is going to uh help the replication crisis uh much more because any given

01:42:25.410 --> 01:42:29.650
finding is just going to be more likely to to replicate because it has a prior edifice of

01:42:29.650 --> 01:42:30.530
knowledge behind it.

01:42:30.690 --> 01:42:31.090
I don't know.

01:42:31.170 --> 01:42:33.010
What do you what do you think about that that way of thinking?

01:42:33.400 --> 01:42:39.080
I I I understand that way of thinking, but I I also think that there's a a problem that

01:42:39.080 --> 01:42:40.840
potentially comes out of it.

01:42:41.000 --> 01:42:46.440
Um, which is that we we want to the things that are

01:42:46.440 --> 01:42:52.200
counterintuitive or surprising, let's say, are actually the more valuable if

01:42:52.200 --> 01:42:53.080
they're right.

01:42:53.880 --> 01:42:59.400
So you know if I tell you that like, hey, I dropped a thousand pennies and all of them fell

01:42:59.400 --> 01:43:04.610
down, Like that's not a interesting like I yeah, I can replicate that, but like no one

01:43:04.610 --> 01:43:05.090
cares, right?

01:43:05.170 --> 01:43:06.530
We know it's gonna happen.

01:43:07.650 --> 01:43:12.690
If I tell you that like I discovered a material that actually goes up when you drop it and

01:43:12.690 --> 01:43:15.570
we never knew about it before, that's amazing.

01:43:15.730 --> 01:43:18.210
Um I mean it's just a helium balloon, right?

01:43:18.290 --> 01:43:20.930
But like it's still seems amaz it's amazing still.

01:43:21.410 --> 01:43:22.610
Helium balloons are incredible.

01:43:23.010 --> 01:43:23.490
They're the best.

01:43:23.650 --> 01:43:25.170
Don't don't poo poo helium balloons.

01:43:25.250 --> 01:43:25.650
They're amazing.

01:43:25.730 --> 01:43:26.290
It's magic.

01:43:27.580 --> 01:43:30.540
Things, you know, things that are counterintuitive, they're they're surprising.

01:43:30.700 --> 01:43:34.620
We pay attention to surprising things for like an a you know a very good evolutionary

01:43:34.620 --> 01:43:35.180
reason, right?

01:43:35.260 --> 01:43:36.700
That it has information.

01:43:37.020 --> 01:43:42.540
Uh and the the issue, right, is like if every single person individually is

01:43:42.540 --> 01:43:47.260
looking for the most surprising thing because individuals get rewarded when they find

01:43:47.260 --> 01:43:48.780
something really surprising.

01:43:49.020 --> 01:43:54.860
Well, everyone's gonna look, try to like find something regardless, especially

01:43:54.860 --> 01:44:00.450
if some of these other incentives are also in place and people are not

01:44:00.930 --> 01:44:05.010
looking for things with good reason, they're just kind of, you know, if you find something

01:44:05.010 --> 01:44:08.450
surprising and it's true and it's real, that's valuable.

01:44:09.010 --> 01:44:12.770
But if you find something that you say is surprising, but actually if you think about it

01:44:12.770 --> 01:44:16.770
for a minute, it's like it's not, it's only surprising because like you've been like really

01:44:16.770 --> 01:44:22.700
myopically and your field has been myopically looking in one place, um, you know,

01:44:22.860 --> 01:44:26.700
uh another area of research I've been getting more and more interested in is the sort of

01:44:26.700 --> 01:44:31.180
like collective intelligence, collective uh problem solving stuff.

01:44:31.260 --> 01:44:36.860
And you know, if if we stop thinking about scientists as a bunch of individuals

01:44:37.020 --> 01:44:42.860
that all are individually incentivized to be the individually best scientist, but

01:44:42.860 --> 01:44:48.110
think about you know teams and collectives as what would be the best strategy if if we have

01:44:48.110 --> 01:44:52.510
aligned interests that we're all trying to find uh good stuff?

01:44:52.750 --> 01:44:57.710
Well, to some extent it would it would be to get it wouldn't be to have every single person

01:44:57.710 --> 01:44:59.550
working on exactly the same problem.

01:45:00.190 --> 01:45:06.030
You would want an enough idiosyncrasity, enough

01:45:06.190 --> 01:45:11.860
you know, uh iconoclasts That people are looking in different places, that people are

01:45:11.860 --> 01:45:15.060
trying things that uh might not work.

01:45:15.380 --> 01:45:15.700
Yeah.

