#108 - Dr. Matt Clancy - The Economics of Innovation, AI, and Collaboration

WEBVTT
00:00:00.000 --> 00:00:09.040

00:00:09.383 --> 00:00:12.423
Welcome to Direction Correct, a freelance podcast with Colin Scott.
00:00:12.723 --> 00:00:15.983
Today's guest, Matt Clancy, Research Fellow at Open Philanthropy.
00:00:17.383 --> 00:00:22.603
Thanks to our sponsors, OneModel. OneModel helps people leaders at large organizations
00:00:22.603 --> 00:00:26.743
make consistently brilliant talent decisions by unlocking the analytical value
00:00:26.743 --> 00:00:28.823
of the data dispersed across your business.
00:00:29.303 --> 00:00:33.343
OneModel's people analytics platform takes all of the heavy lifting out of data
00:00:33.343 --> 00:00:38.163
extraction, cleansing, modeling, analytics, and reporting of enterprise workforce data.
00:00:38.163 --> 00:00:42.443
One Model pioneered people data orchestration and has perfected the ethical
00:00:42.443 --> 00:00:46.603
use of AI and data science for leaders who need transparent and explainable
00:00:46.603 --> 00:00:48.103
decision support systems.
00:00:48.423 --> 00:00:53.363
HR and business teams trust its accurate reports, analyses, and storytelling capabilities.
00:00:53.703 --> 00:00:57.743
Data scientists, engineers, and people analytics professionals love the combination
00:00:57.743 --> 00:01:02.483
of governance and flexibility that no other Workforce Insights platform can provide.
00:01:02.703 --> 00:01:07.083
To learn more, book a demo at onemodel.co slash directionallycorrect.
00:01:07.663 --> 00:01:11.823
All opinions are our own and do not reflect those of any other organization
00:01:11.823 --> 00:01:17.983
there was a few students who like were like in it because they you know like
00:01:17.983 --> 00:01:21.703
wanted to like do something with their career or whatever yeah whereas a lot
00:01:21.703 --> 00:01:23.163
of people were like well you know
00:01:24.363 --> 00:01:27.363
i'm just gonna go live off my daddy's money now
00:01:27.363 --> 00:01:31.303
I'm just getting this like finance degree and i'm gonna go get like an immediate
00:01:31.303 --> 00:01:37.523
golden tax high-paying job and then like go to the golf course and like start it all over again
00:01:37.523 --> 00:01:38.343
Exactly.
00:01:38.343 --> 00:01:43.123
Kids mad hey guys what's happening man you're looking good you're looking
00:01:43.123 --> 00:01:49.403
Like haircut looking clipped up yeah are.
00:01:49.403 --> 00:01:52.523
You a work-at-home guy like you did primarily out of the
00:01:52.523 --> 00:01:58.323
House yeah i mean open phil is based in san francisco and i live in iowa so
00:01:58.323 --> 00:02:04.103
i get in to the office a couple times a year but but that's it i mean i'm not
00:02:04.103 --> 00:02:10.523
unusual open phil is like more than 50 probably people like me so iowans.
00:02:10.523 --> 00:02:13.003
Is that 50 is that what you
00:02:13.003 --> 00:02:17.283
Mean no i just mean remote i'm probably the only iowan yeah.
00:02:17.283 --> 00:02:18.743
How's that culture clash
00:02:18.743 --> 00:02:23.943
Yeah yeah No, it's not bad because it's like an international kind of,
00:02:24.163 --> 00:02:27.903
or I don't know, national community, I guess, or work organization.
00:02:28.803 --> 00:02:31.803
What is it? Just out of curiosity. Yeah.
00:02:32.063 --> 00:02:36.723
Yeah. I mean, it's organization that gives away money is the bottom line.
00:02:36.903 --> 00:02:39.323
Wow. Yeah. So open philanthropy.
00:02:40.143 --> 00:02:43.763
And we have a bunch of programs in different areas. And I run the innovation
00:02:43.763 --> 00:02:45.903
policy one. So I make grants in that area.
00:02:46.923 --> 00:02:52.143
This is where I came across your name. I work with an economist here at my office,
00:02:52.143 --> 00:02:54.543
and he's like, hey, you got to check out Matt Clance.
00:02:54.663 --> 00:02:58.863
He's doing really cool research, essentially aggregating various...
00:02:58.863 --> 00:03:03.063
I imagine you read a million articles a day, and you just get a really good
00:03:03.063 --> 00:03:05.023
overview of the entire field.
00:03:05.583 --> 00:03:10.663
And I found your sub stack, Everything Under the Sun, is just wildly impressed by it.
00:03:11.383 --> 00:03:14.883
Well, thanks. Yeah. I mean, that is kind of the idea. So before I was at this
00:03:14.883 --> 00:03:18.063
job, I was a professor of economics, Iowa State University.
00:03:19.443 --> 00:03:23.503
And yeah that was kind of the idea is like there's like
00:03:23.503 --> 00:03:26.323
like academia is not super well
00:03:26.323 --> 00:03:30.963
set up for to like see the forest for the trees like uh you're kind of rewarded
00:03:30.963 --> 00:03:34.823
for doing research which means you've got to like super specialize and like
00:03:34.823 --> 00:03:38.283
be on this narrow slice the best in the world at it and then push it a little
00:03:38.283 --> 00:03:44.103
bit further and so i thought maybe there was i plus i really like it you know,
00:03:44.183 --> 00:03:45.183
to try and do something different.
00:03:45.503 --> 00:03:49.163
And so created this sub stack, which kind of took off and was really popular.
00:03:49.403 --> 00:03:51.403
And now it's a website too. And yeah,
00:03:52.135 --> 00:03:57.915
I still spend, even though I'm not at a university, I negotiated when I joined,
00:03:58.015 --> 00:04:00.675
30% of my time is on this project still.
00:04:00.975 --> 00:04:06.775
So it's not just like a side hustle or something. It's something I spend significant time on.
00:04:07.115 --> 00:04:09.435
Did you know Marcus Corday over at Iowa State?
00:04:09.875 --> 00:04:11.535
No, I don't think so. What department was he in?
00:04:12.255 --> 00:04:14.995
He's probably business. School of Business, Cole. Is that right?
00:04:15.495 --> 00:04:16.235
I don't know.
00:04:16.875 --> 00:04:19.855
Yeah, it's like across the street, but didn't connect.
00:04:20.615 --> 00:04:24.575
I mean iowa state's a beautiful beautiful place like he came in here and like
00:04:24.575 --> 00:04:28.255
he was essentially saying like 30 of all io research essentially is made up
00:04:28.255 --> 00:04:31.455
like you can't trust it like really coming with a
00:04:31.455 --> 00:04:33.375
Blowtorch of which kind of research.
00:04:33.375 --> 00:04:35.575
Io psychology yeah
00:04:35.575 --> 00:04:40.195
Well i don't know the field but there was this there's an article in the atlantic
00:04:40.195 --> 00:04:46.895
right now about you know fallout from all this fraud and then this this researcher
00:04:46.895 --> 00:04:50.715
who discovers that one of her co-authors is you know been.
00:04:50.715 --> 00:04:53.635
Oh don't worry yeah we are all over
00:04:53.635 --> 00:05:00.775
that okay i've been talking about it for probably a year now but i did see that
00:05:00.775 --> 00:05:04.455
one in the atlantic and it's just like it just keep there they keep peeling
00:05:04.455 --> 00:05:09.255
back the layers of the onion and it just keeps getting more juicy and it's just
00:05:09.255 --> 00:05:11.115
like yeah everything is wrong with everything.
00:05:11.315 --> 00:05:15.315
Yeah. I mean, the twist at the end, right? That the sort of main person there
00:05:15.315 --> 00:05:20.455
who's leading the investigation is themselves possibly engaged in fraud.
00:05:22.415 --> 00:05:23.315
We'll see. Yeah.
00:05:24.300 --> 00:05:26.640
Well, it's like, at a certain point,
00:05:26.960 --> 00:05:32.940
You know, is it systemic, right? Is like, is if everybody's doing it, you know, I don't know.
00:05:33.200 --> 00:05:39.740
Like I had this conversation with some academics a few years ago and they were
00:05:39.740 --> 00:05:45.080
just talking to me pretty openly about kind of how, like, cause they were just
00:05:45.080 --> 00:05:47.920
trying to get published, right? Like their whole goal was to try to get published.
00:05:48.100 --> 00:05:51.700
And they were just talking very openly about like
00:05:51.700 --> 00:05:54.480
the incentives and how misaligned they were and
00:05:54.480 --> 00:05:57.540
like how they were going to have to like kind of skirt the edges
00:05:57.540 --> 00:06:01.040
of the rules to get this by and i was like what the
00:06:01.040 --> 00:06:04.020
f are you guys talking about like yeah like i i
00:06:04.020 --> 00:06:07.740
immediately lost respect for them the thing i realized is as i reflect on it
00:06:07.740 --> 00:06:12.280
over time and as again a lot of these articles have come out is oh they're talking
00:06:12.280 --> 00:06:17.360
about like this is just common practice like this isn't like and so ever since
00:06:17.360 --> 00:06:22.440
that conversation i've you know scott's heard me talk about the open science movement and, and,
00:06:22.620 --> 00:06:27.480
you know, how we need to do pre-publications and all the stuff that's been going on in that space.
00:06:27.480 --> 00:06:30.920
And it's just like, I don't know, it seems endemic to a lot of the sciences
00:06:30.920 --> 00:06:32.320
now. And that's really concerning.
00:06:32.800 --> 00:06:37.340
Yeah. I mean, I actually looked into this a little bit recently and,
00:06:37.340 --> 00:06:42.720
you know, people trying to estimate kind of how prevalent is fraud in different academic fields.
00:06:42.880 --> 00:06:45.540
And there's kind of like two approaches people have taken, which is It's like
00:06:45.540 --> 00:06:50.900
they look at a big set of papers and then they, in some disciplines,
00:06:51.240 --> 00:06:55.820
like you can detect when people are manipulating the images in their,
00:06:55.820 --> 00:06:58.400
like, you know, biology, they're sort of.
00:06:58.540 --> 00:07:00.700
What are you talking about? Like the Stanford president, right?
00:07:00.940 --> 00:07:04.400
Yeah. Yeah. So, so, you know, people like there's this amazing person,
00:07:04.580 --> 00:07:10.220
Elizabeth Bick, who, who can, who is like just a gift for spotting sort of fraud in, in,
00:07:10.480 --> 00:07:15.820
or like article images when they've been doctored or when they're being duplicated
00:07:15.820 --> 00:07:21.220
and sort of rotated to disguise it and things like that. So she looked at like 20,000 papers.
00:07:21.920 --> 00:07:29.240
And I think it was something like 2% of them had one of these kind of images in it.
00:07:30.380 --> 00:07:33.700
And then there's surveys of academics like, have you ever done fraud?
00:07:34.060 --> 00:07:37.440
And some of those find stuff in sort of the same ballpark where people say like,
00:07:37.520 --> 00:07:40.260
well, 2% of people will say they've done something like that.
00:07:40.340 --> 00:07:43.060
But you never know if that's like an under those could both be underestimates
00:07:43.060 --> 00:07:47.060
right because it's just this one kind of fraud in one case it's like this images
00:07:47.060 --> 00:07:51.620
that you detect and then in the other case it's anybody who feels comfortable
00:07:51.620 --> 00:07:54.140
admitting in an anonymous survey that they did this so.
