WEBVTT

1
00:00:00.001 --> 00:00:01.060
So if it's okay.

2
00:00:01.290 --> 00:00:07.840
So Jan, Luca, I am Massey, and I am in the past and chair of the

3
00:00:07.960 --> 00:00:12.490
the site in the and we have now on

4
00:00:12.950 --> 00:00:28.910
E. C.

5
00:00:29.230 --> 00:00:33.189
In a question, discussion, time.

6
00:00:33.270 --> 00:00:44.719
so on. So on side people will, and make his the questionable comment at the at the front of the camera. So please make a discuss

7
00:00:44.820 --> 00:01:01.810
Gianluca Manzo: good. Good. Thank you very much. Thank you very much, master, for handling this on site. it's it's it's it's it's clearly a quite complex organization. But well, hopefully, we will be able to to make it

8
00:01:02.060 --> 00:01:08.950
Gianluca Manzo: Well, so we are almost on on on time. I'm not sure.

9
00:01:09.000 --> 00:01:24.689
Gianluca Manzo: If Clements is already connected. but he certainly will in a few minutes. Well, let me just thank you all for accepting. Well, to give a talk in this

10
00:01:24.730 --> 00:01:35.570
Gianluca Manzo: in this section that I suggested to to title what makes society credible? But the idea was simple, simply

11
00:01:35.770 --> 00:01:39.279
Gianluca Manzo: to discuss or to continue to discuss.

12
00:01:39.320 --> 00:02:01.690
Gianluca Manzo:  what are the principles that can help us to distinguish rigorous sociology from not rigorous sociology. So there is a debate in various areas of of contemporary from this topic. Obviously it is not a

13
00:02:01.800 --> 00:02:13.409
Gianluca Manzo: a a recent debate, but well, we continue to discuss about this, and the idea was to to continue this this debate. And so we we have

14
00:02:13.440 --> 00:02:30.029
Gianluca Manzo: 4 great speakers from different part of the world. Emily Erickson, who is connected from from the States. Hello, Henry, So he's professor also. Sorry. Called Yale University.

15
00:02:30.280 --> 00:02:36.910
Gianluca Manzo: Then we have Andreas who is also Professor of Sociology. Hello, Andreas.

16
00:02:37.040 --> 00:02:49.270
Gianluca Manzo: at the University of Groningen. Then we have Clement's core number who is not connected yet, but he will. Who is Professor of the University of Colon.

17
00:02:49.300 --> 00:02:59.490
Gianluca Manzo: and then we have Lucas and Lucas, who is Max? Labor to do our fellowship at university.

18
00:02:59.510 --> 00:03:14.120
Gianluca Manzo: European University Institute in in in Florence, and finally. we will have Daniel lately. Little who is Professor of philosophy philosophy at

19
00:03:14.489 --> 00:03:33.839
Gianluca Manzo: Michigan University. Sorry and who recorded that? A short video for us? because he finally wasn't able to to attend to the conference for personal reasons. So he apologize. But here was so nice to record. a short video for for us.

20
00:03:33.840 --> 00:03:50.149
Gianluca Manzo: And I suggest that I I will share my screen. We do V to at the end of the of the full for first 4, and and I suggest that we follow the the order of the program so Hamilton can start then

21
00:03:50.150 --> 00:03:53.369
Gianluca Manzo: claimants then.

22
00:03:53.630 --> 00:03:59.379
Gianluca Manzo: sorry Lucas. A second position. Then Clements and finally Andreas

23
00:03:59.490 --> 00:04:20.270
Gianluca Manzo: and Emily. Obviously, if you want to leave after your talk, even the to time, Lot, that's perfectly fine for me. I I I perfectly understand. So please don't don't worry, and that's that's that's fine and So we will try to answer questions for for you, even though obviously, it's not easy. But we will try.

24
00:04:20.510 --> 00:04:28.110
Okay? So well, I can, we can. We can. We can start Well, as you know. I suggested

25
00:04:28.140 --> 00:04:35.530
Gianluca Manzo: that you have a 15 min to talk and then we have 5 min for questions.

26
00:04:35.730 --> 00:04:41.180
Gianluca Manzo: but we can also decide to have questions at the end of let's see how things evolves.

27
00:04:41.490 --> 00:04:42.410
Gianluca Manzo: Okay.

28
00:04:43.480 --> 00:05:03.570
Emily Erikson: okay, terrific. Thank you. Thank you very much. and thank you, John, look for organizing this. I I think it's and thank you everyone for being there. I wish I could be there in person. I'd probably be a little more awake. Possibly But I do think it's a really important question. I think it's a really important issue.

29
00:05:03.570 --> 00:05:30.110
Emily Erikson: and and I do think that sociology has a very significant credibility issue that that needs to be addressed in some way. It's methodologically pluralistic. It's substantively You could even say unfocused and often purely critical. And it's a fairly remarkable situation. given the number of social problems that the world faces

30
00:05:30.110 --> 00:05:56.979
Emily Erikson: and some widespread agreement that purely technological solutions at this point are not going to fix those problems because they are the results of social coordination issues. So if any field is a poised to address these kinds of, you know, emergent social coordination problems that are are so problematic at this point. then I would think it could at least be sociology.

31
00:05:57.100 --> 00:06:08.560
Emily Erikson: but we don't seem to be associated with solutions at this point, at least in mind of the general public or policymakers sometimes. so

32
00:06:08.660 --> 00:06:24.310
Emily Erikson: it will 1 one way that and then a number of of really significant and influential sociologists who tried to address this is Through the introduction of the the some field of analytical sociology.

33
00:06:24.490 --> 00:06:47.950
Emily Erikson: So in analytical sociology, there is a greater theoretical coherence and focus on some specific issues that distinguish it from other types of sociology, and and I think the part of the project was to give it more methodological rigor in order to specifically increase the credibility of the discipline.

34
00:06:47.950 --> 00:07:17.499
Emily Erikson: So these include focusing specifically on the micro macro link using the call and boat as a kind of a regular conceptual tool that that leads researchers to focus on this micro macrolink. there's also the emphasis on it, the sort of dbo framework it desires, beliefs, opportunities, rational choice theorists and causal mechanisms. and just being sort of more rigorous, more transparent, and precise in the language.

35
00:07:18.600 --> 00:07:38.299
Emily Erikson: Now I think that in some ways even though it makes sense that these things would increase the credibility and the influence of of sociology. Unfortunately, it hasn't entirely worked out that way, often because of resistance, even within sociology, to these sort of moves.

36
00:07:38.380 --> 00:07:44.939
Emily Erikson: So I have often heard other sociologists say that in fact

37
00:07:44.960 --> 00:07:56.989
Emily Erikson: they think that analytical sociologist, in fact, just the way of saying that the rest of sociology is not analytical and that the rest of it is is not sufficiently rigorous

38
00:07:57.070 --> 00:08:14.439
Emily Erikson: and in some sense I think it is true. that in analytical sociology, in this particular attempt to to build more credibility is currently mainly defined by its own standard of what constitutes better research.

39
00:08:14.560 --> 00:08:16.890
Emily Erikson: and this hasn't really

40
00:08:17.180 --> 00:08:24.869
Emily Erikson: helped it sort of gain the traction that I I think people have hoped it would gain in the discipline more broadly.

41
00:08:25.940 --> 00:08:34.049
Emily Erikson:  so I what I think is is a possibility, and I suppose this is the idea. And I I took, you know.

42
00:08:34.120 --> 00:08:44.589
Emily Erikson: the the problem as a real opportunity to think through things. So that's what this is. Me thinking through what might be another possible approach.

43
00:08:44.640 --> 00:09:08.740
Emily Erikson: and I think that analytical, I mean, we have so many significant, real world problems that I think we could do a better job linking analytical sociology to the object of providing a solution to those problems. and there is, there's actually a history within analytical sociology and within sociology more broadly.

44
00:09:08.740 --> 00:09:16.429
Emily Erikson: that of a having a strong focus on solving coordination or understanding coordination problems

45
00:09:16.430 --> 00:09:25.309
Emily Erikson: and the the lineage that I'm thinking of the lineage or the theoretical lineage is the work on unintended consequences.

46
00:09:26.350 --> 00:09:54.559
Emily Erikson: so, as everyone probably is very familiar, Merton developed this idea. I mean, you can, of course, trace it back to Mandeville, or or or probably even further. But you formalize and Merton formalize the idea in in 1,935 paper The paper was influential, however, also received significant criticism for confusing an anticipated unanticipating consequences

47
00:09:54.560 --> 00:09:59.579
Emily Erikson: and then conflating both of them with negative outcomes.

48
00:09:59.850 --> 00:10:16.440
Emily Erikson: a at, but it did at least it enter into the sort of public consciousness and broader discourse in an important way, and then I it back up by the Raymond Boudon in 1,977. In the book the Unintended Consequences. Social actions picked up by

49
00:10:16.640 --> 00:10:17.630
Emily Erikson: it.

50
00:10:17.820 --> 00:10:45.099
Emily Erikson: It's it's picked up later by didn't. But at Boudon's book also, was not particularly well received at the time that it was published in 1,977 it. So there's problems apparently as poorly translated. you know, it did focus fairly narrowly on his own research and interest rather than casting sort of a wider net about how the that theoretical frame it could be applied, and further

51
00:10:45.100 --> 00:10:49.429
Emily Erikson: And it was precise for not moving beyond what Merton had already presented.

52
00:10:49.740 --> 00:11:03.040
Emily Erikson: Then it was linked to Merton's functionalism, and then it was linked to rational choice theory. Which again hurt his popularity in the broader discipline in sort of 2 different directions, which is unfortunate.

53
00:11:03.040 --> 00:11:24.170
Emily Erikson: but you know it. It's still persists. However, it's picked up by Gary Ellen. Fine, in 2,006, Alejandro Porter's use it in his Presidential address to the essay in 2,010, and most recently, I There's been a series of books published by A. G on a meek at that. trying to sort of resuscitate.