01:45:15.940 --> 01:45:20.740
And they're not punished for failing as long as they're doing it well, as long as they're

01:45:20.740 --> 01:45:24.740
being rigorous and careful in the way that they're doing it.

01:45:25.060 --> 01:45:30.740
And if you have people who are if what's being, you know, sort of scrutinized is

01:45:31.380 --> 01:45:37.290
is this a good idea and is the way you're doing it good, then

01:45:37.530 --> 01:45:39.770
let people do all kinds of things.

01:45:40.090 --> 01:45:45.850
And if we can if we can evaluate the methods, but you know, and the

01:45:45.850 --> 01:45:51.770
methods can include the method of selecting a hypothesis, it's tricky because

01:45:52.570 --> 01:45:57.530
evaluation is by its very nature a

01:45:57.530 --> 01:45:59.850
variance-reducing process.

01:46:00.410 --> 01:46:00.970
Yeah.

01:46:01.320 --> 01:46:06.200
And variance reducing is good if you want to maintain high quality

01:46:06.600 --> 01:46:11.480
that's um you know corresponds to a fixed set of criteria.

01:46:11.720 --> 01:46:15.800
But you know, as we know well, right, evolution works on two things, right?

01:46:15.960 --> 01:46:17.480
Selection and variance.

01:46:18.040 --> 01:46:22.120
And you know, it's the fundamental theorem of evolution, right?

01:46:22.200 --> 01:46:26.680
Is that the speed of evolution is proportional to the variance in the population.

01:46:26.840 --> 01:46:32.750
So higher variance actually makes a s a population more adaptable and more resilient.

01:46:32.990 --> 01:46:38.910
So we also need mechanisms to inject variation into this system if we

01:46:38.910 --> 01:46:40.830
want it to be you know to be making progress.

01:46:41.070 --> 01:46:43.790
So yeah, no I I I I agree with that.

01:46:43.870 --> 01:46:49.310
Yeah, no, it's it's um it's analogous to like uh you know if I have a a stock portfolio and

01:46:49.310 --> 01:46:53.870
I want to invest in a bunch of different stocks and I want to ideally invest in stocks that

01:46:53.870 --> 01:46:57.960
are likely to give me a return, but maybe I I'll take a risk you know, in a few cases.

01:46:58.120 --> 01:47:02.040
Maybe I'll, you know, I'll I'll include a few risky stocks on my portfolio too.

01:47:02.120 --> 01:47:03.640
And that's probably a good thing, right?

01:47:03.800 --> 01:47:05.160
And the same thing is true of science.

01:47:05.240 --> 01:47:10.040
You know, we wanna uh you know uh invest in uh uh theories and ideas that are likely to

01:47:10.040 --> 01:47:13.960
bear fruit, but maybe a small percentage of us can take some risks on some wilder ideas and

01:47:13.960 --> 01:47:15.000
we'll and we'll see what happens.

01:47:15.240 --> 01:47:19.160
But I think the current distribution is lopsided, where we have too many people investing

01:47:19.160 --> 01:47:24.210
in the risky strategy and not enough incentive for figuring out when those risky findings

01:47:24.210 --> 01:47:25.570
are actually correct or not.

01:47:25.970 --> 01:47:30.530
And not enough people investing in the more uh uh promising ideas.

01:47:30.690 --> 01:47:35.890
Um, and we get a lopsided uh distribution of research.

01:47:35.970 --> 01:47:41.890
Um ideally, we would have um uh you know projects that have some high

01:47:41.890 --> 01:47:43.570
likelihood of bearing fruit.

01:47:43.730 --> 01:47:47.700
Um but the problem is that a lot of people are investing in projects that have a low

01:47:47.700 --> 01:47:48.740
likelihood of bearing fruit.

01:47:48.900 --> 01:47:53.220
And and the deeper problem is they don't even realize that because they haven't bothered to

01:47:53.220 --> 01:47:58.980
familiarize themselves with past research and with past knowledge uh and and with uh say

01:47:59.220 --> 01:48:04.020
evolutionary theory, economics, whatever, whatever is going to give them an edge in uh uh

01:48:04.260 --> 01:48:08.020
making sharper predictions that are more likely to bear fruit.

01:48:08.180 --> 01:48:08.820
I don't know.

01:48:09.290 --> 01:48:09.770
I don't know.

01:48:09.850 --> 01:48:10.170
It's hard.

01:48:10.250 --> 01:48:10.810
It's hard to say.