00:07:54.140 --> 00:07:57.400
It takes a special kind of person to like fabricate their data and be like i'm
00:07:57.400 --> 00:08:03.180
gonna do some sweet data viz here all right you're really gonna like showcase this yeah yeah
00:08:04.667 --> 00:08:09.147
Well, Matt, coming back to your Substack for a second. So I publish every so
00:08:09.147 --> 00:08:11.127
often an article or two on Substack.
00:08:11.227 --> 00:08:15.727
And I looked through yours earlier today and I was like, oh, my God, I suck so bad.
00:08:16.767 --> 00:08:20.987
Like the quality of your work is is just phenomenal.
00:08:21.307 --> 00:08:26.047
And so I guess the question in the actually the most impressive part to me was
00:08:26.047 --> 00:08:31.387
it covered such a variety of different topics that seem somewhat nonrelated.
00:08:31.647 --> 00:08:33.907
And so I'm like, this guy just knows everything about everything.
00:08:33.907 --> 00:08:36.147
And it is called everything under the sun.
00:08:36.307 --> 00:08:41.147
I know no pun intended there, but how, how do you, this is a dumb question.
00:08:41.287 --> 00:08:45.087
How do you do that? Like, how do you know everything under the sun and how is
00:08:45.087 --> 00:08:46.427
it so technically oriented?
00:08:46.767 --> 00:08:50.887
All right. So first correction, it's called new things under the sun.
00:08:51.047 --> 00:08:55.567
It's like a pun on this old biblical verse. Like there is no new thing under the sun.
00:08:55.707 --> 00:08:59.167
And so it's supposed to be about innovation and it's kind of like a cheeky thing
00:08:59.167 --> 00:09:03.287
about these are new things under the sun and yeah
00:09:03.287 --> 00:09:06.127
how does it so you know i i did
00:09:06.127 --> 00:09:09.087
a phd in economics this was sort of my specialty the economics of
00:09:09.087 --> 00:09:12.287
science and innovation i finished in 2015 i was
00:09:12.287 --> 00:09:18.387
always really interested in like making connections between distant stuff so
00:09:18.387 --> 00:09:23.007
one of my district like two of my dissertation papers were about that idea like
00:09:23.007 --> 00:09:24.907
understanding innovation as when
00:09:24.907 --> 00:09:29.247
you spot a connection between previously unconnected comments or concepts.
00:09:30.267 --> 00:09:34.547
And so I kind of, I think for a long time, I just thought it was valuable to
00:09:34.547 --> 00:09:37.687
read widely for that reason, because maybe you'll spot something.
00:09:38.667 --> 00:09:43.707
And then when I started the Substack, you know, I had been reading a lot for
00:09:43.707 --> 00:09:45.627
a long time and I had sort of taken notes on it.
00:09:46.770 --> 00:09:48.350
And so I think that helped.
00:09:48.630 --> 00:09:53.990
And as for why it jumps around, I think that that's to keep it interesting for
00:09:53.990 --> 00:09:55.530
readers and for me, basically.
00:09:55.930 --> 00:10:00.830
And also because I sort of have this view that maybe you'll see something valuable
00:10:00.830 --> 00:10:03.450
if you have that approach that other people are missing.
00:10:03.770 --> 00:10:06.650
I don't know if I have, but that was the theory anyway.
00:10:07.450 --> 00:10:12.170
We're reaching a point where a lot of organizations are mandating return to office.
00:10:12.390 --> 00:10:16.590
We had this four-year period where it's either, please don't come in.
00:10:16.590 --> 00:10:21.250
All the way to like, hey, we're kind of encouraging to come in and now you definitely have to come in.
00:10:22.530 --> 00:10:27.870
What does the research say about how that's going to impact innovation overall?
00:10:28.950 --> 00:10:32.010
Yeah, so this is an area I was interested in for a long time,
00:10:32.130 --> 00:10:37.410
even before COVID, because I live in Iowa, which is not like in the center of
00:10:37.410 --> 00:10:39.690
innovation and frontier stuff.
00:10:39.810 --> 00:10:42.930
So I was always interested in how connected can you be remotely?
00:10:42.930 --> 00:10:49.030
And I think that there's kind of different factors that pull in different directions.
00:10:49.050 --> 00:10:52.110
So I think everyone, it's easy to think of the things why it would be good to
00:10:52.110 --> 00:10:53.590
have people return to the office.
00:10:54.490 --> 00:10:59.490
You have these serendipitous water cooler conversations where two people make
00:10:59.490 --> 00:11:02.830
these connections between ideas they didn't expect because one person in the
00:11:02.830 --> 00:11:04.730
office knows something and another person doesn't.
00:11:05.910 --> 00:11:09.570
And there is research kind of documenting stuff like that.
00:11:09.570 --> 00:11:14.330
So, you know, they've done studies on academics, for example,
00:11:14.530 --> 00:11:19.270
and when two academics collaborate, what's the probability that one of those
00:11:19.270 --> 00:11:22.730
guys goes on to work on a new topic afterwards after their collaboration?
00:11:22.730 --> 00:11:24.470
So maybe they learned something from their peer.
00:11:24.550 --> 00:11:28.050
And that's more likely if they collaborate on an article and they're based in
00:11:28.050 --> 00:11:29.810
the same place than if they're at distant places.
00:11:29.810 --> 00:11:36.330
And there's other research about, you know, if I cite a paper that is kind of
00:11:36.330 --> 00:11:39.270
a weird paper for me to cite, like it's not just standard,
00:11:39.850 --> 00:11:47.010
you know, in my field, it's more likely that that paper is by somebody in my university.
00:11:47.970 --> 00:11:52.110
And so that's one of the advantages of like bringing people into the office.
00:11:52.290 --> 00:11:54.990
A second advantage is that it
00:11:54.990 --> 00:11:59.750
maybe is easier to just work together on projects that are sort of like.
00:12:00.250 --> 00:12:01.710
Like collaborative demand.
00:12:01.970 --> 00:12:07.230
Exactly. Like whiteboard type stuff. I don't think, so I don't know that literature as well, but.
00:12:08.304 --> 00:12:13.224
The kind of stuff that I've reviewing a lot of this emphasizes that in-person
00:12:13.224 --> 00:12:17.364
is really good for introducing people to each other who don't know each other.
00:12:18.144 --> 00:12:25.024
But then with modern technology, it's not as hard as it used to be to stay,
00:12:25.304 --> 00:12:26.524
to maintain relationships with
00:12:26.524 --> 00:12:30.384
people once you've had that kind of first connection, even at a distance.
00:12:30.384 --> 00:12:37.324
So it's not easy to see among people who already know each other that their
00:12:37.324 --> 00:12:42.364
ability to collaborate effectively declines noticeably at a distance.
00:12:43.904 --> 00:12:49.044
So I think that's the rub, right? Where people say like, they get shit mad.
00:12:49.224 --> 00:12:52.004
Like, I am fine working at home. I know all these people.
00:12:52.224 --> 00:12:56.104
I can, you know, just zoom them in. I don't have to like put on pants today.
00:12:56.684 --> 00:12:59.704
And like, you know, they essentially say like, I don't need to come into the
00:12:59.704 --> 00:13:03.464
office. But there are subtle differences in how people interact.
00:13:04.424 --> 00:13:08.704
Yeah. I mean, I think that you certainly, if you're going to have everybody
00:13:08.704 --> 00:13:15.984
work from home, you certainly want to do something to make those kind of connections
00:13:15.984 --> 00:13:17.484
that would happen in the office happen.
00:13:17.664 --> 00:13:23.944
You don't want to just like let everybody's network ossify and nobody's making
00:13:23.944 --> 00:13:25.324
new connections anymore. And then
00:13:25.324 --> 00:13:29.464
as people leave, your network sort of actually starts to shrink maybe.
00:13:30.784 --> 00:13:33.824
Siloed exactly and i think that covid showed
00:13:33.824 --> 00:13:37.244
that those kind of concerns were real
00:13:37.244 --> 00:13:40.404
like there was a famous study of sort of microsoft employees
00:13:40.404 --> 00:13:43.984
and when they went remotely you know they they just were much more likely to
00:13:43.984 --> 00:13:48.944
communicate with people they already knew and so forth so you know our org is
00:13:48.944 --> 00:13:53.384
mostly remote and we have like four weeks a year that are coordinated for people
00:13:53.384 --> 00:13:59.064
to come into the office and try to all be there at the same time so that it's it's like a way to you.
00:13:59.064 --> 00:14:02.984
Know to be intentional about collaborating, get together.
00:14:03.604 --> 00:14:06.784
But as I was saying, there's actually like, so there's, there's factors that
00:14:06.784 --> 00:14:10.104
make it valuable to be in person. And then there's other factors that make it
00:14:10.104 --> 00:14:13.844
valuable to be able to work remotely. And I think the chief one is just about recruiting.
00:14:15.364 --> 00:14:19.884
So if you have to recruit locally, you know, you, you are limited to who's in
00:14:19.884 --> 00:14:23.664
your local labor market and, or you have to entice somebody to move.
00:14:24.424 --> 00:14:27.744
And if you are able to hire remotely you
00:14:27.744 --> 00:14:30.544
can hire people who are the
00:14:30.544 --> 00:14:33.544
best at that position in the world and and it's
00:14:33.544 --> 00:14:36.464
it's a trade-off right like maybe they're 90 as effective as
00:14:36.464 --> 00:14:39.864
if they're in the office but if they're you know 20 better
00:14:39.864 --> 00:14:43.944
then maybe you will take that trade and it's
00:14:43.944 --> 00:14:47.444
kind of interesting like academia and in
00:14:47.444 --> 00:14:50.484
innovation in general so among academics and inventors
00:14:50.484 --> 00:14:56.664
the share of papers or patents that are developed by geographically distributed
00:14:56.664 --> 00:15:02.624
teams is just like continually marching up and i think that's all about this
00:15:02.624 --> 00:15:06.344
we need the specialist guy who knows about this one niche thing and he doesn't
00:15:06.344 --> 00:15:07.684
work here so we're going to,
00:15:08.592 --> 00:15:13.872
fly out there and have a weekend retreat or, and then keep up with zoom and stuff like that.
00:15:14.392 --> 00:15:18.612
One of the things I think that you pointed out about, you know,
00:15:18.712 --> 00:15:24.252
remote versus moving people cross country is I think one of the things that
00:15:24.252 --> 00:15:27.632
was lost during COVID is prior to COVID times,
00:15:28.252 --> 00:15:32.652
organizations spent hefty sums of money moving people around or doing,
00:15:32.652 --> 00:15:36.452
you know, offsites or what have you to make sure that you had the people that
00:15:36.452 --> 00:15:40.312
you want to be in a physical location in the physical location with one another.
00:15:41.072 --> 00:15:43.992
And then they didn't have to pay for that kind of stuff for like two years.
00:15:44.572 --> 00:15:47.752
And so that got taken as a line item out of their budgets.
00:15:48.332 --> 00:15:51.392
But then they came on the other side. Well, it's like, we want to have all the
00:15:51.392 --> 00:15:54.892
same collaboration in person time, but we're just not going to pay you to move cross country.
00:15:54.992 --> 00:15:57.232
We're not going to pay for those off sites anymore.
00:15:58.072 --> 00:15:59.972
And it's like, well, that's not going to work.
00:16:00.992 --> 00:16:08.132
And so it's, I think the, the, the idea has reverted back to the mean.