54
00:11:24.470 --> 00:11:26.999
Emily Erikson: So now is this

55
00:11:27.020 --> 00:11:38.120
Emily Erikson: is this something that an that is already there, for example, within analytical sociology is the unattended consequences, idea, framework, a way that we could

56
00:11:38.390 --> 00:11:54.410
Emily Erikson: bring together the kind of rigor that exist in that local sociology together to an approach to solving problems that would be compelling to people say outside of Academia and also potentially within

57
00:11:54.410 --> 00:12:10.499
Emily Erikson: the social sciences. But outside of the analytical sociology tradition at this point, well, if we use the unattended consequences, I do think that there are some. So it's it's definitely a compelling idea. But there are some still shortcomings.

58
00:12:10.650 --> 00:12:31.160
Emily Erikson: so theoretically, it's been constructed as a way to think about rational action, and individuals as as motors of of activity. But Raymond but don't, for example, poses it as a theoretical response to the structural determinism, so preserving the causal importance of individuals.

59
00:12:31.310 --> 00:12:46.590
Emily Erikson: but so we I don't find that to be actually the most compelling aspect of it. I mean at this point our complete inability to accurately predict almost any social outcome. doesn't make me very concerned about

60
00:12:46.590 --> 00:13:05.700
Emily Erikson: Determinism, although deterministic thinking, of course, is problematic. But I would say that complexity sciences has given us a a really a much more powerful way to think about navigating between the ideas of sort of structure, and an individual free will and choice.

61
00:13:06.640 --> 00:13:22.350
Emily Erikson: So I would think that the greatest strength of unattended consequences and the importance of the relationship between it and analytical sociology is that that usually the action is conceived at the micro level, and and the outcome is conceived at the macro level.

62
00:13:22.490 --> 00:13:46.220
Emily Erikson:  so that's really that's really crucial. But it poses another problem for this, the concept of unintended unintended outcomes, which is that emergent behavior, not this is just the start of the problem. But emergent behavior is then of central interest right? But a version behavior

63
00:13:46.630 --> 00:14:00.019
Emily Erikson: should be of interest to sociologists that should be of interest, analytical sociologists, and it seems to fit somewhat naturally into the framework of unintended consequences, except that because it's not ever intended.

64
00:14:00.020 --> 00:14:21.059
Emily Erikson: it's also not unintended right. It's never a side effect of the intentional act of an individual. It's actually a collection of of many different people. So at what point can you say, for example, that a group or an emergent outcome is either intended or unintended, actually doesn't quite fit into that sort of

65
00:14:21.060 --> 00:14:24.190
Emily Erikson:  evaluative standard.

66
00:14:24.770 --> 00:14:25.860
Emily Erikson: So

67
00:14:26.400 --> 00:14:45.080
Emily Erikson: does it matter. You know what is the subjective location of the definition of the unintended? Can a group in 10 something? Can a group not to intend to something? and in when we get into this sort of problem, also, it is it unintended?

68
00:14:46.980 --> 00:15:04.619
Emily Erikson: how do we determine? Because it's a subjective state of the individual, whether they intend to take some action or not, how can we as reach? There's always be certain whether or not the action is intended or unintended, that that becomes a very sort of a difficult problem. I would say.

69
00:15:06.370 --> 00:15:08.410
Emily Erikson: so

70
00:15:08.660 --> 00:15:24.139
Emily Erikson: for example, you know, what what does this mean for strategic actors is for is robust action, and intended or unintended consequence? Is, is it it necessary for the subject to be considered a a naive.

71
00:15:24.140 --> 00:15:38.850
Emily Erikson: if their actions that have other consequences, and the other than the obvious ones. But the actors are aware of the potential for those outcomes. Does that, then, exclude these outcomes from consideration because they're not unintended.

72
00:15:38.850 --> 00:16:02.410
Emily Erikson: Right? So this is just some of the some of what the problems. But probably the the worst problem is that the idea of the unattended consequences as a theory is self defeating, because once you, as a researcher, reveal the relationship between the intent and the unintended, and it makes the unintended, intended right? And that's the goal is you're trying to understand what are the range of possible outcomes?

73
00:16:02.410 --> 00:16:18.160
Emily Erikson: so you you're essentially narrowing the field of study as you go along, by revealing what the possible outcomes are to individuals, so that they can include it in in the decision that they're making and make them for it intended.

74
00:16:18.380 --> 00:16:21.019
Emily Erikson: Hmm. so

75
00:16:21.320 --> 00:16:39.589
Emily Erikson: so I think I think it's It's it's really important, but that it's it's also problematic in a way that makes it impossible to kind of apply to all the sets of problems that we would want to apply. And this is perhaps,

76
00:16:40.800 --> 00:16:59.250
Emily Erikson: yeah, just a matter of of terminology. but I don't think it is. I think it has. So. I mean in the title of the talk here that I suggest, is secondary consequences, secondary consequences, Macro consequences. These are These are different ways to to think about

77
00:16:59.400 --> 00:17:11.540
Emily Erikson: or or titles that we could potentially use that would allow us to fall into this sort of trap of always thinking about the motivations of of the actors when there's there's, you know.

78
00:17:11.640 --> 00:17:22.789
Emily Erikson:  other things that are also potentially causing outcomes and changing outcomes that are important, that we want to certainly pay attention to such as, for example, network structure.

79
00:17:23.070 --> 00:17:52.009
Emily Erikson: but one issue that I want to raise. That, I think, is really important is not just this terminological issue, but also, then the the relative importance. It puts on computational models versus causal mechanisms. And I'm almost done. And it's going to be 2 2 s, really. But if we're making causal analysis central to a model of of like figuring out unintended consequences or related secondary consequences.

80
00:17:52.010 --> 00:18:19.770
Emily Erikson: The problem is very significant is that causal analysis is based on the counter factual model. which presumes that you can compare a treated unit to an untreated unit. but the most basic effects of that that we see that are usually considered to be unintended. Consequences are usually affects, like the tragedy of the comments where the outcome is different. If one individual acts one way, and all individuals act that way

81
00:18:19.820 --> 00:18:45.569
Emily Erikson: so you can't compare or treated in an untreated because you can't. If you separate out a control group, the effect. The outcome is different. And that's gonna be a problem. That's that that's going to be a problem for all causal analysis. When you have these issues of interlocking aggregate effects, unintended consequences, secondary consequences, whatever. they end up being called

82
00:18:45.570 --> 00:19:00.849
Emily Erikson: So that to me suggests that, I think that we really do need to think about the relative advantage of computational models and thinking about systems and interdependence. and that that they

83
00:19:01.070 --> 00:19:03.010
Emily Erikson: causal

84
00:19:03.420 --> 00:19:12.349
Emily Erikson: the the relative advantage of of analytical sociology and sociology, I think, probably lies in the the modeling.

85
00:19:15.060 --> 00:19:16.530
Emily Erikson: That's it. Thank you.

86
00:19:18.680 --> 00:19:21.240
Gianluca Manzo: Thank you very much, Emily, very.

87
00:19:21.410 --> 00:19:22.940
Gianluca Manzo: very interesting.

88
00:19:25.820 --> 00:19:36.730
Gianluca Manzo:  I see 2 2 hands already, Andreas and Clemens. So we have a 3 min for a couple of questions. And so, Andreas, please go ahead.

89
00:19:37.350 --> 00:19:45.630
Andreas Flache: Thank you. That was mainly a clapping hand, but I also would have questioned, Thank you, Emily. very estimating

90
00:19:46.150 --> 00:19:47.780
Andreas Flache: talk.

91
00:19:47.810 --> 00:19:53.550
Andreas Flache: so yeah, I mean, I could agree more with what you say, that a a very strong

92
00:19:53.770 --> 00:20:04.939
Andreas Flache: erez agmoni element in analytical sociology is the modeling. And I I also very much agree that computation and modeling from a say, complexity, perspective 150

93
00:20:05.240 --> 00:20:10.079
Andreas Flache: probably really can help us to better understand this. Yeah.

94
00:20:10.170 --> 00:20:18.110
Andreas Flache: social phenomena in which things happen which people at least do not necessarily anticipate to happen.

95
00:20:18.650 --> 00:20:27.039
Andreas Flache: Maybe maybe an an an element about which I'm curious to hear your thoughts could be that

96
00:20:27.090 --> 00:20:31.479
Gianluca Manzo: I believe we we can most fruitfully use computational modeling in

97
00:20:32.100 --> 00:20:40.309
Andreas Flache: in combination with and in interaction with empirical work. And then, in particular experimental work also.

98
00:20:40.350 --> 00:20:50.680
Andreas Flache: erez agmoni where experimental work often can help us to get a better, better understanding of the microprocesses, say how people socially influence each other in an online setting, for example, 150,

99
00:20:50.740 --> 00:20:57.819
Andreas Flache: and then computational modeling can help us to better understand to what actually, this aggregates up under certain conditions.

100
00:20:57.860 --> 00:21:00.720
Andreas Flache: So I'm curious what your thoughts are on this.

101
00:21:02.150 --> 00:21:09.109
Emily Erikson: Well, I definitely you know. I I mean, I think all it empirical and theoretical

102
00:21:09.180 --> 00:21:36.710
Emily Erikson: approaches are are welcome and add to our wealth of knowledge. So I'm I'm not saying that we should not do experimental methods. I'm personally very fond of compared to historical methods, because it's a way to do empirically rigorous work that doesn't take people outside of the messy complexity of reality. But it allows you to sort of investigate. What's what's going on?