01:48:10.890 --> 01:48:14.410
I mean, like science is there's a lot of scientists studying a lot of different things, but

01:48:14.410 --> 01:48:20.010
I think in the in the sort of behavioral sciences, I worry that it's it's worse than that,

01:48:20.090 --> 01:48:20.330
right?

01:48:20.410 --> 01:48:25.930
That you have actually like um low quality, low risk soci

01:48:26.250 --> 01:48:27.530
uh work being done.

01:48:28.570 --> 01:48:34.170
It's likely to uh to find something that uh

01:48:34.650 --> 01:48:36.650
doesn't kind of matter.

01:48:36.890 --> 01:48:37.130
Yeah.

01:48:37.370 --> 01:48:37.530
Yeah.

01:48:39.100 --> 01:48:42.300
I mean, and I like one of you might even be end up being a reviewer for this.

01:48:42.380 --> 01:48:45.980
I just submitted a paper that is just to be really quick.

01:48:46.140 --> 01:48:50.540
Um, I got really frustrated because I sent it was it was very much inspired by you, but

01:48:50.540 --> 01:48:55.580
inspired by you, Paul, because I sent I sent a paper to I'm gonna say perspectives on psych

01:48:55.660 --> 01:48:55.980
science.

01:48:56.060 --> 01:48:59.020
It's good, you know, it had a lot of theory, theory crisis.

01:48:59.180 --> 01:49:03.930
Um, no, it didn't go, but it was the editor desk rejected it for reasons that I was like,

01:49:04.010 --> 01:49:07.210
wow, this is selecting against theory, good theory.

01:49:07.450 --> 01:49:11.450
And so I so I actually wrote a rant paper, I submitted to perspective.

01:49:11.530 --> 01:49:17.450
So here's the premise: is that um if you ask for data concurrent with

01:49:17.450 --> 01:49:23.290
theory, uh, so if you select theory based on whether it it comes along with

01:49:23.290 --> 01:49:29.280
data with the theory, it selects for bad theory because

01:49:29.840 --> 01:49:34.560
the it selects for basically the kinds of theories that it's easy to generate data about.

01:49:34.720 --> 01:49:38.240
Um, so you're actually or at least theory that you don't need necessarily.

01:49:38.400 --> 01:49:43.760
It could be you don't need so like so for precisely theories that are about difficult

01:49:43.760 --> 01:49:49.410
things that we are not even sure how to well, like challenges to like we we I have a theory

01:49:49.410 --> 01:49:53.250
that there is this mechanism and there's a deductive logic that like there should be this

01:49:53.330 --> 01:49:54.770
like parenting mechanism, right?

01:49:55.010 --> 01:49:59.490
It would be impossible for organisms to exist if it didn't exist, but we don't know how to

01:49:59.490 --> 01:50:00.210
look for it yet.

01:50:00.370 --> 01:50:01.650
We don't know how to study it.

01:50:01.890 --> 01:50:05.650
The idea that, you know, that gets selected, that paper gets selected out because it

01:50:05.650 --> 01:50:07.090
doesn't have any data with it, right?

01:50:07.330 --> 01:50:12.370
It it it falls out of the um it falls out of the narrative of what we're even discussing.

01:50:12.530 --> 01:50:17.560
So the idea is that I mean, uh the the short version is um asking for data concurrent with

01:50:17.560 --> 01:50:22.360
theory selects for bad theory and I sent it to perspectives and it was rejected because it

01:50:22.440 --> 01:50:23.960
it didn't have data.

01:50:25.640 --> 01:50:26.200
The ultimate.

01:50:26.920 --> 01:50:27.080
Yeah.

01:50:27.160 --> 01:50:28.280
So so I love this.

01:50:28.360 --> 01:50:32.280
And so this is a very much inspired by Paul Smaldino I think, but it comes back to this

01:50:32.280 --> 01:50:35.800
thing of um that, you know, it's just what are you selecting for?

01:50:35.960 --> 01:50:36.120
Right.

01:50:36.770 --> 01:50:42.130
No, and we have you from I mean this is also, you know, my sort of push to like be deep in

01:50:42.130 --> 01:50:46.370
your field, but like look at other fields too, because otherwise you get totally insular

01:50:46.450 --> 01:50:52.050
and you you you you allow things to become normative uh a and that

01:50:52.290 --> 01:50:53.170
shouldn't be.

01:50:53.490 --> 01:50:54.610
Um like that.