00:16:08.352 --> 00:16:12.352
But the budget to do so hasn't gone along correspondingly.
00:16:12.452 --> 00:16:17.832
And I think that is a healthy tension as well as just the in-person remote tension
00:16:17.832 --> 00:16:21.052
that's going on at the same time. I don't know. Are you seeing that Matt at all?
00:16:22.152 --> 00:16:25.692
That's tough for me to like, I haven't seen papers, for example,
00:16:26.272 --> 00:16:29.032
documenting or like looking into it. Not that it doesn't happen.
00:16:29.192 --> 00:16:31.712
It's, it's not like they find, they don't find that when they look,
00:16:31.712 --> 00:16:35.572
it's just, I haven't seen papers that have looked at that particular question.
00:16:35.572 --> 00:16:38.672
And then in my personal experience, you
00:16:38.672 --> 00:16:42.132
know open fill is generous in helping people travel to
00:16:42.132 --> 00:16:45.272
the weekly reports i know they take
00:16:45.272 --> 00:16:49.232
care of us but yeah i think the like
00:16:49.232 --> 00:16:53.312
travel is is an interesting thing that often i feel like it's funny how it kind
00:16:53.312 --> 00:16:56.772
of got left out of the discussion i i think of the value of being in the office
00:16:56.772 --> 00:17:00.452
is like well you might be more productive but there's also like you know you
00:17:00.452 --> 00:17:04.872
might spend an extra it's kind of like extending your work day by an extra however
00:17:04.872 --> 00:17:08.372
long you commute which could be, you know, an hour a day for people.
00:17:08.812 --> 00:17:13.792
And that sort of never counted in these papers as like a cost of,
00:17:14.813 --> 00:17:17.973
It's the productivity while you're in the office. And at the time that you're
00:17:17.973 --> 00:17:21.873
not in the office, it doesn't exist in some sense.
00:17:22.093 --> 00:17:23.993
We definitely hear this. That's
00:17:23.993 --> 00:17:27.353
one of the main complaints. Got you to get up earlier. Got to shower.
00:17:28.053 --> 00:17:32.473
Right. You got to come home. But I guess the flip side is that when you leave
00:17:32.473 --> 00:17:35.293
the office, your day is kind of done at that point, right?
00:17:35.633 --> 00:17:40.653
There's no just kind of like this weird sort of fringe bleed into the evening sort of situation.
00:17:40.653 --> 00:17:45.353
Yeah. And I think in the early days of COVID,
00:17:45.633 --> 00:17:49.453
when everyone was suddenly trying remote work for the first time,
00:17:49.593 --> 00:17:53.993
or at least a lot of people were, this was like a real focus of studies just
00:17:53.993 --> 00:17:59.053
sort of documenting this bleed out and the shifting of when people would work.
00:17:59.633 --> 00:18:02.673
And you know it can be can have advantages
00:18:02.673 --> 00:18:05.913
like i think there's some study i don't remember the name of it about like the
00:18:05.913 --> 00:18:10.313
number of people playing golf during the workday went up dramatically during
00:18:10.313 --> 00:18:13.793
covid you know but then some of these other studies document people working
00:18:13.793 --> 00:18:17.153
like after eight o'clock for an hour and answering emails and things like that
00:18:17.153 --> 00:18:22.153
so maybe they're shifting time or or maybe they're taking advantage of a difficulty
00:18:22.153 --> 00:18:25.253
being held accountable in some in some workplaces.
00:18:25.253 --> 00:18:28.313
I'm gonna blame cole cole's the one keeping the golf
00:18:30.233 --> 00:18:32.933
Well it was like it became impossible to get
00:18:32.933 --> 00:18:36.073
a tea time after because everybody was
00:18:36.073 --> 00:18:39.693
out there doing i'm like do you guys work like what it actually created a lot
00:18:39.693 --> 00:18:45.173
of personal resentment for me is how many people you know were just like in
00:18:45.173 --> 00:18:48.453
the middle of the day just out there got like what do you do you're in your
00:18:48.453 --> 00:18:53.073
jobs paying you right like i i don't know that that really kind of chafed me a little
00:18:53.073 --> 00:18:53.313
Bit.
00:18:53.573 --> 00:18:58.693
But I wanted to come back, you know, Matt, you're doing a lot of research on innovation.
00:18:58.873 --> 00:19:01.853
Like that, that's clearly kind of a theme in terms of what you're doing.
00:19:02.053 --> 00:19:07.073
And we've been talking a lot about AI and what its impacts are going to be.
00:19:07.533 --> 00:19:14.353
Can AI innovate things on its own without pre-human based direction,
00:19:14.513 --> 00:19:18.133
you know, or how soon will that happen? Or just what are your thoughts in general?
00:19:19.273 --> 00:19:23.853
Yeah, I think it's not there yet where it can do interesting,
00:19:24.113 --> 00:19:27.213
important stuff with no human in the loop, I would say.
00:19:29.413 --> 00:19:33.113
It's really interesting. So this is an area that's moving so fast that I feel
00:19:33.113 --> 00:19:38.073
like it's kind of breaking our traditional research ecosystem where economists
00:19:38.073 --> 00:19:44.173
study something and it takes a year or two to study it, and then they write the paper.
00:19:44.313 --> 00:19:48.193
And so it's like four or five years behind when the events happened that you
00:19:48.193 --> 00:19:49.993
sort of see the paper in a journal.
00:19:51.173 --> 00:19:56.873
And reading about four or five years ago is like before GPT-3, I think, right?
00:19:57.013 --> 00:20:00.033
Like this is back when these AIs could do very little.
00:20:00.213 --> 00:20:05.873
And so that's been a challenge, I think, is like assessing how good this stuff
00:20:05.873 --> 00:20:09.853
is given that just in the last two years, it's become a lot more capable. Yeah.
00:20:10.727 --> 00:20:13.607
One really nice study about this that everyone's been talking about in
00:20:13.607 --> 00:20:16.707
the last couple weeks is about a kind of
00:20:16.707 --> 00:20:19.887
a large material company that develops like complex materials
00:20:19.887 --> 00:20:23.027
and they they did like a nice randomized control trial internally
00:20:23.027 --> 00:20:27.367
in their org where they gave some of their teams access to a ai to help them
00:20:27.367 --> 00:20:32.207
design new materials and others didn't and it seems like it really helped like
00:20:32.207 --> 00:20:38.387
the teams that had access produced like 40 more sort of new materials discovered
00:20:38.387 --> 00:20:40.127
They're filing patents for them.
00:20:40.327 --> 00:20:45.267
They're turning them into products, things like that. But the humans stayed in the loop.
00:20:45.467 --> 00:20:48.447
And one thing that's really neat about the study is they have,
00:20:48.587 --> 00:20:52.547
in this company, people kept really close logs on how they use their time.
00:20:52.767 --> 00:20:56.087
And so you can kind of see how their time shifted.
00:20:56.267 --> 00:21:00.627
And it went from, they basically handed off all of the sort of ideation to the
00:21:00.627 --> 00:21:04.467
machines and took on a lot bigger role in like evaluation.
00:21:04.467 --> 00:21:11.267
So the machines, this is a 2022 era AI, would spit out a bunch of sort of ideas
00:21:11.267 --> 00:21:16.947
for here's how you could organize molecules to have a structure that has these kinds of properties.
00:21:16.947 --> 00:21:21.427
And then they would evaluate what seemed feasible and what didn't because the
00:21:21.427 --> 00:21:24.027
AI doesn't actually have, it's sort of pattern matching.
00:21:24.067 --> 00:21:29.587
It doesn't necessarily have a deep understanding of physical laws and so forth. So that's one example.
00:21:30.127 --> 00:21:33.867
I worry about just like human brain rot from this sort of stuff.
00:21:33.867 --> 00:21:40.227
Like i i catch myself doing the same thing like my best friend is my gpt right
00:21:40.227 --> 00:21:44.667
it's my favorite co-worker as it were right and like i i find myself doing the
00:21:44.667 --> 00:21:48.687
same thing like oh you you come up with ideas and i will vet them and then you
00:21:48.687 --> 00:21:50.707
become like a essentially a middle manager at some point
00:21:50.707 --> 00:21:57.767
No i think this is actually going to be like the next big thing people worry about in this yeah.
00:21:57.767 --> 00:21:59.167
Absolutely like
00:21:59.167 --> 00:22:04.167
Even in this paper that i'm talking about they find that you know some people
00:22:04.167 --> 00:22:08.767
are really good at evaluating the ideas of the AI and other people are not.
00:22:09.407 --> 00:22:13.247
And it's sort of my reading of it is the people who are really good at evaluating
00:22:13.247 --> 00:22:16.947
the ideas are the people who had experience previously doing what the AI did.
00:22:17.727 --> 00:22:22.127
And, you know, they, they kind of like, they used to come up with brainstorm
00:22:22.127 --> 00:22:25.827
ideas and they're just, they have good intuitions for what's a good idea versus
00:22:25.827 --> 00:22:29.647
a bad one, but they probably developed all that intuition through you're doing
00:22:29.647 --> 00:22:31.687
the work that the AI is now doing.
00:22:32.167 --> 00:22:36.887
And you can see the same thing at a lot more like mundane level at school, right?
00:22:37.007 --> 00:22:42.627
Like if a GPT can write your essay, do you really learn how to write,
00:22:42.867 --> 00:22:47.767
structure your arguments, think carefully, all that stuff. And yeah.
00:22:47.927 --> 00:22:51.527
It's like driving like an automatic car. Like I know very few people that can
00:22:51.527 --> 00:22:53.467
drive in manual transmission anymore.
00:22:53.787 --> 00:22:58.567
Like why would you, right? Like everything, it's probably hard to get a manual transmission.
00:22:59.600 --> 00:23:01.060
But it's tricky because like,
00:23:01.200 --> 00:23:06.780
In some sense, I don't know how to drive a manual transmission and it hasn't held me back in life.
00:23:07.140 --> 00:23:09.320
Exactly, exactly. No need.
00:23:09.520 --> 00:23:16.080
But if I'm evaluating the output of a model and having done that work myself,
00:23:16.280 --> 00:23:17.720
it makes me a very good evaluator.
00:23:17.980 --> 00:23:21.780
Then, you know, you worry about people who don't have that experience.
00:23:21.780 --> 00:23:25.240
They don't have any training of going through what it's like to write an essay
00:23:25.240 --> 00:23:27.920
yourself and developing taste maybe or so on.
00:23:28.300 --> 00:23:31.840
And I was talking to some economists, too, about writing theory papers now.
00:23:32.280 --> 00:23:38.180
You can ask the latest models to do a lot of the math, and it's kind of like
00:23:38.180 --> 00:23:39.340
a superpowered calculator.
00:23:39.500 --> 00:23:43.020
And I guess people have always been worried about calculators eroding your skills.
00:23:43.020 --> 00:23:47.160
But it's the kind of thing where like he they would say, I would now prefer
00:23:47.160 --> 00:23:52.660
to work with this than a grad student to to do certain kinds of proofs and things.
00:23:52.960 --> 00:23:56.420
And, you know, where does the grad student then get the sort of mentorship and
00:23:56.420 --> 00:23:59.340
experience if people are not interested in working with them?