103
00:21:36.730 --> 00:21:39.570
Emily Erikson: in the settings? And

104
00:21:40.300 --> 00:21:49.689
Emily Erikson: to again, you know, you don't have the the right environment for precise causal measurement, but

105
00:21:50.050 --> 00:22:02.810
Emily Erikson: it also seems as though you know people act somewhat differently inside the laboratory than outside of the laboratory. so I so I do think that you know, absolutely there's definitely

106
00:22:03.240 --> 00:22:05.950
Emily Erikson: still an important place for experimental methods

107
00:22:06.150 --> 00:22:15.269
Emily Erikson:  and and causal inference. I mean, I I do think I think causal inference has

108
00:22:15.900 --> 00:22:37.289
Emily Erikson: taken center stage and political science and economics in a way that it hasn't in sociology, and and that that again, that I see. Maybe the value added in sociology, being in moving a little bit away from this, you know, sort of like.

109
00:22:37.480 --> 00:22:49.009
Emily Erikson:  very driven focus on on causal identification. So that. That's where I'm sort of suggesting we maybe move a little bit in a different direction.

110
00:22:52.900 --> 00:22:53.580
The

111
00:22:53.780 --> 00:23:13.869
Gianluca Manzo: I I I'm I'm so sorry, but we we we have a finished the 20 means lots for you. I mean, I have my myself some question that I perhaps I will share with you by by by email. And so thank you very much. And if you want to stay with us, I'm

112
00:23:13.870 --> 00:23:24.509
Gianluca Manzo: more than happy if you want to leave because of the time. Lag. Please don't don't worry. Okay, thank thank you very much again for for being with with us.

113
00:23:24.530 --> 00:23:35.980
Gianluca Manzo: okay, so I I suggest to you more of what to do. The second presentation, which is given by Lucas. Right? According to the

114
00:23:36.180 --> 00:23:43.530
Gianluca Manzo: the program. Order. And so, Lucas. please go ahead for 50 min.

115
00:23:44.410 --> 00:23:50.589
Lucas sage: Okay, I hope everybody can hear me and see my screen.

116
00:23:51.930 --> 00:23:54.520
Gianluca Manzo: Yeah, we can.

117
00:23:55.340 --> 00:23:59.200
Lucas sage: All right. so

118
00:23:59.520 --> 00:24:00.430
Lucas sage: yes.

119
00:24:00.630 --> 00:24:16.550
Lucas sage: So I'm going to talk about the paper, which is not yet a paper. But so that is more. an illustration. Let's say of what I I I hope, is a

120
00:24:16.830 --> 00:24:30.269
Lucas sage: credible piece of sociology rather than a a a theoretical discussion, a percent about the principles of so

121
00:24:30.300 --> 00:24:55.290
Lucas sage:  What is the the the motivation that's in the starting point behind this? this study? well, they recently they have been, some pieces of research interested in, like the decline of disruptivity or innovativeness of of research.

122
00:24:55.370 --> 00:25:04.410
Lucas sage: So the the here I illustrate with some a few articles, but they are more than this, and covering different disciplines.

123
00:25:04.500 --> 00:25:20.700
Lucas sage: And so even with various methods, there is this this conclusion, this empirical pattern that science is becoming less and less innovative and less disruptive over time.

124
00:25:21.020 --> 00:25:26.310
Lucas sage:  I. Yes, okay.

125
00:25:26.430 --> 00:25:40.090
Lucas sage: So my question is, what what can we do about it? So and I have a I would like to bring an intuition from, let's say, from social networks up from complex system into it.

126
00:25:40.300 --> 00:25:52.719
Lucas sage:  so there are 2 pieces 2 2 parts in in in the idea that I will explore in this. In this paper

127
00:25:52.770 --> 00:26:02.060
Lucas sage: the the first thing is So certain authors are probably certain scientists, certainly more innovative than others.

128
00:26:02.440 --> 00:26:03.530
Lucas sage: And

129
00:26:03.760 --> 00:26:09.710
Lucas sage: is it possible? Do do do this spread? they are

130
00:26:09.740 --> 00:26:22.240
Lucas sage:  to their to their quarters. Let's say to the people they collaborate with, so that, in other words, can you learn from your quarters to be more innovative?

131
00:26:22.480 --> 00:26:29.800
Lucas sage:  and if this is the case, and that would be the second part of this of this project.

132
00:26:29.900 --> 00:26:55.370
Lucas sage: maybe there's the possibility that we under exploit this this mechanism. collectively, let's say, because there is this tendency this, a sort activity tendency, which is that? let's say, the most innovative scientist only collaborate with the most innovative orders. And so collectively, we we? We don't exploit

133
00:26:55.460 --> 00:26:59.590
Lucas sage: this this mechanism.

134
00:27:00.050 --> 00:27:05.269
Lucas sage: enough, let's say  So yes.

135
00:27:05.620 --> 00:27:12.539
Lucas sage: So the idea is first to estimate whether there is this a social learning effect.

136
00:27:12.640 --> 00:27:18.230
Lucas sage: and this is what I'm going to focus on mainly today. and

137
00:27:18.400 --> 00:27:32.680
Lucas sage: then, if the second part of the the project would be to then integrate this finding into a nonprofit collaborated agent, based model and then

138
00:27:33.550 --> 00:27:47.729
Lucas sage: modify count of actually who collaborates with Zoom. So the the emergent network structure and assess the overall effect of these of this modification of the team formation

139
00:27:47.890 --> 00:27:53.479
Lucas sage: mechanism on the overall level of innovators.

140
00:27:53.880 --> 00:28:08.570
Lucas sage: So as I said so far, it's it's a work in progress. So I will. only Mentioned the the the the first part in the preliminary results of this first part.

141
00:28:08.960 --> 00:28:38.449
Lucas sage: so, How will I do this? A few words on the on the measures? So well, I'm going to use bibliometric data that allows us to to track precisely interactions between scientists. Because, like, we can assume that. if people co-OP their paper, they have really been interacting before. And this is So so they they, they have this chance to learn from from from from each other.

142
00:28:38.590 --> 00:29:07.219
Lucas sage: And the second point that is very important is the measure of innovativeness that I will use, and I rely on on the at typical combination of knowledge idea by Uzi and his colleagues. And so let me explain to you. So here is just a an illustration of a a paper, and at the end of each paper you have a reference list. So the the quotes that the A paper make

143
00:29:07.320 --> 00:29:21.869
Lucas sage: and for each pair of I'm I'm giving you the intuition. Here, let's say, for each pair of references in the overall reference list

144
00:29:21.910 --> 00:29:38.050
Lucas sage: you can define, you can measure whether this quoting these 2 journals, let's say you the P. And A. S. And plus one is it something that is done very often. And so it's quite conventional, let's say.

145
00:29:38.050 --> 00:29:58.389
Lucas sage: or here, if you look at this other pair that I mentioned here is quoting a a journal of a medicine, and another one of a working paper and economics. And so this para is quite atypical. And this is going to be, let's say, our measure of of innovativeness. So

146
00:29:58.480 --> 00:30:09.979
Lucas sage: overall for each at school I get a distribution of zed scores, and I'm going to take the the the value of the first decide

147
00:30:10.050 --> 00:30:17.119
Lucas sage: as as my measure of of this distribution of this course. that will be my measure of

148
00:30:17.420 --> 00:30:23.199
Lucas sage: the paper. The papers at typicality or innovators.

149
00:30:23.650 --> 00:30:33.680
Lucas sage:  then, what we need to do is to get a data set in which we have a

150
00:30:33.800 --> 00:30:45.700
Lucas sage: we, we need panel data to identify this this social interaction effect. And we need time consistent.

151
00:30:45.760 --> 00:31:00.629
Lucas sage: author. Ids, let's say. And so this is not so easy, actually, the the online platforms that give you a access to this kind of data. they, they have this ambiguation algorithm not working very well.

152
00:31:00.650 --> 00:31:18.499
Lucas sage: So what we decided to focus on is we, we took the disambiguation algorithm that was used in this paper by a sinatra Barabazi and and their colleagues. in which basically, we have

153
00:31:19.140 --> 00:31:28.969
Lucas sage: all papers that we are published in the family of a physical review between the 1893 and 2,010.

154
00:31:29.220 --> 00:31:33.869
Lucas sage: And so I I I use the time, constant author, id.

155
00:31:34.490 --> 00:31:44.700
Lucas sage: and then I gather the data about the the each publications reference list. And I computed my metric of Innovativeness.

156
00:31:45.010 --> 00:31:56.820
Lucas sage: And I'm going to focus only on the first 15 years of the carriers of the scientists, the physicists that I have. in in this data.

157
00:31:58.200 --> 00:32:23.290
Lucas sage:  how am I going to estimate the effect of social interactions on on the innovativeness. So I I I was inspired. Let's say I I copied the the methodology of a paper that is going to to appear soon in econometric that

158
00:32:23.520 --> 00:32:31.149
Lucas sage: And so the idea is that we have a publication. Queue is the the measure of innovativeness

159
00:32:31.250 --> 00:32:35.960
Lucas sage: of the team. J, that produce this paper

160
00:32:36.230 --> 00:32:43.800
Lucas sage: at time. T. And what I do is, I define for each publication a team leader which is.

161
00:32:43.890 --> 00:32:48.280
Lucas sage: among all the authors of this publication.

162
00:32:48.440 --> 00:32:52.149
Lucas sage: the the the out there with the

163
00:32:52.710 --> 00:32:56.360
the most experienced author in the team.

164
00:32:56.370 --> 00:32:59.280
Lucas sage: And so for this person I compute.

165
00:32:59.310 --> 00:33:13.409
Lucas sage: I'm going to explain. I'm going to come back on the interactions measure. But the let's say, these are the past interactions up until the the T minus one of these

166
00:33:13.410 --> 00:33:33.699
Lucas sage: of this team leader. and then we have a measure of the age of the team leader age square, the team size and a set of a fixed effect notably individual or team leader fixed effect, but also region, and the yearly fixed effect.