01:50:54.770 --> 01:51:00.530
I mean there's so many examples from other fields of theories being published

01:51:00.840 --> 01:51:06.280
and then you know not finding you know experimental evidence for them for years,

01:51:06.440 --> 01:51:07.560
sometimes decades.

01:51:07.880 --> 01:51:08.200
Yeah.

01:51:08.360 --> 01:51:08.760
Yeah.

01:51:09.160 --> 01:51:14.120
Paul, I want to just because we're running short on time or we're ending here, um, you we

01:51:14.120 --> 01:51:18.040
haven't, you know, we've we've ebbed in flowed in terms of like talking to guests about

01:51:18.040 --> 01:51:20.280
like themselves or like their their life.

01:51:20.440 --> 01:51:26.350
Um but I am curious because you know you have this um in my

01:51:26.430 --> 01:51:32.430
in my opinion, you know, you have like a very fierce uh scientific identity

01:51:32.510 --> 01:51:38.270
in the sense of, you know, you're you're really pushing not in like a um a pushy way, but

01:51:38.270 --> 01:51:41.550
you're um you're a spokesperson for doing models.

01:51:41.630 --> 01:51:44.670
I mean, you have papers that are like, you know, models are stupid and we need more of

01:51:44.670 --> 01:51:44.910
them.

01:51:45.070 --> 01:51:47.470
Like that's a very special sort of stance to take.

01:51:47.550 --> 01:51:52.020
Like I'm an expert in something and it's stupid, but it's a good thing.

01:51:52.180 --> 01:51:53.940
Like that's a very special thing.

01:51:54.100 --> 01:51:57.700
It feels very different to me than what we talked about before, which is sort of people

01:51:57.700 --> 01:52:00.740
have these status games and they're just like, I'm the best.

01:52:00.900 --> 01:52:04.260
Like to say something like, I'm an expert in something that's stupid and we need more of it

01:52:04.340 --> 01:52:05.460
is very subversive.

01:52:05.620 --> 01:52:11.620
So I'm just sort of curious, like we talked about um, you know, the punk rock uh East

01:52:11.620 --> 01:52:13.220
German thing.

01:52:14.070 --> 01:52:18.230
I mean, are have you always sort of wanted to is that you?

01:52:18.390 --> 01:52:20.470
Like are you the punk rocker in East Germany?

01:52:20.550 --> 01:52:21.350
Uh if for Saga.

01:52:21.750 --> 01:52:23.350
Well punk rocker would never admit.

01:52:24.390 --> 01:52:25.430
Yeah, but well, you know.

01:52:26.550 --> 01:52:27.750
I yeah, I don't know.

01:52:27.830 --> 01:52:28.630
Uh I don't know.

01:52:28.710 --> 01:52:34.630
I mean I like I do think like a little bit a little bit of a punk attitude

01:52:34.710 --> 01:52:35.830
I think is healthy.

01:52:36.070 --> 01:52:41.020
And uh I I I definitely you know respond to that.

01:52:41.100 --> 01:52:45.180
And I also just feel like for me, like I've tried being more mainstream and it doesn't

01:52:45.180 --> 01:52:45.420
work.

01:52:45.660 --> 01:52:48.380
Like I don't feel comfortable and people reject it.

01:52:48.700 --> 01:52:53.740
Like uh I tried to give more like um when I was on the job market looking for academic

01:52:53.820 --> 01:52:57.900
permanent jobs, I tried to give much more typical jobs like the way that I had seen other

01:52:57.900 --> 01:53:00.780
people give them, and it just never worked, and I didn't get anything.

01:53:01.020 --> 01:53:04.050
And then I finally was just like, you know, screw it.

01:53:04.130 --> 01:53:07.970
I'm just gonna give the the the talk that I wanna give.

01:53:08.130 --> 01:53:13.810
And I I gave a talk at UC Merced called, I think the talk was called that I got that got me

01:53:13.810 --> 01:53:18.130
my job was called Learning About Social Evolution with Extremely Unrealistic Models.

01:53:19.250 --> 01:53:21.970
And I was like, these models are totally unrealistic.

01:53:22.290 --> 01:53:24.450
But let's see what we can learn from them anyway.

01:53:25.010 --> 01:53:30.960
And I think that, you know, we we we tell this like fake narrative that

01:53:31.040 --> 01:53:35.200
like we're the science we're doing is like so precise and we know so much.

01:53:35.280 --> 01:53:40.480
And I think like acknowledging that like we're flailing and we like it's not like people

01:53:40.480 --> 01:53:41.120
don't know anything.