00:24:00.060 --> 00:24:03.200
We've seen some other like really interesting research essentially like
00:24:03.200 --> 00:24:05.960
high skilled worker high skilled workers and low skilled
00:24:05.960 --> 00:24:10.280
workers are really effective at using sort of like these augmented assistants
00:24:10.280 --> 00:24:15.020
but it's like those intermediate level people with like sort of like modest
00:24:15.020 --> 00:24:19.940
skills that are perhaps too too much hubris to use you know these additional
00:24:19.940 --> 00:24:25.320
devices they really fail right they don't get an performance boost at all that's
00:24:25.320 --> 00:24:29.840
Interesting i Yeah, I would imagine it's sort of like if you're middle skill,
00:24:30.340 --> 00:24:36.560
then you're like kind of in the sweet spot where, you know, like an AI can do what you do pretty well.
00:24:36.920 --> 00:24:39.640
Maybe you're thinking differently in terms of augmenting me.
00:24:39.700 --> 00:24:41.100
And I'm thinking more of like it replacing.
00:24:41.640 --> 00:24:44.500
I'm thinking if you're middle skill, it's going to boost you up.
00:24:44.600 --> 00:24:47.880
Just like an automatic transmission can make you a better driver. Got it.
00:24:48.380 --> 00:24:50.440
I'm feeling very middle skilled right now.
00:24:53.460 --> 00:24:57.200
Yeah. Sweet spot. I think that different papers have found different things on this.
00:24:59.200 --> 00:25:02.840
I've definitely seen the papers you're talking about where middle or lower skilled
00:25:02.840 --> 00:25:06.200
people benefit a lot from these things, but people who are the best,
00:25:06.240 --> 00:25:10.180
they can't learn anything from how to write better code, for example,
00:25:10.400 --> 00:25:12.860
from an LLM that's really good at coding.
00:25:13.788 --> 00:25:18.128
Then I've also seen papers like this, this one about the material science company,
00:25:18.248 --> 00:25:21.828
where it was sort of the people who are already best at their jobs,
00:25:21.828 --> 00:25:26.148
had their performance boosted by this because of this ability to evaluate the
00:25:26.148 --> 00:25:30.448
AI and work complementary with it was disproportionately favored by them.
00:25:30.448 --> 00:25:35.788
So there's a guy named Balaji Srinivasan who popularized this concept of the
00:25:35.788 --> 00:25:37.528
internet just creates variants.
00:25:38.508 --> 00:25:45.128
And so more people might become idiots, but some people become superhuman intelligent
00:25:45.128 --> 00:25:47.068
because of the variants that's created.
00:25:47.348 --> 00:25:51.628
And I think AI is just a continuation of that. It's going to create even more
00:25:51.628 --> 00:25:56.208
variants where some people, you always heard about the 10x engineer and the 100x engineer.
00:25:56.428 --> 00:25:59.628
Maybe it's going to create the 1000x or the millionx engineer.
00:25:59.628 --> 00:26:05.708
But also it might create it to where, you know, 99.9% of people don't know how to write.
00:26:07.248 --> 00:26:11.748
Like, and that could be, you know, the variance that's created from this.
00:26:11.848 --> 00:26:17.988
And, and I think the bigger question that we're really asking is what is lost in that scenario?
00:26:18.168 --> 00:26:22.248
Like, if you don't know how to drive an auto or a standard, you know,
00:26:22.408 --> 00:26:26.768
shifting car anymore, what is lost? If, you know, if the United States doesn't
00:26:26.768 --> 00:26:30.968
know how to manufacture things anymore, because all the manufacturing is overseas, what is lost?
00:26:31.208 --> 00:26:33.928
You know, if you don't, kids don't know how to write an essay anymore,
00:26:33.928 --> 00:26:37.468
what is actually lost from that? If they can just type in a few prompts and
00:26:37.468 --> 00:26:38.868
then that writes the perfect essay.
00:26:39.188 --> 00:26:41.128
Like what is lost in this scenario?
00:26:41.928 --> 00:26:46.588
Yeah, I think it's, it becomes like, are those things that are lost like manual
00:26:46.588 --> 00:26:49.028
transmission or are they these kind of,
00:26:49.228 --> 00:26:52.228
I think the danger zone is that they're skills that are valuable,
00:26:52.228 --> 00:26:56.248
but we're used to we're used
00:26:56.248 --> 00:26:59.288
to living in a world where being only okay at
00:26:59.288 --> 00:27:02.208
them still produced enough value
00:27:02.208 --> 00:27:05.188
to pay you to do them and then you were that gave
00:27:05.188 --> 00:27:09.688
you an opportunity to develop those skills like you know if you're an artist
00:27:09.688 --> 00:27:14.028
and you're competing with and you're just starting out it'd be really hard to
00:27:14.028 --> 00:27:18.888
compete and sell your your paintings i guess or your your images when people
00:27:18.888 --> 00:27:22.988
People can just use a model to make images that are better than you,
00:27:23.148 --> 00:27:26.648
but that's not going to create a career path that allows you to develop your
00:27:26.648 --> 00:27:29.708
skill and then become eventually better perhaps than these things.
00:27:29.848 --> 00:27:36.048
So I think that's the danger is things that are on this path where we need to have a,
00:27:36.798 --> 00:27:39.618
you need to have like a period of okay is good
00:27:39.618 --> 00:27:42.338
enough to to keep doing it like you can support that
00:27:42.338 --> 00:27:48.518
career and you know we have things like high schools where we train people to
00:27:48.518 --> 00:27:52.838
write things that are not actually good enough to do economic value like people
00:27:52.838 --> 00:27:56.598
are not paying high schoolers for their essays but maybe we'll have to have
00:27:56.598 --> 00:27:59.698
that kind of training system for just way more things and,
00:28:00.238 --> 00:28:03.218
but maybe that doesn't maybe the economics of that don't work out so i don't
00:28:03.218 --> 00:28:04.798
know it's an interesting i think.
00:28:04.798 --> 00:28:07.578
You you alluded to it like on one hand you could
00:28:07.578 --> 00:28:10.358
have like someone produce a an essay in
00:28:10.358 --> 00:28:15.278
chad gbt but then have chad gbt you read it back to you like and summarize that
00:28:15.278 --> 00:28:19.958
essay there's no real gain there but there's other things that go along with
00:28:19.958 --> 00:28:24.058
learning how to construct an argument and like lay out your plan etc it's really
00:28:24.058 --> 00:28:28.918
just like critical thinking and if like that gets eroded we're real real trouble
00:28:28.918 --> 00:28:35.278
Yeah i mean people have this reminds me of like the debate about when we moved
00:28:35.278 --> 00:28:40.438
to automation of manufacturing instead of having these artisanal people who
00:28:40.438 --> 00:28:47.378
had these skill sets to make furniture that had features that you don't see
00:28:47.378 --> 00:28:48.838
in mass produced furniture.
00:28:49.038 --> 00:28:53.458
And the mass produced furniture is cheaper, but it's not as high quality perhaps along some dimensions.
00:28:53.878 --> 00:29:01.078
And in some sense, yeah, there's kind of a substitution of like high quality
00:29:01.078 --> 00:29:06.358
furniture is maybe still available to a very small set of people who pay a high premium for it.
00:29:06.478 --> 00:29:10.318
But most people have access to more furniture, but it's just not as good.
00:29:10.758 --> 00:29:15.078
That's like the whole premise of higher variance. You have a small number that
00:29:15.078 --> 00:29:19.258
have extremely high quality and then a long tail of just a bunch of shitty Ikea
00:29:19.258 --> 00:29:21.398
furniture that no one ever wanted.
00:29:23.138 --> 00:29:24.358
Shout out to Ikea.
00:29:24.518 --> 00:29:25.398
It's not so bad.
00:29:26.498 --> 00:29:31.518
Sorry, Ikea. My apologies. Well, do you want to do a little bit of confusion metrics, Matt?
00:29:32.278 --> 00:29:33.778
It's happy to learn what it is.
00:29:35.016 --> 00:29:36.216
The Confusion Matrix.
00:29:36.796 --> 00:29:40.896
Real quick, we wanted to say thanks again to OneModel, the AI and people analytics
00:29:40.896 --> 00:29:44.656
pioneer for sponsoring the episode. And now back to the Confusion Matrix.
00:29:46.216 --> 00:29:50.736
Both of you have just like killer office set up. So I need to like emulate you guys.
00:29:52.016 --> 00:29:53.536
Well, nothing's stopping you.
00:29:55.176 --> 00:30:00.296
I have terrible internet at the house. There is like a technical tax that needs
00:30:00.296 --> 00:30:01.556
to be played. Hey, brother.
00:30:02.156 --> 00:30:07.196
So when this pod is released, we're going to be coming up on the Christmas holidays.
00:30:07.496 --> 00:30:10.416
Are you a Christmas celebrator, Matt? I am.
00:30:11.076 --> 00:30:14.316
Okay. Okay. Do you love it? You get way into it?
00:30:15.116 --> 00:30:19.936
I think we don't put big lights up in the house, but we obviously have a tree
00:30:19.936 --> 00:30:23.896
and I have three kids and they love Christmas and all that stuff.
00:30:24.416 --> 00:30:27.256
Magical. I bet I was actually magical at christmas time
00:30:27.256 --> 00:30:30.456
it can be yeah i can get nice snow all right
00:30:30.456 --> 00:30:33.276
so i have five questions for you loosely based on
00:30:33.276 --> 00:30:36.316
the big five personality inventory all about
00:30:36.316 --> 00:30:40.376
christmas it's just a chance to talk and get to know you all right all right
00:30:40.376 --> 00:30:46.936
play along if you want to guess the construct of interest as well as well uh
00:30:46.936 --> 00:30:54.856
okay matt uh are you into giving exotic gifts like you know like atypical gifts to other folks?
00:30:55.556 --> 00:31:01.596
I think if you could think of something that's atypical, I'd be excited about it, but it's hard work.
00:31:01.816 --> 00:31:08.196
And I don't think I do enough work to deliver on a lot of people.
00:31:08.356 --> 00:31:10.556
But when you can do it, it's great.
00:31:11.196 --> 00:31:15.316
It is really wild. Essentially, we're just sending Amazon gifts to each other.
00:31:15.556 --> 00:31:19.236
At some point, what do you want from Amazon? I love that
00:31:19.236 --> 00:31:23.396
The question was about openness to experience, but it became about conscientiousness.
00:31:24.616 --> 00:31:29.336
It's a little hard i mean i don't know but i i know scott that you love giving
00:31:29.336 --> 00:31:31.676
exotic gifts so you're definitely high in that
00:31:33.561 --> 00:31:38.301
I used to make a music playlist for people in my family, and I would take a
00:31:38.301 --> 00:31:43.541
lot of pride into finding stuff that I didn't think they would know in that, but that got hard, too.
00:31:44.401 --> 00:31:49.901
Give us some hipster jams, some under-the-ground radar that you love.
00:31:50.681 --> 00:31:54.001
It's been a long time since I've done it, or at least a few years.
00:31:54.361 --> 00:32:01.421
What are some hipster jams that they would like? around there jeez i don't even know like.
00:32:01.421 --> 00:32:03.981
I'm putting you on the spot i apologize here
00:32:03.981 --> 00:32:06.641
I know you can watch me scroll through my phone or something.