167
00:33:33.820 --> 00:33:42.339
Lucas sage: And so what is behind this interaction? measure. So here I illustrated the one other. Let's say, author, one

168
00:33:42.380 --> 00:33:53.349
Lucas sage: that, as in period t, as as the so far, 2 collaborations. So with the with one with 2 other, and one with one of those.

169
00:33:53.480 --> 00:34:03.150
Lucas sage: and when we move on to t one This person starts a new collaboration with the the fifth node in in this network.

170
00:34:03.510 --> 00:34:17.369
Lucas sage: And so I'm going to compare the the let's say, the cumulative innovativeness, the cumulative score of innovativeness of these of these 2 authors. And so, if

171
00:34:17.570 --> 00:34:23.329
Lucas sage: so far in T one person, 5

172
00:34:23.430 --> 00:34:32.529
Lucas sage: the fifth author as a higher productivity than a productivity a higher innovativeness.

173
00:34:32.540 --> 00:34:46.330
Lucas sage: Then, my, my, the note I'm focusing on which is the person one then discounts as one one interaction with someone that was more innovative.

174
00:34:46.380 --> 00:34:55.290
Lucas sage: Okay? And so if we come back to this idea interaction here is, that is, the is the cumulative

175
00:34:55.980 --> 00:35:08.949
Lucas sage: Some of all past interactions with people that were at the moment of the collaboration, more innovative than the team leader

176
00:35:09.270 --> 00:35:17.779
Lucas sage: of of this paper. So I hope this was clear, because it's not so easy. And I move on to the

177
00:35:17.860 --> 00:35:36.370
Lucas sage: descriptive immediately. So my measure of innovativeness is standardized. And so I I'm I'm transformed it to make it a positive. Only so that it's easier to to to interpret

178
00:35:37.100 --> 00:35:51.710
Lucas sage: And then so what this is telling us is that on average, over the first 15 years of their carriers individuals have 2 point 35 interactions with

179
00:35:51.790 --> 00:35:53.609
Lucas sage: higher

180
00:35:54.260 --> 00:36:16.550
Lucas sage: innovative co-authors, and they have a bit more interaction with lower with with lower innovative. And you have the descriptive also about the the team team leader age in the overall data set in the final.

181
00:36:16.930 --> 00:36:30.979
Lucas sage: So these are more classic results. But let's say, The number of papers per author is a asymmetrically distributed, such as the same is true for the team size.

182
00:36:31.280 --> 00:36:48.200
Lucas sage:  so we see that the cumulative innovativeness of authors is obviously increasing over time quite fast in the first years, and they are progressively decreasing.

183
00:36:48.340 --> 00:36:55.199
Lucas sage: And yeah, I also plotted the the number of interactions over over time.

184
00:36:55.280 --> 00:37:13.149
Lucas sage: over the carrier of people. maybe I go quick because I'd like to move on to the results. But the the paper at T which is my dependent viable, is slightly asymmetrically distributed.

185
00:37:13.290 --> 00:37:23.170
Lucas sage:  yeah. So I move on to the to the the results of the regression.

186
00:37:24.130 --> 00:37:35.390
Lucas sage: so basically, we found we found what we expected to find, which is that So the the the valuable of interest is the is the first valuable

187
00:37:35.470 --> 00:37:49.509
Lucas sage: the first role of this data set. And so either if I take a log of the innovative or the the innovativeness on its own, I find that

188
00:37:49.720 --> 00:37:59.520
Lucas sage: more interactions, more past interactions by the team leader in increase the innovators

189
00:37:59.620 --> 00:38:23.490
Lucas sage: of the of the project. So there seem to be this effect. Obviously, the question is about whether this effect is is truly causal. Well, so there is an endogenity issue. Maybe I I believe that the individual fixed effect already remove part of the of the concern. Also.

190
00:38:24.189 --> 00:38:26.540
Lucas sage: because

191
00:38:26.950 --> 00:38:34.149
Lucas sage: I mean we we we can discuss about it. But the idea is then to instrument to to use an instrument to estimate this effect.

192
00:38:34.160 --> 00:38:47.830
Lucas sage: but we we can talk about it more in the original paper. It doesn't change so much the the difference. But the user citations rather than the innovative of papers.

193
00:38:48.390 --> 00:38:50.849
Lucas sage: And yeah, so the next step

194
00:38:50.920 --> 00:39:10.740
Lucas sage: would be then to integrate this into an Abm where I would transform the the, the, the team formation mechanism and see the spread of the innovators and the overall innovators in the the population. And yeah, so this

195
00:39:11.510 --> 00:39:12.620
Lucas sage: let's say.

196
00:39:12.650 --> 00:39:30.449
Lucas sage: to to come back to the title of the of the session. my idea of credible sociology would be this integration of rigorous. I I hope you found it the rigorous empirical data analysis

197
00:39:30.880 --> 00:39:47.329
Lucas sage: and bring it into the formal Atms, where? Where the mechanisms, the effect of this mechanism that we care about have been documented in the in the first part. yes, and I I stop here. because I'm already out of time.

198
00:39:47.550 --> 00:39:48.520
Lucas sage: Thank you.

199
00:39:57.730 --> 00:40:03.420
Gianluca Manzo: Thank you very much. Lucas, for this very innovative.

200
00:40:03.470 --> 00:40:09.970
Gianluca Manzo: and recall stork. is there any questions in the room? Or from the

201
00:40:10.040 --> 00:40:12.210
Gianluca Manzo: online world?

202
00:40:17.500 --> 00:40:18.660
one President.

203
00:40:19.290 --> 00:40:20.690
Gianluca Manzo: yes, please go ahead.

204
00:40:23.150 --> 00:40:23.830
Yeah.

205
00:40:24.090 --> 00:40:28.509
You know. Look at this is the I'm talking about. I go making him his pronouns.

206
00:40:28.550 --> 00:40:48.489
I'm a Phd. Spin, too. My question is, when you define innovativeness. How are you? How is it different from interdisciplinariness versus creativity versus innovation. Because people can cite other people into disciplinary without

207
00:40:49.270 --> 00:41:06.260
creating anything new or innovation. Because, for example, I'm I'm in computer science. I'm bringing intersectionality theory into some of my research. Sometimes it's just improving demographic analysis without in awaiting anything. So I'm just curious about interdisciplinary versus creativeness versus innovation.

208
00:41:09.010 --> 00:41:09.960
Lucas sage: Hmm.

209
00:41:10.120 --> 00:41:15.439
Lucas sage: thank you. Thank you very much for your question.

210
00:41:15.940 --> 00:41:28.480
Lucas sage: I haven't. I mean, it's a very good point I I don't have like a a maybe I haven't thought enough to through this.

211
00:41:29.040 --> 00:41:41.960
Lucas sage: I would say that it's a it's a measure that it as became a conventional, let's say, in in a specific literature, on it's a sociology of science or science of science.

212
00:41:42.110 --> 00:41:56.410
Lucas sage: I would say it measures a a tendency, or at least in that time, to be innovative, but also like when in the original paper, they analyze like they they

213
00:41:56.410 --> 00:42:12.479
Lucas sage: they look at the impact of publications that are. let's say that do more atp call combination and they find that like, there is a a sort of sweet spot in which

214
00:42:13.010 --> 00:42:26.469
Lucas sage: papers, I mean, it's a combination of being both conventional and introducing some at typical combinations that make a a paper more likely to be very impactful.

215
00:42:26.770 --> 00:42:29.410
Lucas sage:  So

216
00:42:29.450 --> 00:42:43.079
Lucas sage: yeah, there, there is a link with the with impact, at least. But then you're right. Maybe it's not a I. I do not differentiate between the everything, all, all the concept that you mentioned.

217
00:42:43.120 --> 00:42:46.890
Lucas sage: I I will think about this more. Thank you.

218
00:42:50.980 --> 00:43:20.410
Gianluca Manzo: Thank you for the question, and thank you, Lucas, for for your reply. again, I'm afraid. But we we we have to to move on to the next presentation. Then we can go back to your port during the the overall. This cash on So thank you very much again, Lucas. And now I I will give the the floor to Clemens. again for a 15 min, and then for a couple of questions. so, Clements, please go go ahead.

219
00:43:21.960 --> 00:43:26.150
Clemens Kroneberg: Yeah, thank you very much. Andaluka. Thanks everyone for joining our session.

220
00:43:26.210 --> 00:43:37.260
Clemens Kroneberg: which I find really important that these days there's also other sciences, I think, become more reflexive about the question of credibility of findings and so on.

221
00:43:37.270 --> 00:43:42.210
Clemens Kroneberg: So it's good to have this discussion. and even also to have it with such as team colleagues.

222
00:43:42.560 --> 00:43:48.140
Clemens Kroneberg: And I'm going to talk about analytical sociology and some methodological

223
00:43:48.210 --> 00:43:52.169
Clemens Kroneberg: the questions that that have came up in in recent years.

224
00:43:52.960 --> 00:44:00.909
Clemens Kroneberg: so just as a background. I think you all know that characteristic of analytical sociology is this interest and social mechanisms.

225
00:44:01.270 --> 00:44:13.540
Clemens Kroneberg: and the idea that we also formulate theoretical models that specify, specify how actions and interactions with rise to collect the phenomena. And I think one of the biggest challenge of this agenda has always been.

226
00:44:13.620 --> 00:44:17.140
Clemens Kroneberg: but, on the one hand, you would like to have abstract and precise models.

227
00:44:17.620 --> 00:44:27.149
Clemens Kroneberg: but, on the other hand, we also would like to have more realism than in more traditional approaches, such as rational choice theory. So, for example, we would like to allow by heterogeneous

228
00:44:27.220 --> 00:44:35.889
Clemens Kroneberg: agents, right? And this combination. and let, of course, to the importance of agent-based models about which there has flatter, will talk

229
00:44:36.030 --> 00:44:40.989
Clemens Kroneberg: later. And so, if I

230
00:44:41.060 --> 00:44:46.550
Clemens Kroneberg: directly start with the question of the session. What makes sociology credible, I think, as in other fields.