01:53:41.200 --> 01:53:46.320
We know a lot of things, but there's also just so much we don't know and being a little m

01:53:46.480 --> 01:53:50.110
humble about that, I think it is actually a real strength.

01:53:50.270 --> 01:53:53.790
Gives, well, for me anyway, I feel like it gives me a power.

01:53:53.950 --> 01:53:59.550
Um, and I think also like I just my trajectory, like I studied

01:53:59.550 --> 01:54:00.990
physics in as an undergrad.

01:54:01.230 --> 01:54:06.510
I didn't take a single class on biology, cognitive science, or psychology until grad

01:54:06.510 --> 01:54:07.070
school.

01:54:07.310 --> 01:54:09.550
And when I did, I was like, really?

01:54:10.190 --> 01:54:13.710
This that we're are we this is what we're seriously doing?

01:54:13.870 --> 01:54:15.540
And I I love the topics.

01:54:15.620 --> 01:54:17.540
I love they're my favorite kinds of questions.

01:54:17.620 --> 01:54:21.700
I think like psychology, cognitive science, the social sciences have the best questions.

01:54:21.780 --> 01:54:23.780
They're the things I'm most interested in.

01:54:24.020 --> 01:54:28.260
And sometimes and there's it's not like everyone's doing crap work but me.

01:54:28.340 --> 01:54:29.700
Like, obviously, it's not what I'm saying.

01:54:29.780 --> 01:54:33.940
There's so many people doing work that I think is amazing and inspiring.

01:54:34.180 --> 01:54:38.580
But there's also like plenty of stuff where I'm like, come on, come on, guys.

01:54:38.900 --> 01:54:44.170
Um, and I think that, you know,

01:54:44.730 --> 01:54:48.170
being a little bit willing to push back on that.

01:54:49.530 --> 01:54:53.850
Also, if I'm gonna say other people's be, you know, critical of other people's work, then

01:54:53.930 --> 01:54:58.490
like it's beholden to me to be just as critical about my own work.

01:54:58.970 --> 01:55:04.730
And so I I really try to be like I am very careful and I haven't always

01:55:04.730 --> 01:55:10.720
succeeded, but I try you know, to be pretty cautious in the uh in the claims

01:55:10.720 --> 01:55:11.520
that I'm making.

01:55:11.760 --> 01:55:16.640
And uh and but that doesn't mean that I don't I love speculating.

01:55:16.800 --> 01:55:20.240
Like I it hate I hate it when people are afraid to speculate because they're like, well, I

01:55:20.240 --> 01:55:21.360
don't know if I don't have the evidence.

01:55:21.520 --> 01:55:25.840
And I'm like, right, that's why you speculate, and then you see maybe you're right and

01:55:25.840 --> 01:55:26.560
maybe you're wrong.

01:55:26.640 --> 01:55:31.590
Like, but if you are an expert, then like your speculation should be better than people who

01:55:31.590 --> 01:55:32.950
are not an expert, right?

01:55:33.110 --> 01:55:38.710
So like the stuff we can prove is such a small amount of the things, a small subset of the

01:55:38.710 --> 01:55:40.150
things we actually want to talk about.

01:55:40.230 --> 01:55:40.550
Yeah.

01:55:40.870 --> 01:55:46.070
So um I I think that like I'm all for like,

01:55:46.870 --> 01:55:48.310
you know, I don't know.

01:55:48.550 --> 01:55:53.830
Um I I I like I don't think I'm that special in the like sort of

01:55:54.860 --> 01:55:58.620
field of scientists, but I have carved out my little niche.

01:55:58.860 --> 01:56:04.860
And uh I I just keep you know, I'll just

01:56:04.860 --> 01:56:08.780
say the last thing I'll say is like I've I've done this thing, this stuff on like working

01:56:08.780 --> 01:56:13.740
on models of science and talking about like science and improving science, and I and it's

01:56:13.740 --> 01:56:15.100
my most cited stuff.

01:56:15.260 --> 01:56:20.140
Like the natural selection of bad science, that paper is by far my most cited paper.

01:56:20.380 --> 01:56:25.930
Um, but the idea that working on science itself is like

01:56:26.250 --> 01:56:29.770
my main life's mission is is like I reject that.

01:56:29.930 --> 01:56:31.530
I don't want that to be the case.

01:56:31.690 --> 01:56:36.410
And I I'm happy to be known for that, but I um there's so many other things.