00:32:06.641 --> 00:32:09.361
One of my favorite jokes is like if
00:32:09.361 --> 00:32:12.101
you're driving on the highway like there's a car on the on fire on the
00:32:12.101 --> 00:32:16.901
side of the road be like dude they listen to my mixtape finally it's like a
00:32:16.901 --> 00:32:23.021
car got on fire all right all right uh when you're like giving gifts assuming
00:32:23.021 --> 00:32:27.281
that you wrap them or do you make sure it's like super precise and like it looks
00:32:27.281 --> 00:32:31.301
like super good or is it just kind of like a we'll throw it in a bag i'm
00:32:31.301 --> 00:32:36.861
A rapper i won't put i mean i'll use bags occasionally but i think most things i wrap,
00:32:38.321 --> 00:32:42.501
santa wraps presents at our house and and they have to have a certain level
00:32:42.501 --> 00:32:46.561
of quality for that i think you know you can't get away with sloppy work there.
00:32:46.561 --> 00:32:49.841
I mean both of y'all have kids at the house are y'all elf on the shelf people
00:32:49.841 --> 00:32:52.361
like like terrorize the kids for a little while no
00:32:52.361 --> 00:32:55.921
I do not believe in the surveillance state that we're trying to condone.
00:32:57.721 --> 00:33:02.421
Same, same. Also, it's one of those things where it's a lot of work and I didn't
00:33:02.421 --> 00:33:05.641
grow up with it. And so it's sort of unnatural and I don't need to do it, I guess.
00:33:06.221 --> 00:33:10.461
What about you, Cole? Are you a tight gift guy? Or just leave that up to Santa.
00:33:10.681 --> 00:33:13.081
Someone else outsource it. You're going to learn
00:33:13.081 --> 00:33:17.501
How much of a shitty person I am. There's this answer. I literally give my spouse
00:33:17.501 --> 00:33:20.281
money to buy gifts for everybody, including themselves.
00:33:20.501 --> 00:33:25.621
I don't do any of this stuff. Like I'm just gift giving in any context. I don't like birthdays.
00:33:25.881 --> 00:33:29.661
Don't like, I like, I don't mind like the principle behind it.
00:33:29.721 --> 00:33:32.421
Like I love other people receiving things that I give.
00:33:32.621 --> 00:33:36.181
I just, the pro the process of it, I hate. Hmm.
00:33:37.375 --> 00:33:42.335
Oh, are y'all into like big get togethers like Home Alone?
00:33:42.535 --> 00:33:46.335
There's like 12 people in the house. That's how they lost Kevin or like smaller
00:33:46.335 --> 00:33:50.475
sort of like I can't think of a pop culture, small get together,
00:33:50.735 --> 00:33:53.035
Tiny Tim and his little family getting together.
00:33:53.635 --> 00:33:57.055
I like a big get together. That's how I grew up. Go over to grandma's house
00:33:57.055 --> 00:34:01.535
and, you know, there'd be so many people by the end when people started getting
00:34:01.535 --> 00:34:04.015
married and have kids of their own that people were like worried about the structural
00:34:04.015 --> 00:34:05.895
integrity of like the floors.
00:34:06.415 --> 00:34:07.955
Yeah, you have branch-offs.
00:34:07.975 --> 00:34:08.795
Like where the
00:34:08.795 --> 00:34:12.795
One big thing became smaller things just out of necessity. I totally agree,
00:34:12.875 --> 00:34:14.135
Matt. I love that kind of stuff.
00:34:14.715 --> 00:34:18.495
I wish I grew up in this sort of situation. I have a very small family,
00:34:18.495 --> 00:34:24.975
and it's so foreign to me to see 45 people get together on the Hallmark Channel.
00:34:25.415 --> 00:34:31.015
I mean, 45 people, yeah, that was me getting together. Christmas Eve every year.
00:34:31.535 --> 00:34:33.515
That's a lot of gifts. That's a lot of gifts.
00:34:33.515 --> 00:34:37.815
So we did a, we did like a kind of a white elephant gift thing where everybody
00:34:37.815 --> 00:34:42.495
brings one gift for the pool and then there's a competition over them.
00:34:42.915 --> 00:34:45.475
You're like, here's my, here's my mixtape. Exactly.
00:34:46.315 --> 00:34:51.175
And if it's like every Hallmark movie, the, the, you know, the lonely single
00:34:51.175 --> 00:34:53.355
girl comes back for Christmas with her family.
00:34:53.575 --> 00:34:56.915
And then the guy who wasn't in during high school,
00:34:57.155 --> 00:35:02.295
who's recently widowed and now has pursued an interest in her for the first
00:35:02.295 --> 00:35:06.695
time and it's about her new candle making business and how he really has found
00:35:06.695 --> 00:35:12.075
the candle making to be enlightening they fall in love the end there you go I ruined it for you
00:35:12.855 --> 00:35:18.395
Shout out to Holiday in Handcuffs 2009 Melissa Joan Hart and AC Slater from
00:35:18.395 --> 00:35:24.715
Saved by the Bell that's some solid stuff baby Holiday in Handcuffs it's a must see
00:35:26.695 --> 00:35:29.775
we probably already covered this Like, do you worry about the perfect gift?
00:35:30.235 --> 00:35:31.875
Trying to get like, do you wrap yourself up in knots?
00:35:33.396 --> 00:35:37.476
I mean, I think that there's two things.
00:35:37.616 --> 00:35:42.356
There's my spouse and where we both have enough, we're doing well enough that
00:35:42.356 --> 00:35:47.216
if you really want something at any point in the year, you probably can get it.
00:35:47.316 --> 00:35:51.896
And if you don't get it because it's expensive and then it's hard to make the
00:35:51.896 --> 00:35:55.656
unilateral Christmas decision, like we're spending our money on this now for you.
00:35:56.196 --> 00:35:57.576
And then there's the kids where
00:35:57.576 --> 00:36:01.456
they are in the opposite situation where they're very much on the mercy.
00:36:01.456 --> 00:36:05.656
They can't afford to get the things they really want and so for them we want
00:36:05.656 --> 00:36:11.136
to I think we do a pretty good job but it's a little bit of like they give us a list of things,
00:36:12.556 --> 00:36:17.576
there's not this careful figuring out of what would exactly match their personality or anything.
00:36:18.116 --> 00:36:21.396
There's like a technology play here back in the day
00:36:21.936 --> 00:36:25.456
things had to be handcrafted it would take forever to be expensive and you took
00:36:25.456 --> 00:36:29.316
care of your things, now everything's just like cheap plastic from China right
00:36:29.316 --> 00:36:33.696
like you just buy whenever it doesn't really matter just
00:36:33.696 --> 00:36:35.556
Fills up landfills eventually yeah
00:36:35.556 --> 00:36:40.056
Yeah dude you go like the like aisles in target and you'll be like everything
00:36:40.056 --> 00:36:43.736
in here is just going straight to landfill like it's like stuffed animals and
00:36:43.736 --> 00:36:48.756
you're like there's there's no home for this i have no idea what happens to this so depressing we
00:36:48.756 --> 00:36:53.376
Have a yeah like we we take you know kids to fast food places and they get the
00:36:53.376 --> 00:36:56.856
happy meal stuff and then we i think i just put out a garbage bag full of like
00:36:56.856 --> 00:37:01.476
happy meal toys that have accumulated over like the last five years if they're listening those.
00:37:01.476 --> 00:37:04.876
People who hand out flyers on the side of the road it's like oh you just give
00:37:04.876 --> 00:37:07.396
this to me to put into the trash can thank you
00:37:07.396 --> 00:37:12.896
Let's put that right in the gutter last one like we're coming up on it holidays
00:37:12.896 --> 00:37:18.676
are stressful times do you stress about this or do you just like whatever we'll let it happen
00:37:18.676 --> 00:37:23.796
I don't think i'm too stressed about it it's a little stressful you know there's
00:37:23.796 --> 00:37:28.756
some crunch times where you have to figure out how to get things but like we
00:37:28.756 --> 00:37:34.876
usually take time off around the holidays and yeah it's it's i'd say net net
00:37:34.876 --> 00:37:37.936
relaxing although there's like some parts that are stressful.
00:37:39.338 --> 00:37:42.098
There should be a little tension right like you can't just like leave your christmas
00:37:42.098 --> 00:37:46.798
lights on the house all year that's right right and just be like we're rolling into it yeah
00:37:46.798 --> 00:37:49.438
What about you scott how do you approach the holidays
00:37:49.438 --> 00:37:57.978
Holidays it's a magical time give me give me your amazon wishlist and we'll take care of it so
00:37:57.978 --> 00:37:59.478
You get them to do all the work for you
00:37:59.478 --> 00:38:02.718
We'll throw money at it well if
00:38:02.718 --> 00:38:08.398
someone doesn't have like a discernible like interest eventually like one will
00:38:08.398 --> 00:38:12.998
be assigned to them so it's like you don't drink you don't you know i don't
00:38:12.998 --> 00:38:17.938
know golf you get pigeonholed all of a sudden you're collecting uh porcelain
00:38:17.938 --> 00:38:21.618
cows that's what i'm deciding for you right so yeah you're
00:38:21.618 --> 00:38:24.198
Gonna get one for every holiday for the rest of your life
00:38:24.198 --> 00:38:26.038
Yeah if
00:38:26.038 --> 00:38:28.438
You don't have a collection yet you will by the end of this
00:38:28.438 --> 00:38:30.998
Absolutely all
00:38:30.998 --> 00:38:35.438
Right let's uh let's do some nerdery. The nerdery.
00:38:36.818 --> 00:38:38.738
I think the one about collaboration,
00:38:38.998 --> 00:38:42.258
Scott, might be a good one per the conversation we're having from earlier.
00:38:42.438 --> 00:38:43.778
Do you want to go there first?
00:38:44.458 --> 00:38:48.658
Yeah, this is a really cool study. And I'd love to get Matt's thoughts on this sort of thing.
00:38:49.278 --> 00:38:52.718
Collaboration, physical proximity, and serendipitous encounters,
00:38:53.058 --> 00:38:55.478
avoiding collaboration in the collaborative building.
00:38:55.678 --> 00:38:59.838
So a lot of companies have these sort of like open office sort of buildings
00:38:59.838 --> 00:39:04.198
or you'd expect that the layouts would encourage serendipitous connections,
00:39:04.478 --> 00:39:08.918
you know, people to meet one another, but they rarely like really show the benefits.
00:39:09.198 --> 00:39:13.638
So the author's really cool study. So they went to a collaborative building,
00:39:13.818 --> 00:39:14.518
Some sort of like.
00:39:14.518 --> 00:39:17.698
Specifically designed to encourage these sort of connections.
00:39:17.898 --> 00:39:20.198
They sat there kind of like Jane Goodall.
00:39:20.438 --> 00:39:24.058
They, you know, monitored everyone in the office. They interviewed several folks
00:39:24.058 --> 00:39:29.658
and just kind of tried to discern like how they are avoiding each other and essentially
00:39:30.358 --> 00:39:34.858
developing serendipitous connections was exceedingly rare due to four tactics
00:39:34.858 --> 00:39:36.578
that folks would employ.
00:39:36.758 --> 00:39:41.618
One is focusing on existing connections, so only talking to those people that you already know.
00:39:42.218 --> 00:39:46.218
Reinforcing group boundaries, so failing to invite other people to sit at your
00:39:46.218 --> 00:39:48.398
lunch table while you're eating, this sort of thing.
00:39:48.958 --> 00:39:51.638
Enacting legacy policies, so essentially claiming like, well,
00:39:51.738 --> 00:39:55.018
that group just does things different than us, or they're on a different financial system.
00:39:55.898 --> 00:39:59.978
And minimizing social interactions, that's your typical, just put on your headphones sort of stuff.