231
00:44:46.750 --> 00:45:02.029
Clemens Kroneberg: it's credible. And I would also say, more productive if we achieve a tight coupling of theory. And then parents. And and I think this is really also the the main challenge, because we easily have a development of research fields that are heavily empirical.

232
00:45:02.460 --> 00:45:13.569
Clemens Kroneberg: or, you know, theoretical communities that are rather detached from in vehicle research. So I think it's all about this a link. And that's why a methodology

233
00:45:13.690 --> 00:45:15.040
Clemens Kroneberg: for me. It's so important.

234
00:45:16.350 --> 00:45:25.299
Clemens Kroneberg: And I recently came across a very nice article by John Squirts, in the Journal of Mathematical Sociology. They can. I can only recommend to everyone

235
00:45:25.350 --> 00:45:27.269
Clemens Kroneberg: who has hasn't read it yet.

236
00:45:27.490 --> 00:45:34.280
Clemens Kroneberg: and he gets this very nice figure here on the Gauss between 2 you and daytime.

237
00:45:34.720 --> 00:45:44.569
Clemens Kroneberg: which I'm going to use to to organize my discussion in the first minutes. So basically, the idea is that we have these theoretical ideas just like,

238
00:45:44.600 --> 00:45:47.320
Clemens Kroneberg: Innovativeness and Lucas talk

239
00:45:47.330 --> 00:45:50.630
Clemens Kroneberg: right? And some people really start with data

240
00:45:50.760 --> 00:45:54.139
Clemens Kroneberg: and statistical tools like this, methodological models.

241
00:45:54.200 --> 00:45:59.950
Clemens Kroneberg: And to draw conclusions about, for example, the conditions that lead to more innovation

242
00:46:00.580 --> 00:46:09.460
Clemens Kroneberg: and others. try to translate the theory in theoretical models. such as tipping point models, game, theoretic models

243
00:46:09.680 --> 00:46:13.630
Clemens Kroneberg: and models of cumulative advantage, a model of polarization.

244
00:46:13.830 --> 00:46:22.279
Clemens Kroneberg: and then try to derive implications of such models and and test them with data. So this that's really the top down route

245
00:46:22.350 --> 00:46:29.670
Clemens Kroneberg: of development, or of making this link between theory and data. Where is the other? One is model the bottom up route.

246
00:46:30.170 --> 00:46:40.950
Clemens Kroneberg: And I, I would argue that in recent years we have seen most progress in this bottom up way of doing and social science. So one part is the causal revolution.

247
00:46:41.230 --> 00:46:44.369
Clemens Kroneberg: for example, in economics and also political science.

248
00:46:44.420 --> 00:46:50.830
Clemens Kroneberg: And the idea that we need to really try to identify causal effects and a more defendable assumption.

249
00:46:51.270 --> 00:46:59.509
Clemens Kroneberg: So we need to adopt experimental or quas experimental designs. And often this also is more policy relevant because we will try to

250
00:46:59.670 --> 00:47:12.909
Clemens Kroneberg: change something in the world and see whether it has the desired effect or not. At the same time one, sometimes one. That's what we did. The the the significance of such studies is so. For example, in my area of interest of crime

251
00:47:13.010 --> 00:47:18.149
Clemens Kroneberg: we have a rigorous research showing that most street lights would use crime.

252
00:47:19.040 --> 00:47:26.980
Clemens Kroneberg: It's important finding and really admire the research, but at the same time, I think there is no single theory on crime

253
00:47:26.990 --> 00:47:32.130
Clemens Kroneberg: that would be challenged by this finding right. It's compatible with all our few is about crime

254
00:47:33.140 --> 00:47:41.750
Clemens Kroneberg: and This has led some authors to say that the present method of logic and and statistical solutions to the replication crisis

255
00:47:41.970 --> 00:47:47.470
Clemens Kroneberg: that include such causal effect identification, but also open science practices

256
00:47:47.690 --> 00:47:55.550
Clemens Kroneberg: that these solutions will only help ensure solid stones. They don't help us build the house. I think building the house really means

257
00:47:55.900 --> 00:48:01.020
Clemens Kroneberg: to understand under what conditions call the effects are stronger or even absent.

258
00:48:01.250 --> 00:48:05.610
Clemens Kroneberg: and to come up with theoretical models that summarize our state of knowledge

259
00:48:05.690 --> 00:48:06.859
Clemens Kroneberg: in it is something.

260
00:48:08.080 --> 00:48:12.870
Clemens Kroneberg: and a similar challenge arises when you, when we do computational social science.

261
00:48:13.360 --> 00:48:21.940
Clemens Kroneberg: so we often start from process produce big data. And we have these great tools at at our disposal, such as machine learning

262
00:48:21.970 --> 00:48:27.309
Clemens Kroneberg: what other computationally intensive models that also allow us to predict.

263
00:48:27.440 --> 00:48:30.180
Clemens Kroneberg: patents and the data attend

264
00:48:30.680 --> 00:48:35.090
Clemens Kroneberg: sometimes better, sometimes less. But here, again.

265
00:48:35.100 --> 00:48:40.939
Clemens Kroneberg: one can ask, okay, is this the same as you know, achieving theoretical understanding.

266
00:48:41.050 --> 00:48:45.549
Clemens Kroneberg: it models. And and I would now that this is the same.

267
00:48:46.350 --> 00:48:58.150
Clemens Kroneberg: So This is where I would like to to come in. So one proposal by Peter Hadstrom recently has been to add theoretical models

268
00:48:58.180 --> 00:49:04.769
Clemens Kroneberg: to this line of research by focusing really only on those kind of entities and behaviors

269
00:49:04.860 --> 00:49:09.720
Clemens Kroneberg: on which data is available. So when we think about big data such as Twitter

270
00:49:09.780 --> 00:49:15.860
Clemens Kroneberg: we have information on behavior of actors, for example.

271
00:49:15.980 --> 00:49:20.979
Clemens Kroneberg: the content of the tweets. whether they comment on a certain protest. For example.

272
00:49:21.310 --> 00:49:25.949
Clemens Kroneberg:  and we have data on how they are connected to the network task.

273
00:49:26.250 --> 00:49:31.160
Clemens Kroneberg: And the idea is if we have, like a matrix on the network. And we also know

274
00:49:31.490 --> 00:49:38.560
Clemens Kroneberg: across time, when people start with that, the certain adopt a certain behavior. And this allows us to model

275
00:49:38.870 --> 00:49:43.199
Clemens Kroneberg: the micro macro link. In a way, yeah, so it's it's rather

276
00:49:43.630 --> 00:49:52.620
Clemens Kroneberg: yeah. Object to this. So especially, we don't need any information on intentions. And that's basically the punchline of Peter's argument. Here. Let me just

277
00:49:52.810 --> 00:49:57.600
Clemens Kroneberg: quote from this piece on the past and Future of Analytical Sociology.

278
00:49:58.150 --> 00:50:04.580
Clemens Kroneberg: He wrote the importance many analytical sociologists attached to intention and explanations of behavior.

279
00:50:04.690 --> 00:50:13.920
Clemens Kroneberg: combined with the lack of information on the individual's intention. has led to a proliferation of just so stories. This sort of storytelling

280
00:50:14.380 --> 00:50:33.350
Clemens Kroneberg: contradicts one of the main tenants of analytical sociology. The explanations should refer to the causal process is known to be responsible for the outcome to be explained. and I suggest that we should avoid intention, explanations, and all but the rare cases where we have access to reliable information on the mental states of the acting individuals.

281
00:50:33.420 --> 00:50:44.510
Clemens Kroneberg: Instead, as I just said. we should concentrate our explanatory efforts on interaction structures and the macro patterns they give rise to. So that's basically the

282
00:50:44.750 --> 00:50:58.909
Clemens Kroneberg: the position that he advocates. And I read this as and you positioning. And you also is very self-effective about that. It responds to the rise of competition, social science, and the availability of big data

283
00:50:59.140 --> 00:51:02.120
Clemens Kroneberg: on behaviors and networks.

284
00:51:02.520 --> 00:51:05.439
Clemens Kroneberg: but not on individuals. Intentions, you know.

285
00:51:06.380 --> 00:51:11.519
Clemens Kroneberg: So what I would like to to ask in the remainder of my talk is.

286
00:51:12.730 --> 00:51:22.989
Clemens Kroneberg: if our objective is, we need to identify the actual cause of processes. As has always been the case, analytical sociology is this really good strategy?

287
00:51:23.230 --> 00:51:27.560
Clemens Kroneberg: So that's the avoidance of intentional explanations help us to reach our objective.

288
00:51:28.210 --> 00:51:35.120
Clemens Kroneberg: should we, in other words, exclusively concentrate our explanatory efforts on interaction structure, such as networks

289
00:51:35.320 --> 00:51:42.960
Clemens Kroneberg: and the resulting macro patterns. And I will illustrators using one poster child. Example of Analytical Sociology.

290
00:51:42.980 --> 00:51:44.439
Clemens Kroneberg: namely,

291
00:51:45.040 --> 00:51:52.999
Clemens Kroneberg: The famous paper by Beerman Moody. Install on chase of of On chase, a chains of effect.

292
00:51:53.640 --> 00:51:59.040
Clemens Kroneberg: and many of you will notice, so I can quickly summarize the main argument here. So the database

293
00:51:59.280 --> 00:52:04.139
Clemens Kroneberg: and is at health, and one school in the midwest. The us.

294
00:52:04.160 --> 00:52:10.210
Clemens Kroneberg: Well, we have data on the network of sexual and romantic relationships among respondents.

295
00:52:10.620 --> 00:52:15.149
Clemens Kroneberg: And what has been so remarkable is that and the networks shows this very unique.