01:56:36.570 --> 01:56:40.810
Like I'm interested in culture and the world and human behavior more generally.

01:56:40.890 --> 01:56:44.090
And it's just you know, that's just one aspect of it.

01:56:44.330 --> 01:56:48.860
And I I um I don't know.

01:56:49.020 --> 01:56:50.540
I don't know where I'm going with that.

01:56:50.700 --> 01:56:51.580
But uh Yeah.

01:56:51.740 --> 01:56:57.340
I I I also think this all comes back to like the original identity thing where we're all

01:56:57.340 --> 01:57:02.060
looking around and like psychologists are like, it's sort of the emperor has no clothes.

01:57:02.140 --> 01:57:07.100
We're sort of you're you're anticipating like what's gonna keep what's gonna give me

01:57:07.100 --> 01:57:09.740
status, how are other people gonna react to stuff?

01:57:10.060 --> 01:57:14.940
And there's this, I think there's a paper about this where it's sort of the surprisingly

01:57:14.940 --> 01:57:20.930
popular thing where um I remember we had um for our wedding meal we asked people like we

01:57:20.930 --> 01:57:24.610
have to have like these three entrees and and we had like I don't know something like quail

01:57:24.930 --> 01:57:29.730
and everyone was like oh I would love quail but I don't think anyone else is going to and

01:57:29.890 --> 01:57:33.970
that it turned out to be a really good idea that we had quail for the meal because

01:57:33.970 --> 01:57:36.770
everybody wanted but nobody thought anyone else did what they thought they did.

01:57:36.850 --> 01:57:40.370
And what this paper is about surprisingly popular is something where you respond to

01:57:40.370 --> 01:57:42.370
something but you don't think anyone else will.

01:57:42.720 --> 01:57:47.760
And I think like the the value of like having a paper like models are stupid, we should

01:57:47.760 --> 01:57:53.280
have more of them, is it removes that, it's it's sort of that courage to sort of be like,

01:57:53.680 --> 01:57:57.600
um, it's it's standing out, like it's it's rocking the boat a little bit.

01:57:57.760 --> 01:58:01.680
But then but then people can be like, oh, there is somebody else who's like these things

01:58:01.680 --> 01:58:03.590
are useful, but they are kind of stupid.

01:58:03.670 --> 01:58:08.310
And we don't have to like pretend like everybody agrees that they're not stupid.

01:58:08.470 --> 01:58:14.230
You know, and maybe it it kind of um it g it gives some error to

01:58:14.390 --> 01:58:19.030
otherwise everybody's like pretending like, well, I think everybody must think they're

01:58:19.030 --> 01:58:19.270
right.

01:58:19.430 --> 01:58:21.910
I might think they're stupid, but nobody else does.

01:58:22.150 --> 01:58:26.390
And you know, you being a little bit courageous to be like, Well, they're dumb, but let's

01:58:26.390 --> 01:58:26.870
use them.

01:58:27.030 --> 01:58:30.550
Like, I think that helps people realize that they're not alone.

01:58:30.900 --> 01:58:33.540
Yeah, I I think I think that's right.

01:58:33.700 --> 01:58:38.740
Um I just I can't remember what book it's in, but there was this I remember a long time ago

01:58:38.740 --> 01:58:42.500
reading this this something by Kurt Vonnegut where he talks about in one of his books and

01:58:42.500 --> 01:58:48.180
about how like one of the great things about writing is that like somebody you've

01:58:48.180 --> 01:58:53.540
never met can can you know read something that like expresses a perspective on the world

01:58:53.540 --> 01:58:54.820
and says hey me too.

01:58:54.980 --> 01:58:56.020
I think that way too.

01:58:56.710 --> 01:59:01.910
that like that feeling of, hey, I'm not alone in thinking this is hugely powerful.

01:59:01.990 --> 01:59:05.830
And it's like, you know, one of the things that really like ties me, and I think a lot of

01:59:05.830 --> 01:59:10.470
us to like the works of art that we love and the works of writing fiction or non that we

01:59:10.470 --> 01:59:10.710
love.

01:59:10.790 --> 01:59:15.990
And so I think like, you know, to the extent that like I can be that for even like a

01:59:15.990 --> 01:59:20.070
handful of people in, you know, is is that that's huge to me.

01:59:20.150 --> 01:59:21.500
I I love Yeah.

01:59:22.460 --> 01:59:23.820
Well it it was for me, honestly.