00:40:00.278 --> 00:40:07.038
But for those rare few scientists who did overcome this, they simply just said
00:40:07.038 --> 00:40:10.578
hello to folks, right? Just were open to meeting people.
00:40:11.773 --> 00:40:18.813
Yeah, so this reminds me of a study about a startup accelerator where they had
00:40:18.813 --> 00:40:21.093
a bunch of different startup teams.
00:40:21.273 --> 00:40:25.473
I think this was some kind of short-term accelerator thing.
00:40:25.553 --> 00:40:29.393
So it wasn't just their permanent home. It was like they were together for a period of time.
00:40:29.793 --> 00:40:31.673
And this was based in India.
00:40:32.793 --> 00:40:38.033
And they kind of wanted to see who do people interact with, who do they meet,
00:40:38.353 --> 00:40:42.253
who are their connections. And the people who came in already knowing people,
00:40:42.253 --> 00:40:47.013
you know, came out essentially knowing fewer people than the people who didn't know anybody going in.
00:40:47.133 --> 00:40:51.613
And it was because they just disproportionately relied on their pre-existing
00:40:51.613 --> 00:40:56.593
ties and didn't sort of reach out or they weren't forced to reach out and meet new people.
00:40:56.793 --> 00:41:00.233
So that kind of, that, you know, that, that matches what I've heard.
00:41:00.933 --> 00:41:04.133
I think you'd post an article, I'll attribute to you anyway,
00:41:04.373 --> 00:41:08.313
of folks going to like say conferences or like these sort of like training seminars.
00:41:08.553 --> 00:41:11.393
If they already knew people going in, they met fewer people,
00:41:11.473 --> 00:41:15.953
but people that don't know anybody were much more open and developed new connections overall.
00:41:16.593 --> 00:41:20.293
Yeah, I don't, I'm not sure, but like this reminds me of a thing I've often
00:41:20.293 --> 00:41:22.893
wondered about is like if you were going to design a conference and you just
00:41:22.893 --> 00:41:27.953
want to maximize how many people you can meet. I think actually you might have
00:41:27.953 --> 00:41:30.193
more luck with a smaller conference than a bigger one.
00:41:30.373 --> 00:41:33.393
And because if you have a big conference, there's more likelihood that people
00:41:33.393 --> 00:41:34.433
you already know are going to be there.
00:41:34.933 --> 00:41:39.013
And then you can kind of just cluster together. You can get like 10 people to
00:41:39.013 --> 00:41:41.013
go out that you already know for drinks.
00:41:41.253 --> 00:41:44.053
And if you're in a small conference, you might know a few people,
00:41:44.173 --> 00:41:48.873
but you don't have enough to just fill your weekend getting to chatting with
00:41:48.873 --> 00:41:49.613
the people you already know.
00:41:50.093 --> 00:41:53.353
Well, I think it goes back to all that paradox of choice. if
00:41:53.353 --> 00:41:56.253
you have like three choices of people to talk to you're more likely
00:41:56.253 --> 00:41:59.313
to talk to somebody than if you have a thousand people to
00:41:59.313 --> 00:42:02.313
choose from and so you're just going to be paralyzed by it
00:42:02.313 --> 00:42:06.233
but the thing i liked about the article scott is
00:42:06.233 --> 00:42:12.053
i kept thinking this as i was looking through it you can't make me right like
00:42:12.053 --> 00:42:16.153
like the organization is saying we're going to create this environment and we're
00:42:16.153 --> 00:42:20.113
going to force you to sit in it and it's going to make you collaborate it's
00:42:20.113 --> 00:42:23.333
going to make you serendipitously run into somebody else.
00:42:23.473 --> 00:42:28.553
And the employees are saying, we've got these four strategies that say, you can't make me.
00:42:29.033 --> 00:42:32.793
So another thing that I've always thought interesting about this whole line of work is.
00:42:33.793 --> 00:42:39.553
It's like a pretty consistent finding that once you're even a little bit far away from somebody,
00:42:39.813 --> 00:42:43.093
So I think it's like 25 meters or something, it's like they might as well like
00:42:43.093 --> 00:42:46.913
be remote working because like your probability of talking to people who are
00:42:46.913 --> 00:42:51.213
like quite close to you is significantly elevated if you're in the same place.
00:42:51.433 --> 00:42:54.753
But you can't really actually scale that as much as you think.
00:42:55.033 --> 00:42:56.693
Like you think, you know, you can.
00:42:57.373 --> 00:43:00.993
That's really wild. Like, yeah, I mean, if they're not in that immediately area,
00:43:00.993 --> 00:43:06.073
I often wonder about like, Like I purpose, I mad that you don't know me that
00:43:06.073 --> 00:43:08.673
well, but I will talk to anybody at any time about anything.
00:43:09.133 --> 00:43:12.993
And I purposely sit myself next to the bathroom here at the office.
00:43:13.173 --> 00:43:15.413
And people hate me because I'm like, hey, what's up?
00:43:16.093 --> 00:43:17.733
Whether they go pee or whatever.
00:43:18.133 --> 00:43:19.333
They're like, oh, oh shit.
00:43:19.353 --> 00:43:26.253
He's here again. You know, that sort of thing, you know. But while the proximity surely matters,
00:43:26.373 --> 00:43:31.393
like you've also posted like really interesting, like naturalistic studies of say asbestos,
00:43:31.393 --> 00:43:36.153
like throws two different science departments together and they i can't remember
00:43:36.153 --> 00:43:40.273
the exact stats but it's like two to five x public co-publication rate than
00:43:40.273 --> 00:43:42.453
you would expect at the rest of the campus yeah that's
00:43:42.453 --> 00:43:45.313
A really interesting study where they were like removing asbestos so
00:43:45.313 --> 00:43:48.433
they moved all the labs around it's in france and as
00:43:48.433 --> 00:43:51.313
you said like if you get moved together you're a lot more likely to collaborate
00:43:51.313 --> 00:43:54.313
but the thing that's interesting is when you move them apart they don't actually
00:43:54.313 --> 00:43:57.093
get less likely to collaborate and that gets back to that idea
00:43:57.093 --> 00:44:00.653
that like this is great for like meeting people uh but
00:44:00.653 --> 00:44:02.873
then like once you're apart and i think it's important to remember these guys
00:44:02.873 --> 00:44:07.973
are still all in the same like university campus so it's not like if if i meet
00:44:07.973 --> 00:44:10.393
somebody and we become friends and then we decide we're going to collaborate
00:44:10.393 --> 00:44:14.713
on a project but years later we're now on opposite ends of the campus it's it's
00:44:14.713 --> 00:44:17.493
not that hard for us to like meet up and do all that.
00:44:17.493 --> 00:44:21.913
Stuff well what can i pick on this for a second because this gets to kind of
00:44:21.913 --> 00:44:27.893
the axe i have to grind with companies versus the example you used is the example you used was like
00:44:28.093 --> 00:44:31.093
again the asbestos could be another argument for
00:44:31.093 --> 00:44:33.933
this is just a serendipitous happening where the teams
00:44:33.933 --> 00:44:36.913
got moved together it was completely random it wasn't like
00:44:36.913 --> 00:44:40.373
strategically created the reason why organizations
00:44:40.373 --> 00:44:43.433
are doing this usually isn't to increase collaboration
00:44:43.433 --> 00:44:46.473
that's the guys they're putting on it it's usually to
00:44:46.473 --> 00:44:49.353
re to reduce the real estate footprint and save money
00:44:49.353 --> 00:44:52.133
right and employees are wise to
00:44:52.133 --> 00:44:54.973
that right and so i i think
00:44:54.973 --> 00:44:58.313
you know again that back to my point about you can't make me is
00:44:58.313 --> 00:45:01.013
organizations are saying hey you know how you used to have a little bit of
00:45:01.013 --> 00:45:05.273
private space that was like yours well you can't have that anymore now your
00:45:05.273 --> 00:45:08.333
private space is everyone's public space and we're just going to cram as many
00:45:08.333 --> 00:45:12.973
of you in as possible and and so people are saying well i'm rightfully going
00:45:12.973 --> 00:45:17.233
to put up my kind of my mental and my social walls because you took away my
00:45:17.233 --> 00:45:20.833
actual walls right and so there's a part of this where it's like
00:45:21.532 --> 00:45:27.552
If you really want to create collaboration, you need to incentivize collaboration
00:45:27.552 --> 00:45:33.012
or incentivize the serendipitous connections, but don't force it because you
00:45:33.012 --> 00:45:34.492
actually have an ulterior motive.
00:45:34.752 --> 00:45:38.392
That is where I come down, like the university example versus the organizational
00:45:38.392 --> 00:45:40.812
example, I think are qualitatively different in that regard.
00:45:41.172 --> 00:45:44.652
Yeah, I think that it's interesting how often, I mean, I get the sense too,
00:45:44.812 --> 00:45:49.292
that a lot of return to office rhetoric is kind of like, you know,
00:45:49.512 --> 00:45:53.252
motivated reasoning. Like that's not really why we want you to return to office.
00:45:53.472 --> 00:45:56.792
It's because we don't often, it's like, maybe we don't have systems in place
00:45:56.792 --> 00:45:59.572
to monitor that you're really doing work and we just don't trust you and we
00:45:59.572 --> 00:46:00.712
don't have a high trust culture.
00:46:00.812 --> 00:46:03.172
So we want everyone to be sort of in a place.
00:46:03.452 --> 00:46:09.812
And, and the serendipity line is like a useful, you know, fig leaf about why
00:46:09.812 --> 00:46:11.652
we're asking everyone to come back to the office.
00:46:11.932 --> 00:46:14.712
Like, you know, there was a, there's a study that people were tweeting about
00:46:14.712 --> 00:46:16.092
this morning about that.
00:46:16.172 --> 00:46:19.132
I think it's interesting because I haven't read the study. So I'm going off
00:46:19.132 --> 00:46:23.212
of a hot take tweet, I guess, but it was like monitoring software engineers
00:46:23.212 --> 00:46:24.392
and their productivity.
00:46:24.772 --> 00:46:29.692
And it looks at sort of their GitHub commits as a way to measure how productive they are.
00:46:29.952 --> 00:46:36.772
And the people who are remotely working are both more likely to be like 5X engineers
00:46:36.772 --> 00:46:41.212
and more likely to be what they called like ghost engineers who basically do not work.
00:46:42.039 --> 00:46:44.479
And so, again, that's, I guess, that high variance thing.
00:46:44.639 --> 00:46:48.779
But, you know, one response to that is sort of like, well, the org just needs
00:46:48.779 --> 00:46:51.859
to do a better job of monitoring people's GitHub commits.
00:46:51.939 --> 00:46:56.159
Like if these researchers can do something like this, why is it,
00:46:56.639 --> 00:46:59.179
you know, why is this an issue that the firm has never thought about?
00:46:59.359 --> 00:47:04.219
No, I love this point, Matt. And the thing is, is like organizations try to
00:47:04.219 --> 00:47:07.779
create a singular policy for one camp or the other camp.
00:47:07.979 --> 00:47:12.659
They either say, we're all 5X engineers now, clearly. And so here's our policy
00:47:12.659 --> 00:47:16.639
or we're all deadbeats who don't work at all. So here's our policy.
00:47:16.819 --> 00:47:21.319
And it's like, how about you, you know, again, use some organizational science
00:47:21.319 --> 00:47:25.859
to show that we have a different policy per the type of performance that we're seeing.
00:47:26.159 --> 00:47:33.059
Yeah, yeah. The other thing I think is interesting is like ways to try to create
00:47:33.059 --> 00:47:38.199
serendipity or new connections among kind of remote employees.