296
00:52:15.310 --> 00:52:28.820
Clemens Kroneberg: spending 3 like structure that you see here visualized in this figure. And, as you can see from the circle, so that's just the largest component of the of the network.  what is really very rare.

297
00:52:28.840 --> 00:52:31.460
Clemens Kroneberg: and what leads to the circle like structure.

298
00:52:31.560 --> 00:52:35.780
Clemens Kroneberg: is a a configuration which students

299
00:52:36.320 --> 00:52:46.270
Clemens Kroneberg: date the former partner of an ex partner's new partner. But it sounds very very strange. So let me

300
00:52:46.300 --> 00:52:53.710
Clemens Kroneberg: to show you one example where this tri structure that this otherwise absent is present. Actually. So that's the this gray shaded area here.

301
00:52:53.990 --> 00:52:56.949
Clemens Kroneberg: So here we have 2 couples, this one

302
00:52:57.200 --> 00:53:14.499
Clemens Kroneberg: and that one on. Also, you know. Let's let takes the one. Take the look at the couples at sorry at time. Point T. One. So here we have 2 couples. Then they split up both of these couples, and then we have a date on new romantic relationship between those students.

303
00:53:14.580 --> 00:53:18.530
Clemens Kroneberg: and then also among the X partners. And this

304
00:53:19.590 --> 00:53:28.240
Clemens Kroneberg: final, you formation of a tie here completes this cycle of length 4. And this is really really the case.

305
00:53:28.410 --> 00:53:32.049
Clemens Kroneberg: Yeah. And and the argument made by human and others is that

306
00:53:32.260 --> 00:53:33.100
Clemens Kroneberg: and

307
00:53:33.480 --> 00:53:37.949
Clemens Kroneberg: that couples is the one forming in in at time 23

308
00:53:38.060 --> 00:53:42.820
Clemens Kroneberg: can be thought of as a second partnership. and therefore involves a public loss of

309
00:53:42.850 --> 00:53:49.289
Clemens Kroneberg: that was because, in a way, these 2 are the left over us from the previous covers, and when they now also form.

310
00:53:49.580 --> 00:53:57.410
Clemens Kroneberg: this is in a way, you know. yeah, not so good for their self confidence and for their public. standing in the school.

311
00:53:58.230 --> 00:54:09.349
Clemens Kroneberg: Okay? So this is basically, and the argument in this famous piece of research. It is also cited in the introductory section of the Oxford Handbook of Analytical Sociology, for example.

312
00:54:10.100 --> 00:54:16.000
Clemens Kroneberg: But recently, on 2,017. there has been a nice paper by so Stratford and others.

313
00:54:16.290 --> 00:54:20.779
Clemens Kroneberg: We show that if you employ a more advanced statistical modeling.

314
00:54:21.320 --> 00:54:30.890
Clemens Kroneberg: There's actually no support for this idea that such configurations are avoided by students. What seems to be rather behind this pattern is

315
00:54:31.030 --> 00:54:38.849
Clemens Kroneberg: that it's simply very unlikely that 2 individuals are simultaneous. These singles after the ex partner start dating.

316
00:54:39.010 --> 00:54:41.649
Clemens Kroneberg: Yeah. So this configuration is simply

317
00:54:41.740 --> 00:54:44.730
Clemens Kroneberg: where we are from a purely opportunity point of view.

318
00:54:45.090 --> 00:54:58.460
Clemens Kroneberg: So the whole idea of a social norm governing these choices might have been wrong. But of course, maybe also the statistical model by structured, and Co. With us is wrong, and Pm. And others were right about the existence of the social norm.

319
00:54:58.900 --> 00:55:07.110
Clemens Kroneberg: The point is without data on intentions. We will never be able to tell whether that's such a norm existed in this case or not.

320
00:55:08.740 --> 00:55:17.350
Clemens Kroneberg: So this shows this example. I think that it's really challenging to identify the true cause of processes behind micro level patterns.

321
00:55:17.450 --> 00:55:20.739
Clemens Kroneberg: And we can't say that it's always good to avoid

322
00:55:20.900 --> 00:55:27.590
Clemens Kroneberg: intentional explanation, right? Because we will never know in advance whether they met and how much they met.

323
00:55:28.840 --> 00:55:36.680
Clemens Kroneberg: So the interest in the interest of time. And let me just make it this quite a little bit more provocative here for our discussion.

324
00:55:36.740 --> 00:55:59.209
Clemens Kroneberg: So I think it's really difficult to diagnose complex phenomena based on just such such objective data. Right. And if you think about other sciences such as medicine. it's so, it's so multi-fessed that how they approach a diagnosing a complex phenomena, such as whether into what the green individual has a certain disease, right? There's so many different kinds of tests.

325
00:55:59.450 --> 00:56:01.760
Clemens Kroneberg: And I think that we also have lots of

326
00:56:02.010 --> 00:56:20.019
Clemens Kroneberg: and ways to test the existence of a social norm. So rather than being so, pacimonious. And we should really also use survey data, use qualitative interviews, and so on use some joint experiments to really nail down the existence of a social norm in a certain setting. So I think investing in measurement is really important even

327
00:56:20.060 --> 00:56:22.399
Clemens Kroneberg: in the age of computational social science.

328
00:56:22.460 --> 00:56:29.380
Clemens Kroneberg: and also complementing such a Dss studies with qualitative work as Mario Small and others

329
00:56:29.510 --> 00:56:31.590
Clemens Kroneberg: also have recommended.

330
00:56:32.420 --> 00:56:36.089
Clemens Kroneberg: So This will, in a way have a 3 values.

331
00:56:36.130 --> 00:56:48.210
Clemens Kroneberg: So for our advantages. So first of all. just what make us better equipped to identify the actual caller processes at work. And, secondly, we can also benefit from analytical advances that are underway

332
00:56:48.290 --> 00:57:02.509
Clemens Kroneberg: in other neighboring science, that study human behavior, and also the subjective sites, such as virtual reality methods and and and other tools. And finally, it's also, of course, helpful to engage with these neighboring side sciences rather than

333
00:57:02.570 --> 00:57:14.749
Clemens Kroneberg: focusing too narrowly on a very specific understanding. often legal sociology, that it could lead to an isolation from such neighboring approaches. So yeah, but I'm very.

334
00:57:14.920 --> 00:57:18.690
Clemens Kroneberg: You get to hear your thoughts on this method methods. Thank you.

335
00:57:24.750 --> 00:57:35.510
Gianluca Manzo: Thank you very much, Clemens, for for your talk and this invitation to make the logical bureaucracy and and and data pluralism.

336
00:57:35.720 --> 00:57:40.249
Gianluca Manzo: so any questions for Clements,

337
00:57:40.990 --> 00:57:43.560
Gianluca Manzo: from the room or

338
00:57:48.490 --> 00:57:51.030
Gianluca Manzo: I. Yes, no question.

339
00:57:52.040 --> 00:58:00.279
Gianluca Manzo: So I I I have 1 one general remark, if if I can, if I can. It's just a remark. Clements.

340
00:58:00.800 --> 00:58:04.669
Gianluca Manzo:  to some extent.

341
00:58:05.960 --> 00:58:16.860
Gianluca Manzo:  The Peter's statement. about not to invest too much time in a

342
00:58:17.200 --> 00:58:25.979
Gianluca Manzo: focus on on intentions. Seems to some extent the symmetric positions that

343
00:58:26.520 --> 00:58:32.139
Gianluca Manzo: orders

344
00:58:32.160 --> 00:58:47.160
Gianluca Manzo: part in the past, and I'm I'm thinking about scholars that you know, in the field of social certification. Research. who for a long time Well, they claimed exactly to pose it.

345
00:58:47.420 --> 00:59:13.659
Gianluca Manzo: meaning that. Well, we, we have data on preferences and cost perception. For instance, in the field of education, educational choices. But we don't have data on interactions, structures. And so to at least the implicit idea was, don't look into interactions and networks structures. If you want to understand the link between micron and macro.

346
00:59:13.760 --> 00:59:27.470
Gianluca Manzo: levels of analysis in the field of social, mobile, or educational preference. And so you you know, lots of of of autos. even even in in German. So they they well.

347
00:59:27.520 --> 00:59:43.769
Gianluca Manzo: they tend to study educational charges just by looking at preference and and cost and now we have litter by little partly because of analytical solutions. a a move

348
00:59:44.290 --> 01:00:02.160
Gianluca Manzo: to more focus on social interact. So it's just a remark. But I I well, we we, we, we we we are only in in the middle of a of a dynamic process in which we have to find the good balance between

349
01:00:02.410 --> 01:00:05.120
Gianluca Manzo:  between

350
01:00:05.390 --> 01:00:19.180
Gianluca Manzo: data procedure collection to to combine the best data on intentions and the base data on interaction in terms. It's not to defend Peter's statement just to to notice that

351
01:00:19.600 --> 01:00:34.800
Gianluca Manzo: we we had symmetric claims in order part of sociology unrelated with analytical societies. So I don't know if if you, if you have a an opinion on on this of if if you think that my

352
01:00:34.820 --> 01:00:37.399
Gianluca Manzo: my observation is wrong.

353
01:00:38.230 --> 01:00:57.699
Clemens Kroneberg: well, I think that's very valuable and interesting observation, and I I share your view. I think all our fields have certain inertia right? And it's important from time to time to approach a topic from a different angle, using different methods, focusing on different kinds of phenomena. And I agree that restaurant choice sociology in Europe that focused

354
01:00:57.740 --> 01:01:06.900
Clemens Kroneberg: on on educational decision making as for a long time neglected the relevance of networks. And and your work has, of course, also be very important to change that.