01:59:23.900 --> 01:59:27.420
Like honestly the the the like I'm like this is ridiculous.

01:59:27.580 --> 01:59:31.980
Um there's good work and there's clearly selection for status.

01:59:32.220 --> 01:59:38.220
Um it was really uh it it it it kept me in this to to to know that other people

01:59:38.220 --> 01:59:39.340
were thinking the same way.

01:59:40.060 --> 01:59:46.010
To tie it back full circle it is perhaps the feeling of a covert signal well

01:59:46.090 --> 01:59:46.730
received.

01:59:47.850 --> 01:59:48.170
Exactly.

01:59:48.570 --> 01:59:52.330
And to be to be part of a community that you didn't know existed.

01:59:53.850 --> 01:59:59.690
So Paul, is there anything uh that we should have asked you uh during this

01:59:59.690 --> 01:59:59.930
time?

02:00:00.090 --> 02:00:01.930
Or did we did we get it all?

02:00:02.810 --> 02:00:08.410
I mean, I'm sure that you got a very small amount, but because we all

02:00:08.410 --> 02:00:12.560
contain multitudes, but uh I feel very, very happy with the conversation.

02:00:12.640 --> 02:00:13.760
I thought I I had a lot of fun.

02:00:14.000 --> 02:00:14.160
Yeah.

02:00:15.200 --> 02:00:21.040
Can you uh um uh sh t demonstrate at least one thing behind like what there for the people

02:00:21.040 --> 02:00:23.600
who are just listening, but like you know, we also are on YouTube.

02:00:24.000 --> 02:00:27.840
Um, you have guitars and you've got a solar system or something behind you.

02:00:28.000 --> 02:00:30.720
What is the well I'm not gonna play guitar on the podcast?

02:00:31.200 --> 02:00:33.200
Okay, no, that's uh that's for subscribe.

02:00:33.440 --> 02:00:36.080
We don't have subscribers, but if we did, that would be bonus content.

02:00:36.160 --> 02:00:37.440
You have to pay to listen to Paul.

02:00:38.000 --> 02:00:43.990
Uh you you can um the the some of the music by my old uh garage band, the

02:00:43.990 --> 02:00:48.310
small dinosaurs, is available on so it's called small dinosaurs.

02:00:48.470 --> 02:00:49.750
I will put a link in the show notes.

02:00:49.830 --> 02:00:51.110
So it's small dinosaurs.

02:00:51.430 --> 02:00:51.750
Yeah.

02:00:51.910 --> 02:00:52.150
Okay.

02:00:52.390 --> 02:00:54.870
Um and then this is a Lego technic.

02:00:54.950 --> 02:01:00.230
It's uh sun and moon and uh an earth, and the earth spins and the sun spins, and the earth

02:01:00.230 --> 02:01:04.880
goes around the sun, and the moon goes around the earth, and I'm twisting a little lever,

02:01:05.040 --> 02:01:09.840
and it can show you like where the position the relative positions of the sun, the moon,

02:01:10.080 --> 02:01:12.400
and the earth are at different times of year.

02:01:12.640 --> 02:01:12.960
Awesome.

02:01:13.200 --> 02:01:17.760
And uh it's it's a very old technology, but Lego made one and I got it for Christmas.

02:01:17.920 --> 02:01:19.440
And we were talking about it earlier.

02:01:19.600 --> 02:01:20.720
I think it's really neat.

02:01:21.200 --> 02:01:22.880
Um so you know, props to Lego.

02:01:23.200 --> 02:01:24.640
So awesome.

02:01:24.800 --> 02:01:25.280
Yeah.

02:01:25.650 --> 02:01:27.650
Lego, please sponsor the podcast.

02:01:29.490 --> 02:01:30.290
Oh, look at that.

02:01:30.450 --> 02:01:32.210
At the very end, I'm shilling to corporate.

02:01:32.290 --> 02:01:33.250
Yeah, anyway, but yeah.

02:01:33.570 --> 02:01:36.450
I mean, if you're gonna shill to a corporation, Lego's a pretty good one.

02:01:36.770 --> 02:01:42.450
Yeah, I mean a lot of uh a lot of wonderful changes to human brains done t by Lego,

02:01:42.850 --> 02:01:44.530
you know, and that's that's saying a lot.

02:01:44.850 --> 02:01:46.610
That sounds like really dark.

02:01:47.490 --> 02:01:49.330
It's a very reductive way to put it.