00:47:38.859 --> 00:47:41.739
So I think we connected because I don't
00:47:41.739 --> 00:47:44.639
know if this was if you went down this route but I
00:47:44.639 --> 00:47:47.539
have like on my website a like button to like book virtual coffee
00:47:47.539 --> 00:47:50.539
with me yeah and like that's an intentional choice
00:47:50.539 --> 00:47:54.359
because I'm based in Iowa and I value this sort of serendipity stuff and I have
00:47:54.359 --> 00:47:59.359
this kind of public internet profile through writing the substack so it is likely
00:47:59.359 --> 00:48:03.739
that people might stumble upon this virtual coffee button but anyway that's
00:48:03.739 --> 00:48:07.859
like you can do other things too Our org has like this, you know,
00:48:08.139 --> 00:48:12.059
you can sign up to be randomly matched with people to have like remote virtual
00:48:12.059 --> 00:48:14.899
donut coffee, you know, once every two weeks or something.
00:48:15.279 --> 00:48:18.999
Well, Matt, you just, you don't know how many people are going to virtually
00:48:18.999 --> 00:48:21.879
book coffee with you now because this podcast.
00:48:21.899 --> 00:48:23.819
So you may regret that decision.
00:48:23.999 --> 00:48:24.799
Remove that button.
00:48:25.599 --> 00:48:28.799
I do have it set up so that like you can only book one per day.
00:48:28.799 --> 00:48:30.339
So I don't, I can't get over wrong.
00:48:31.299 --> 00:48:34.779
I mean, like I can only do one per day. It's so.
00:48:35.657 --> 00:48:39.717
There are other, I mean, we can move on, but like, I think that moving forward,
00:48:39.917 --> 00:48:44.817
like FaceTime with the boss is never going to go away as well as like just the home office bias.
00:48:45.197 --> 00:48:46.937
I don't think we're ever going to get away from that.
00:48:47.817 --> 00:48:55.257
Well, I got an article here since we're moving on to about exit surveys versus engagement surveys.
00:48:55.457 --> 00:48:59.537
And so there's been kind of this question as old as time is,
00:48:59.677 --> 00:49:04.697
you know, are people honestly responding to the surveys that organizations provide?
00:49:04.697 --> 00:49:07.857
And so some people have argued, well, exit surveys,
00:49:08.037 --> 00:49:13.057
because there's no consequences anymore, people are being more accurate in their
00:49:13.057 --> 00:49:17.597
responses, especially about politically engaged topics like,
00:49:17.757 --> 00:49:21.337
do I like my manager? Do I not like my manager?
00:49:21.777 --> 00:49:25.517
And, you know, I would be more afraid to share that on an engagement survey
00:49:25.517 --> 00:49:28.817
because it might come back to haunt me versus an exit survey.
00:49:28.817 --> 00:49:30.977
Well, they can't do anything to me more because I don't work here.
00:49:31.597 --> 00:49:35.577
And friend of the podcast, Ludic Stelic. And by the way, Ludic, I need to reach out.
00:49:35.657 --> 00:49:37.997
We haven't connected in a while. So I'll be reaching out to you.
00:49:38.077 --> 00:49:42.017
Maybe if you put a book of coffee time on your LinkedIn, I'll book some coffee.
00:49:42.737 --> 00:49:47.497
But he did some research to show that actually that's really not true.
00:49:48.177 --> 00:49:53.417
And so what they did is they time lagged the number of months between when somebody
00:49:53.417 --> 00:49:55.337
took an engagement survey and quitting.
00:49:55.657 --> 00:49:59.717
And when they took the exit survey, presumably the exit survey came right after
00:49:59.717 --> 00:50:01.097
they quit or right before they quit.
00:50:01.437 --> 00:50:05.737
And what they found is, is essentially the number of months the way an engagement
00:50:05.737 --> 00:50:10.637
survey was, was they were less likely to be an accurate prediction of them leaving
00:50:10.637 --> 00:50:12.677
in terms of the relationship with their manager.
00:50:12.897 --> 00:50:17.177
But if the exit or the engagement survey was right around the time the person
00:50:17.177 --> 00:50:21.837
left, their responses were almost identical to the exit survey responses.
00:50:22.057 --> 00:50:25.497
So I thought this was an interesting finding.
00:50:25.857 --> 00:50:30.557
How do you guys react to this? And are there other kind of implications you
00:50:30.557 --> 00:50:33.897
see in terms of other research along these lines?
00:50:35.017 --> 00:50:38.897
Because I've been thinking about this question of how prevalent is scientific
00:50:38.897 --> 00:50:44.657
fraud and wondering if people are under-reporting when they're doing these anonymous surveys.
00:50:44.737 --> 00:50:48.677
And now I'm wondering if you could learn something by surveying retirees,
00:50:48.777 --> 00:50:54.457
basically, or people who are maybe, you know, less, less worried about being caught out.
00:50:55.277 --> 00:51:00.957
Yeah. Well, I love, what? Can we just like completely pivot on this for a second?
00:51:01.057 --> 00:51:04.397
Cause I don't feel like we got to squeeze enough juice out of the lemon earlier
00:51:04.397 --> 00:51:05.897
on the scientific fraud point.
00:51:06.477 --> 00:51:09.497
Do you think, like, why is it so prevalent?
00:51:09.937 --> 00:51:13.477
And again, my, my perspective has always been, it's just the incentives.
00:51:13.717 --> 00:51:17.657
There's a limited number of top tier journals. There's so many academics out
00:51:17.657 --> 00:51:20.077
there. It's so competitive to publish in them.
00:51:20.617 --> 00:51:23.737
And frankly, a lot of science has already been conducted over the last hundred
00:51:23.737 --> 00:51:27.477
years. So the chances that you're going to find something new are quite low.
00:51:27.737 --> 00:51:33.217
Is it all just incentives as to why this is happening or are academics just bad people?
00:51:34.811 --> 00:51:38.311
I think the academics are people and there's good and bad people.
00:51:39.051 --> 00:51:42.791
I think the incentives, like I'm an economist, we really focus on the incentives.
00:51:43.071 --> 00:51:47.951
And I think they're certainly like you say, and there are these sort of interesting
00:51:47.951 --> 00:51:52.231
traps fields can get into where the Atlantic article alludes to one of these
00:51:52.231 --> 00:51:56.191
were sort of like you're in a dynamic in, I think it was business psychology there.
00:51:56.371 --> 00:52:00.991
They kind of argued where everybody's doing fraud. And so if you don't do fraud,
00:52:01.351 --> 00:52:02.251
you're at a real disadvantage.
00:52:02.771 --> 00:52:05.991
I mean, fraud is maybe too strong a word for what a lot of people do.
00:52:06.091 --> 00:52:08.951
Like it's not often necessarily like literally making up data,
00:52:09.111 --> 00:52:12.531
but often it's cutting corners on your research.
00:52:12.751 --> 00:52:15.631
And, you know, you didn't find the relationship.
00:52:15.931 --> 00:52:16.831
Harking, harking.
00:52:16.951 --> 00:52:20.571
Exactly. You know, so all the P hacking type stuff. I think that's probably
00:52:20.571 --> 00:52:23.411
pretty common, like P hacking type stuff.
00:52:24.311 --> 00:52:29.311
And another kind of interesting trap you can get into is if there's a field
00:52:29.311 --> 00:52:34.991
where it's just really hard to do compelling work.
00:52:35.231 --> 00:52:39.951
You could argue a lot of economics has this problem where measuring stuff is harder and so forth.
00:52:40.091 --> 00:52:44.731
And then if measuring stuff is really hard because you don't have a good measure
00:52:44.731 --> 00:52:47.871
of innovation, for example, you can use patents to measure innovation,
00:52:47.871 --> 00:52:50.611
but they are really imperfect and so forth.
00:52:50.731 --> 00:52:54.331
Then that means a lot of the time your research doesn't find anything because
00:52:54.331 --> 00:53:01.151
it's just the data is too noisy and then that means it's very easy to,
00:53:01.551 --> 00:53:06.651
it can be hard to sort of get a finding the credible way. But also...
00:53:07.840 --> 00:53:12.800
If you don't get the result you expect, so you kind of expect there to be a
00:53:12.800 --> 00:53:16.380
positive relationship, but you don't see it, you know, in some sense,
00:53:16.460 --> 00:53:17.460
you should take that as like a signal.
00:53:17.580 --> 00:53:20.700
Maybe there's nothing there, but maybe you say, well, it's a lot of noise in
00:53:20.700 --> 00:53:22.980
the data. So like it's I'll just try again.
00:53:23.340 --> 00:53:27.400
And and, you know, you don't you don't sort of publish it as a null result that
00:53:27.400 --> 00:53:29.000
people say, oh, maybe there's actually nothing here.
00:53:29.100 --> 00:53:32.640
You just kind of bury it because you assume it's just file file drawer.
00:53:32.640 --> 00:53:39.880
Yeah but yeah you wonder if the file drawer is is a bigger problem in fields where it's hard to,
00:53:40.380 --> 00:53:44.240
like the data is really noisy like if you found a null result in physics for
00:53:44.240 --> 00:53:47.540
example that everyone expects that you know there's going to be this we have
00:53:47.540 --> 00:53:49.460
a really good theoretical physics explanation,
00:53:49.980 --> 00:53:53.180
and we looked and we actually didn't see something like we couldn't detect the
00:53:53.180 --> 00:53:56.260
higgs boson it's not where we thought it would be you'd publish that and that'd
00:53:56.260 --> 00:54:00.280
be exciting but in other fields you know you didn't find what you thought you'd
00:54:00.280 --> 00:54:03.300
see and you're like well we'll just put that in the file drawer because it's
00:54:03.300 --> 00:54:06.400
probably the data was too noisy to detect it and we should just try.
00:54:06.400 --> 00:54:09.900
Again engineers like claiming that we shouldn't probably you know get anything
00:54:09.900 --> 00:54:14.240
out that isn't like 0.8 0.9 effect size like we're trying to predict humans
00:54:14.240 --> 00:54:15.940
like you're trying to predict a steel beam
00:54:15.940 --> 00:54:16.440
Yeah this.
00:54:16.440 --> 00:54:17.240
Is different man
00:54:17.240 --> 00:54:22.200
Yeah yeah exactly exactly but then that lets that makes it harder to sort of
00:54:22.200 --> 00:54:25.780
build up the evidence base in the fields that sort of need it the most which
00:54:25.780 --> 00:54:27.580
is which is the trap especially.
00:54:27.580 --> 00:54:31.740
When you look over and your co-worker is like what's up like raining money and
00:54:31.740 --> 00:54:33.400
all the publications everywhere.
00:54:33.840 --> 00:54:37.540
Yeah, and the other thing is if it's a really hard field to do work in,
00:54:37.600 --> 00:54:42.600
then that means you might need giant samples and then that becomes expensive and.
00:54:43.543 --> 00:54:45.443
Yeah it's it's it's
00:54:45.443 --> 00:54:48.783
Just it gets harder to learn and all sorts of so that like you get stuck in
00:54:48.783 --> 00:54:52.023
that in that area without knowing much or you make really slow progress anyway.