355
01:01:07.110 --> 01:01:09.239
Clemens Kroneberg: And and and I'm

356
01:01:09.420 --> 01:01:24.599
Clemens Kroneberg: and this over emphasis on action to you is something that I sympathize with. I can understand Peter's point there. But I just think that we shouldn't go too far. Yeah, and and throw out action, theory, or data on intentions all together. Because I think what makes

357
01:01:24.610 --> 01:01:29.089
Clemens Kroneberg: analytical sociology credible in a way, is that we

358
01:01:29.450 --> 01:01:32.290
Clemens Kroneberg: an approach within our

359
01:01:32.380 --> 01:01:35.960
Clemens Kroneberg: not so pluralistic community, one should say, compared to the

360
01:01:36.040 --> 01:01:40.020
Clemens Kroneberg: all discipline, and we can still approach.

361
01:01:40.330 --> 01:01:47.560
Clemens Kroneberg: and our phenomena rather comprehensively, using different methods. And I think this is really a strength that we shouldn't let go of.

362
01:01:52.030 --> 01:01:55.630
Clemens Kroneberg: Thank you. Thank you. Clements

363
01:01:56.330 --> 01:01:57.740
Gianluca Manzo: question remarks.

364
01:01:58.760 --> 01:02:07.620
Gianluca Manzo: Oh, okay, so we can. We are I? Yes, there is a there is a question from you, I think if I if I see correctly in the room.

365
01:02:13.660 --> 01:02:15.110
that's how I got there.

366
01:02:16.690 --> 01:02:25.689
Okay. Heckle, Ms. And a Hajian Luca, thanks for a presentation. Have a very simple question, are you? I wonder whether

367
01:02:25.830 --> 01:02:54.610
sociology is expected to use fiore? Because sometimes the government as as asks me to conduct service on, say, poverty or marriage market. And well, and then I I do it. And I want to analyze them about using, say Russian to theory, but they don't want it, or they they only want distributions at most. media, for instance, by, say, gender or educational attainment.

368
01:02:54.610 --> 01:03:07.720
So my question is very simple, if it is sociologists to use theory. But how do you think?

369
01:03:09.140 --> 01:03:29.350
Clemens Kroneberg: Yeah, thank you very much. That's very important. Question. I try to be brief, not to take way too much time from from the next talk. So I totally share this, and I think also, descriptive work has been undervalued in sociology. So doing. Really rigourous description is really challenging. And it's really important right? You don't have to have a new theory all the time. But still I would think that

370
01:03:29.350 --> 01:03:47.829
Clemens Kroneberg: theory is very important, just as Emily said before, no, we have to have an idea of whether, when we identify, for example, unintended consequences. But the knowledge about unintended consequences will change Actus behavior, and so on. So answering such questions really requires theoretical thinking, as I, I would say.

371
01:03:48.060 --> 01:03:49.080
Clemens Kroneberg: thank you.

372
01:03:53.540 --> 01:04:07.659
Gianluca Manzo: thank you, Clements, and thank you you for for the question. So I think we? We have now to move on to the next talk, and then we can continue the discussion. at the end. so it's

373
01:04:07.880 --> 01:04:09.450
Gianluca Manzo: I think.

374
01:04:09.520 --> 01:04:17.419
Gianluca Manzo: yes, Andreas is is your your turn. So please go ahead for 15 min.

375
01:04:17.460 --> 01:04:19.400
Gianluca Manzo: and you you know the rule.

376
01:04:19.450 --> 01:04:21.270
Andreas Flache: You can see my screen.

377
01:04:22.130 --> 01:04:24.420
Gianluca Manzo: I can. Okay. Good

378
01:04:24.620 --> 01:04:35.969
Andreas Flache: Well, then, thanks a lot for the possibility to present you some of my work and thoughts on the credible sociological model building 100 and

379
01:04:35.980 --> 01:04:41.950
Andreas Flache: yeah, let me jump into it right away. So I wanna wait.

380
01:04:42.260 --> 01:04:43.330
Andreas Flache: why do we?

381
01:04:43.660 --> 01:05:04.560
Andreas Flache: I I want to start from. Let's say what, in my view, is a very important reason why we need model building in sociology. And we have discussed the we the previous speakers have discussed this also already, that typically, if we look at social phenomena, they are complex phenomena in the sense of

382
01:05:05.100 --> 01:05:18.419
Andreas Flache: erez agmoni being generated being brought about by interactions between individuals between individuals and different other entities at different levels in society micro interactions. And it's very hard to 150

383
01:05:18.430 --> 01:05:21.629
Andreas Flache: understand and to figure out how exactly

384
01:05:21.850 --> 01:05:38.779
Andreas Flache: a certain phenomenon has arisen, and in particular, whether this is due to the intentions or anticipations of individuals who acted, or maybe not, maybe unintended, unanticipated, and also understood by us as which the researchers 150.

385
01:05:38.820 --> 01:05:47.030
Andreas Flache: So from that perspective, credible sociological explanation at least requires to show.

386
01:05:47.130 --> 01:05:56.680
Andreas Flache: And now I'm referring, of course, to the wording here in the the I'm quote how the parts of a social system actually interact

387
01:05:56.900 --> 01:06:08.840
Andreas Flache: how this jointly brings about the whole the social phenomenon at the collective level I'm interested in, like, say, segregation polarization, inequality. And more

388
01:06:09.090 --> 01:06:19.219
Andreas Flache:  I need to explicate what exactly I actually these parts. And also, how exactly do they interact?

389
01:06:19.390 --> 01:06:27.949
Andreas Flache: And I believe a very, a powerful tool for that. And that is, of course, also stressed in much of analytical sociology.

390
01:06:28.240 --> 01:06:36.060
Andreas Flache: is computational modeling, and in particular, agent-based computation and modeling, because that also focuses on

391
01:06:36.070 --> 01:06:52.749
Andreas Flache: the individual actors and their interaction. And then the aim. The goal is that we want to develop a generative theory of a social phenomenon. So a computational theory that really allows us to see how this phenomenon comes about.

392
01:06:53.440 --> 01:07:11.590
Andreas Flache: But there, the story doesn't end, because it is indeed one of many possible ways, probably to generate the outcome. So we need or we can use this tool, then, to conduct systematic computational experiments also to give us a better understanding

393
01:07:11.760 --> 01:07:20.360
Andreas Flache: which elements in our explanation are critical, which could be changed, which could be relaxed? Under what conditions would different

394
01:07:20.540 --> 01:07:37.380
Andreas Flache: erez agmoni generative theories lead to the same result, or to different results, which in the end, of course, provides us with testable, empirically testable claims, also that may help us to differentiate and distinguish between different explanations. 250.

395
01:07:37.670 --> 01:07:46.479
Andreas Flache: Now, when do we actually need a really computational modeling? And I think we we need it the most.

396
01:07:46.710 --> 01:07:56.740
Andreas Flache:  If we have a theory for which it's not really very obvious. What exactly are the implications? Because it's so complex?

397
01:07:57.000 --> 01:07:59.550
Andreas Flache: And then we need to have a tool

398
01:07:59.840 --> 01:08:22.349
Andreas Flache: with which we in silico can figure out really exactly under which conditions, what? Which implication follow from the theory. And of course. it's most needed. If other tools do not suffice, and very often verbal deduction will not suffice, and very often purely mathematical. analytical analytical tools will not be available.

399
01:08:23.020 --> 01:08:32.340
Andreas Flache:  then. But then an important discussion. And we have indeed heard about that also before. Here in various talks, is

400
01:08:32.380 --> 01:08:38.240
Andreas Flache: well, how empirically realistic. should these computational models be.

401
01:08:38.590 --> 01:08:44.259
Andreas Flache: And in analytical sociology there's a very clearly very clear stance taken

402
01:08:44.310 --> 01:08:59.650
Andreas Flache: amongst other people, also by John Luca right and had some, of course. And here also in the paper by Koch, Nick and Hadstrom. that we you should use in computational models

403
01:09:00.040 --> 01:09:29.730
Andreas Flache: explanations and which are closely linked to as much as possible of the empirical knowledge that we have about the actual process. So the actual actors, the actual intentions, the actual way of decision making, and this should never be overridden by the aim or, or, let's say, considerations of elegance, simplicity, so and so on, should never override. This aim of an accurate, empirically accurate

404
01:09:29.730 --> 01:09:34.720
Andreas Flache: description of the mechanism that actually produced the phenomena we want to explain.

405
01:09:35.090 --> 01:09:47.620
Andreas Flache: This, of course, is a position that in principle I fully fully would subscribe to. If we have the choice between different explanations, then always the empirically more realistic one is the better one.

406
01:09:48.190 --> 01:09:56.049
Andreas Flache: But I think there's also a snack, and I think it's important to be aware of that, and and to think about what we can do about it.

407
01:09:56.080 --> 01:10:12.819
Andreas Flache: And the snag is that if we make models more complex and more realistic than typically, that also means we make them more complex, harder to understand. And sometimes we can then, not figure out very well.

408
01:10:12.820 --> 01:10:32.840
Andreas Flache: whether adding, this piece of realism has actually really improved our model, or maybe has led us on the wrong track. And I want to illustrate this with what an example, of course, All of us know very well the famous residential segregation model independently proposed by shelling and sakota

409
01:10:32.990 --> 01:11:01.479
Andreas Flache: so the core idea here is to show how segregation can be the unintended result of the interactions of many autonomous agents, many individuals who actually have only very weak. If you want homophilias preferences, they just don't want to be in the minority, locally in their neighborhoods, and as a result of their actions and their decisions, they end up being very highly segregated.

410
01:11:02.470 --> 01:11:13.999
Andreas Flache: The story, as you all know, goes from let's say, the core assumption people have these mild ethnic preferences not want to be in the

411
01:11:14.010 --> 01:11:26.959
Andreas Flache: local minority. but then you have, say, random local concentrations in an initial situation. Some individuals are dissatisfied, leave their neighborhoods there by the neighborhoods, become more homogeneous

412
01:11:26.970 --> 01:11:36.200
Andreas Flache: and leading leading even more members in the local minority to leave. And in the end we have a very segregated world. There's an important lesson here.