02:01:49.970 --> 02:01:53.730
I don't know if you know about Lego's secret laboratory where they're uh changing people's

02:01:53.730 --> 02:01:53.890
brains.

02:01:54.610 --> 02:01:55.860
It's the you know, exactly.

02:01:56.020 --> 02:01:59.700
I'm just the you know, I don't know, like Coca-Cola's just dump people full of sugar.

02:01:59.780 --> 02:02:03.300
Lego, it's like it's you know, it's like it's been but yeah.

02:02:03.380 --> 02:02:06.660
I feel like I kind of think in Legos and I feel like that's a good thing.

02:02:06.820 --> 02:02:10.900
I don't know, to to think in terms of how the the pieces fit together and how it works.

02:02:10.980 --> 02:02:11.380
I don't know.

02:02:11.780 --> 02:02:17.620
And in like perseverance and planning and mental rotation and creativity and autonomy

02:02:17.780 --> 02:02:19.060
and all those things.

02:02:19.220 --> 02:02:20.740
But yeah, no, no, fair enough.

02:02:20.820 --> 02:02:21.140
All right.

02:02:21.300 --> 02:02:24.530
Well, uh maybe on that note, I don't know.

02:02:24.610 --> 02:02:26.450
Paul, do you want to have the last word?

02:02:27.810 --> 02:02:28.610
Potato.

02:02:28.850 --> 02:02:29.170
Awesome.

02:02:29.330 --> 02:02:29.490
All right.

02:02:29.650 --> 02:02:29.970
On that note.

02:02:31.810 --> 02:02:32.530
All right, Paul.

02:02:32.610 --> 02:02:36.930
Yeah, I'll hit I'll hit um I'll stop recording, but uh this is great, man.

02:02:37.090 --> 02:02:37.490
Yep.

02:02:37.650 --> 02:02:38.210
Oh yeah.

02:02:38.290 --> 02:02:38.530
Thanks.

02:02:38.690 --> 02:02:39.490
This is fun.

02:02:42.370 --> 02:02:45.410
David, I really enjoyed this conversation with Paul.

02:02:45.650 --> 02:02:46.850
Did you enjoy it?

02:02:47.100 --> 02:02:47.740
I loved it.

02:02:47.900 --> 02:02:48.540
I had a blast.

02:02:48.860 --> 02:02:49.260
All right.

02:02:49.420 --> 02:02:53.820
Um, and if you did enjoy it, please help us by spreading the word.

02:02:53.980 --> 02:02:59.340
If you just tell one other person, then um then they'll might listen to it as well.

02:02:59.500 --> 02:03:05.500
I can't tell you how many people who I've emailed as like even a um a guest to come on who

02:03:05.500 --> 02:03:08.140
said I don't know, I didn't know about the podcast.

02:03:08.300 --> 02:03:14.170
So um, and then they listen to it and at least um you know 90% of the

02:03:14.170 --> 02:03:16.810
time they say, Oh, it actually was pretty good.

02:03:17.210 --> 02:03:17.930
Yeah, that's true.

02:03:18.010 --> 02:03:21.450
I mean, unless they're all lying to us uh to so they don't hurt our feelings.

02:03:21.530 --> 02:03:25.770
Uh yeah, maybe the people who never got back to me are like that was bad.

02:03:25.930 --> 02:03:29.770
I mean, occasionally people spl leave bad reviews, but that's okay.

02:03:30.250 --> 02:03:34.260
But if you're one of the people who does enjoy the podcast and is willing to tell us the

02:03:34.260 --> 02:03:39.460
truth about that, um then maybe share that truth with others by uh

02:03:39.860 --> 02:03:45.220
saying a kind word about us on social media, leaving us a kind review, visiting us on

02:03:45.220 --> 02:03:46.100
Patreon.

02:03:46.260 --> 02:03:52.100
Uh there are many ways to show your support, and uh your support is what keeps us going

02:03:52.100 --> 02:03:53.300
and keeps this thing going.

02:03:53.460 --> 02:03:54.500
So please show it.

02:03:54.820 --> 02:03:56.260
That and caffeine.

02:03:56.500 --> 02:03:56.900
Yeah.

02:03:57.140 --> 02:04:00.790
You You'd send us some coffee grounds, uh, preferably in a fanny pack.

02:04:00.870 --> 02:04:02.550
That would be fantastic.

02:04:04.950 --> 02:04:10.150
And we'll leave you with that image until we see you next week or hear from you next week.

02:04:10.790 --> 02:04:11.750
See you next week.