00:54:52.023 --> 00:54:57.503
Like on this study specifically i i don't think there's like a big mystery here
00:54:57.503 --> 00:55:01.243
it's like proximal outcomes are just like easier to predict like if you hate
00:55:01.243 --> 00:55:05.983
your boss if you snap you're like okay i'm out of here right i think it was
00:55:05.983 --> 00:55:09.383
camping and moretz at profiles and quitting so there's like four different
00:55:10.223 --> 00:55:13.043
profiles and people that quit some people like you can't save them at all
00:55:13.043 --> 00:55:16.083
like they're gonna go off to school something like that some people
00:55:16.083 --> 00:55:20.443
have like a slow burn some people like there's like a stab judgment like you
00:55:20.443 --> 00:55:23.463
know their boss yells at them when they're out of there this sort of stuff really
00:55:23.463 --> 00:55:28.603
interesting paper but yeah i i think that like uh quitting is kind of like an
00:55:28.603 --> 00:55:31.943
it's kind of like breaking up a relationship you're like i'm done okay like
00:55:31.943 --> 00:55:34.663
you reach a point you're like i'm out of here i'm looking for a new job
00:55:34.663 --> 00:55:41.643
Yeah when i think adam grant has famously talked about the wisdom in finding
00:55:41.643 --> 00:55:45.363
the thing you thought you should find like hey proximal effects should affect
00:55:45.363 --> 00:55:48.923
things proximally right it would be weird if they didn't
00:55:48.923 --> 00:55:50.223
Yeah well
00:55:50.223 --> 00:55:54.643
Scott do you want to move on to the last one about transmission of mental disorders do you
00:55:54.643 --> 00:55:56.383
Have time matt a couple minutes yeah
00:55:56.383 --> 00:55:57.203
I can do a couple minutes.
00:55:57.203 --> 00:56:01.943
Sweet transmission of mental disorders in adolescent peer networks so uh previous
00:56:01.943 --> 00:56:06.643
research had indicated that mental disorders may be transmitted from one individual to another.
00:56:07.203 --> 00:56:11.683
Famously, the Christakis and Fowler Farmington studies, where they found people
00:56:11.683 --> 00:56:15.803
being obese and happiness, etc., will flow through people.
00:56:15.943 --> 00:56:18.823
But it's never been carried out in a large, large sample.
00:56:19.023 --> 00:56:26.723
So what they did is they took all Finnish citizens, born between 1985 and 1997, for whom they had data.
00:56:26.903 --> 00:56:29.943
So 700,000 people, cohort members, whatever.
00:56:30.743 --> 00:56:34.323
And they assessed them in, I believe it was 9th grade.
00:56:34.583 --> 00:56:39.243
And the results show that having one or more diagnosed classmates in your class
00:56:39.243 --> 00:56:44.983
increased your risk by 5% for a later diagnosis. The most common or the most
00:56:44.983 --> 00:56:47.683
prevalent were mood, anxiety, and eating disorders.
00:56:49.423 --> 00:56:54.403
The risk was most pronounced in the first year follow-up. So you have like this like proximal effect.
00:56:54.623 --> 00:56:58.443
And there was also an effect of a dose response relationship.
00:56:58.443 --> 00:57:02.003
So the higher the dose, the higher number of people in your class that were
00:57:02.003 --> 00:57:06.063
diagnosed with a mental disorder, the more likely the students were to also
00:57:06.063 --> 00:57:08.483
develop a mental disorder of the same type.
00:57:09.398 --> 00:57:11.378
Interesting stuff. Interesting stuff. So like.
00:57:11.658 --> 00:57:17.098
Well, it says they controlled for area level co-founders or confounders. Sorry.
00:57:17.618 --> 00:57:19.698
They controlled for a lot in the study. Yeah.
00:57:19.838 --> 00:57:22.818
I wonder like it was like lead in the water controlled for.
00:57:22.898 --> 00:57:26.898
Because I imagine if you're drinking the water and it has lead in it and you
00:57:26.898 --> 00:57:30.498
get a mental disorder, probably everyone else in the class is more likely to
00:57:30.498 --> 00:57:33.838
develop a mental disorder later on in life because the lead in the water there too.
00:57:33.838 --> 00:57:40.238
They controlled, it was a Cox hazard model, but it was also a multi-level regression
00:57:40.238 --> 00:57:41.978
sort of formula that they worked out there.
00:57:42.258 --> 00:57:46.958
And they controlled for parents' incomes and education and all sorts of factors.
00:57:46.958 --> 00:57:51.358
Yeah, but that problem you're alluding to, like the related literature I know
00:57:51.358 --> 00:57:56.518
that has the exact same problems is like social contagion of interest in entrepreneurship,
00:57:56.518 --> 00:57:57.958
which I've written about before.
00:57:58.238 --> 00:58:03.258
And so do people like they find similar things, like if you have more co-workers
00:58:03.258 --> 00:58:06.138
who are ex-entrepreneurs, you're more likely to found a business yourself.
00:58:06.978 --> 00:58:13.178
And the big challenge in that research is maybe people who are interested in
00:58:13.178 --> 00:58:15.538
entrepreneurship work at the same companies.
00:58:15.538 --> 00:58:18.418
Maybe they all just happen to work in an industry
00:58:18.418 --> 00:58:23.098
that is easier to start a business in and that's why they all know each other
00:58:23.098 --> 00:58:28.978
but the the papers i think do a lot of interesting stuff there's like actually
00:58:28.978 --> 00:58:31.598
a lot of different lines of evidence that i think shows that this like social
00:58:31.598 --> 00:58:36.018
contagion thing for entrepreneurship is is like a real deal and so i imagine
00:58:36.018 --> 00:58:37.178
you could do the same kinds of,
00:58:37.698 --> 00:58:41.398
tricks there like they do everything from like randomly place people together
00:58:41.398 --> 00:58:45.778
with an entrepreneur versus another entrepreneur versus, yeah, so.
00:58:46.118 --> 00:58:49.838
I mean, this is sort of like, are they positing it's like a placebo,
00:58:50.038 --> 00:58:56.038
nocebo effect where you see somebody else experiencing something and just like
00:58:56.038 --> 00:58:59.838
because of the contagion of it, you experience it too, even if it might not be real.
00:59:00.882 --> 00:59:04.802
Are you talking about adolescent mental illness or entrepreneurship? Okay.
00:59:05.782 --> 00:59:08.102
Well, I mean, you also see what gets rewarded and what doesn't,
00:59:08.242 --> 00:59:12.062
right? So if someone gets a lot of social attention, especially at that point,
00:59:12.342 --> 00:59:17.282
maybe you start seeing what gets rewarded as well.
00:59:18.022 --> 00:59:21.102
One of the cooler studies I've seen is in their book, Connected,
00:59:21.242 --> 00:59:26.882
Fowler and Kostakis show that the rate of back pain in East and West Germany,
00:59:27.162 --> 00:59:30.942
in West Germany, probably I'm to make up numbers.
00:59:31.102 --> 00:59:34.462
20% of the people had back pain. East Germany, none.
00:59:35.082 --> 00:59:39.202
Wall falls down. All of a sudden, all those people in East Germany start developing back pain symptoms.
00:59:39.962 --> 00:59:41.822
Why? I mean, it's like a naturalistic experiment.
00:59:42.542 --> 00:59:45.762
It's interesting. I think the thing you have to, and maybe they do this already,
00:59:45.942 --> 00:59:50.582
is you could imagine a thing where it's like, I didn't know that there was hope
00:59:50.582 --> 00:59:54.742
for back pain. I didn't know there was treatments for it. I just lived with it.
00:59:54.882 --> 00:59:56.862
And then I'd meet these guys who have it. And they're like, oh,
00:59:56.882 --> 01:00:00.222
no, I went to the doctor and then you go talk to your doctor and now you get
01:00:00.222 --> 01:00:02.082
clicked as also having back pain.
01:00:02.082 --> 01:00:03.842
I can complain about this sweet
01:00:03.842 --> 01:00:07.762
Yeah i thought that i thought you were making like a pharmaceutical commercial
01:00:07.762 --> 01:00:14.542
like it's just your back hurt you need like have you tried exebo you know that
01:00:14.542 --> 01:00:19.062
helps with lower back pain and that spread from west germany to east germany
01:00:19.062 --> 01:00:24.202
Well i i think that clearly like the social contagion works especially like
01:00:24.202 --> 01:00:27.082
a micro level like we've all experienced this situation like we're in a good
01:00:27.082 --> 01:00:31.822
mood and like seven walks in and they just like kills the vibe right like they're
01:00:31.822 --> 01:00:35.402
mean as hell or something like oh jesus christ yeah yeah
01:00:36.602 --> 01:00:39.542
well that's a good way to end it yeah
01:00:39.542 --> 01:00:40.982
You know matt like
01:00:42.023 --> 01:00:45.763
I mean, meeting you today has been cool and seeing your sub stack. It's fantastic.
01:00:46.063 --> 01:00:49.443
You've clearly got a lot of interesting thoughts on a lot of topics,
01:00:49.443 --> 01:00:52.563
and I feel like I've already learned a lot from you. So, well,
01:00:52.603 --> 01:00:53.643
thanks for joining us today.
01:00:53.883 --> 01:00:57.163
Scott, any final words for Matt before we give him the final word?
01:00:57.603 --> 01:00:59.943
Yeah, Matt, it's great to see you
01:00:59.943 --> 01:01:03.463
again. Tell folks how they can get in contact with you if they want to.
01:01:03.743 --> 01:01:10.823
Yeah, yeah. So thanks for having me on. I'm on Twitter in blue skies, Matt S. Clancy.
01:01:12.023 --> 01:01:15.943
There's my website, newthingsunderthesun.com. And then there's the Substack
01:01:15.943 --> 01:01:19.243
version, which I think is also mattsclancy.substack.com.
01:01:19.323 --> 01:01:22.763
But you just Google New Things Under the Sun or Matt Clancy and I'll probably pop up.
01:01:23.323 --> 01:01:25.103
Repeat, New Things Under the Sun.
01:01:25.243 --> 01:01:28.323
That's right. New Things Under the Sun.
01:01:28.623 --> 01:01:31.583
My disclaimer is Scott told me the wrong name. I did.
01:01:33.083 --> 01:01:36.383
It is kind of about, well, it's about all things innovation,
01:01:36.383 --> 01:01:37.883
I'd say, in science and technology. week.
01:01:38.203 --> 01:01:42.103
So when you heard it here, Matt's an ex and a blue sky guy, but not a threads
01:01:42.103 --> 01:01:43.483
guy. So you can't find him there.
01:01:44.943 --> 01:01:48.103
I think I do have a threads account. I just never check it. Yeah.
01:01:48.403 --> 01:01:51.603
Well, you've been listening to Direction and Correct, the People Analytics podcast
01:01:51.603 --> 01:01:54.923
with Colin Scott and today's guest, Matt Clancy. Thanks for joining us, Matt.
01:01:55.323 --> 01:01:55.883
Thanks a lot, guys.
01:01:56.583 --> 01:02:00.423
Direction and Correct is dedicated to you, our listeners, to help educate and
01:02:00.423 --> 01:02:03.383
entertain you on how to effectively do people analytics.
01:02:03.723 --> 01:02:07.963
By supporting this podcast, you're helping us continue to provide valuable insights
01:02:07.963 --> 01:02:09.443
and knowledge to our listeners.
01:02:09.843 --> 01:02:12.363
Please consider becoming a patron of the podcast.
01:02:12.863 --> 01:02:18.023
You can find the link to sign up in the show notes or at patron.podbean.com
01:02:18.023 --> 01:02:20.463
slash directionallycorrect. Thanks for your support.