413
01:11:36.400 --> 01:11:54.670
Andreas Flache: It could be that segregation comes about even without having to assume that, for example, landlords discriminate housing prices are the course. But of course all these other factors may play a role, too, and we need to consider them then in more complex and more realistic models.

414
01:11:54.820 --> 01:12:03.010
Andreas Flache: But for the moment, let's just look at one of the let's say. very intriguing implications of this perspective.

415
01:12:03.020 --> 01:12:16.930
Andreas Flache: So there's a very, I would say, celebrated result. if, for example, we take a situation where individuals look at only a very small neighborhood in their local environment, 250

416
01:12:17.120 --> 01:12:23.680
Andreas Flache: and are satisfied with, say, a 20% individuals of their own type.

417
01:12:23.690 --> 01:12:43.329
Andreas Flache: So that would mean one out of 4. We start from this initial random situation. Here, then, with this one out of 4, we end up with very mild segregation. Say everyone has now 60 similar neighbors. In the beginning they had 50, and if we increase this individual level intolerance only slightly.

418
01:12:43.330 --> 01:12:57.540
Andreas Flache: we have a big jump and a big increase in the level of segregation, and this is an impressive and intriguing example of a face transition in the social system warning us against too rarely inferring from

419
01:12:57.740 --> 01:13:07.219
Andreas Flache: little things change on the micro level. Well, that little change will there be on the micro level? This might be wrong. So this could be a very important insight.

420
01:13:07.420 --> 01:13:11.570
Andreas Flache: But of course it's important to go on and to make your model more realistic.

421
01:13:11.900 --> 01:13:14.650
Andreas Flache: And we have spelled through this in the

422
01:13:14.700 --> 01:13:23.219
Andreas Flache: paper. I will tell a little bit more about that later on. And this is just the very first and very simple step of making a model a bit more realistic

423
01:13:23.300 --> 01:13:35.590
Andreas Flache: erez agmoni So let's move to, let's say, a larger number of groups. Let's mimic a typical big Us. City with, let's say, 4 major ethnic or racial groups. If you want 150

424
01:13:35.730 --> 01:13:51.150
Andreas Flache: different sizes, one majority group a bigger minority to smaller minorities. And let's also assume that people do not only look around them in these very small neighborhoods, but take a bit of a wider look at the wider neighborhood.

425
01:13:51.360 --> 01:13:56.569
Andreas Flache: and now we can again see and just replicate the same analysis we did before.

426
01:13:57.080 --> 01:13:59.359
Andreas Flache: But now the

427
01:14:00.030 --> 01:14:04.860
Andreas Flache: impressive face transition result has just vanished and disappeared.

428
01:14:05.540 --> 01:14:15.999
Andreas Flache: So now, if we increase the level of intolerance by one percentage point, segregation also increases by one percentage point so well, something has changed here.

429
01:14:16.070 --> 01:14:25.280
Andreas Flache: Now we can say, Okay, maybe our initial theory was just wrong, and we shouldn't expect such sudden face transitions empirically.

430
01:14:25.470 --> 01:14:29.939
Andreas Flache: But I think the more important thing to do now at this point would be.

431
01:14:30.440 --> 01:14:38.939
Andreas Flache: try to understand what happened. Look at your model, at the system. You want to understand the dissect your model?

432
01:14:39.080 --> 01:14:55.319
Andreas Flache: so you need to. yeah, not only grow the phenomenon, but also really, understand how you grew the phenomenon and what this means for our our theory. So let me go through this. For this simple example, 150,

433
01:14:55.620 --> 01:15:11.569
Andreas Flache: I would say, what we now need is to develop a sort of tentative explanation of the behavior of our model, because we want, of course, in the end to use the model as a tentative explanation of social reality. But then we better understand our model first.

434
01:15:11.820 --> 01:15:22.589
Andreas Flache: erez agmoni So one thing we can do here now is to to propose. It's the radius that matters, because if we increase the size of this neighborhood radius, every 2

435
01:15:22.650 --> 01:15:28.620
Andreas Flache: local neighborhood resembles more a representative sample of the initial population.

436
01:15:28.830 --> 01:15:33.890
Andreas Flache: which means, if we make agents now increasingly

437
01:15:34.200 --> 01:15:46.750
Andreas Flache: erez agmoni intolerant to out group members in their local neighborhood, we will see also a more smooth change of the proportion of agents who become unhappy one.

438
01:15:46.840 --> 01:15:57.380
Andreas Flache: this, the need to be needs to be combined with the second element, say, in our tentative theory, which basically would be the idea that the more agents are unhappy initially.

439
01:15:57.470 --> 01:16:03.210
Andreas Flache: the more segregation we will get in the end. because these moving cascades become bigger.

440
01:16:03.260 --> 01:16:11.900
Andreas Flache: and that connects the radius to the degree of segregation and the smoothness of the increase in the degree of segregation.

441
01:16:12.150 --> 01:16:16.560
Andreas Flache: if we complete this tentative explanation

442
01:16:17.270 --> 01:16:25.580
Andreas Flache: mit ctl, and the cool thing then, about age and best modeling is that we can use our model to test our tentative explanation. And that is what we did here. 150

443
01:16:25.650 --> 01:16:49.380
Andreas Flache: So we have a radius, one radio, 3 version of the model. I otherwise exactly the same simulation experiment with 4 groups, unequal sizes, and so on. And here you see all the proportion of initially unhappy agents, indeed, smoothly increases. Now, if we increase the radius so so well understood. How? Why, that is the case. Right? If we have

444
01:16:49.640 --> 01:17:03.540
Andreas Flache: a very small neighborhood, then the crucial thing is, do you need one? Or do you need to neighbors to be a satisfied in group names to be satisfied, and that makes for these jumps here

445
01:17:04.040 --> 01:17:15.849
Andreas Flache: erez agmoni. But if we have a larger neighborhood, it becomes a more smoother, a change. And now we can see, does this actually translate in a smoother change also in the level of segregation, and well in the way it does 150,

446
01:17:16.110 --> 01:17:21.210
Andreas Flache: but not quite right. So there are the things that are not quite in line with our

447
01:17:21.610 --> 01:17:31.240
Andreas Flache: tentative theory of our model. and I want to focus now on this little bump here as the next cycle in the going through possible explanations.

448
01:17:31.490 --> 01:17:34.360
Andreas Flache: try to understand what goes on here

449
01:17:34.520 --> 01:17:37.160
Andreas Flache: again. There's a tentative explanation.

450
01:17:37.370 --> 01:17:56.650
Andreas Flache: I don't want to show you now this whole or go with you through this whole wording. But just show you how you get this intuition if you run the model. So let's start here from the 4 group situation. Now we have very mild as a preferences, only 20 in group neighbors.

451
01:17:56.770 --> 01:18:24.970
Andreas Flache: We run the model, and we see that these small minorities here have a really hard time, or they they cluster so the yellow and the blue ones they cluster in very lose clusters, which is good enough for them to find the 20 in group neighbors. That gives us a certain degree of clusting and segregation in the population. But if we make them now more demanding, as I do now. So I go to 35,

452
01:18:25.210 --> 01:18:36.440
Andreas Flache: and I do the same thing again. Then these very small minorities don't actually find enough space to settle in. because now they need denser clusters, more in-group neighbors.

453
01:18:36.590 --> 01:18:52.560
Andreas Flache: but they are so small that they are sort of don't get enough room by the larger minorities who actually do succeed in clustering to some extent, as you see here. So they walk around randomly, which means as an implication.

454
01:18:52.880 --> 01:19:05.820
Andreas Flache: there is less segregation now because they fail to cluster. and if we now make everyone more demanding again. then also the larger minorities start to contract and form clusters.

455
01:19:05.990 --> 01:19:15.390
Andreas Flache: and that gives enough room for the smaller minorities to actually also build their own clusters. And now segregation goes up again.

456
01:19:17.160 --> 01:19:30.590
Andreas Flache: So, not boring you with the slide with many words. I can just show you a simulation experiment, then testing this new tentative explanation. And that is, we just reduce the density of

457
01:19:30.910 --> 01:19:55.640
Andreas Flache: the world. Now, it's not 90% populated, but 70% populated. And now this is what we get for the clustering of the smallest minority in the 90% populated world if we increase the tolerance threshold. But this is what we get in the 70% populated world. So you could said, say, second tentative theory So

458
01:19:55.830 --> 01:20:02.989
Andreas Flache: of our model has been supported by this test. So we seem to have understood our model now even better.

459
01:20:04.190 --> 01:20:13.220
Andreas Flache: And of course we can now go on and focus on more UN understood phenomena and improve our understanding of the model, inject more realism

460
01:20:13.280 --> 01:20:30.280
Andreas Flache: like, for example, adding it to discrete choice models, reflecting more heterogeneous preferences. If you are interested, you can read more about this in this paper to which I refer to here, which, by the way is an open access paper, so you can just download and read it if you like.

461
01:20:30.690 --> 01:20:47.310
Andreas Flache: okay. So what our general lessons one could draw from this I proposed a few here, so I believe, the tentative best practice in trying to explain social phenomena with computational models in analytical sociology, 150

462
01:20:47.370 --> 01:20:58.190
Andreas Flache: approach is, don't inject realism too early, and what is too early before you really have understood what it does it to your model.

463
01:20:58.390 --> 01:21:00.789
Andreas Flache: And sometimes, if you don't understand

464
01:21:00.860 --> 01:21:16.100
Andreas Flache: what happens, you need to backtrack. You need to make the model more simple again and less realistic again, and then develop a solid understanding of why something changed in the model and tested again until you understand it better.

465
01:21:16.500 --> 01:21:29.170
Andreas Flache: The good thing is, the Abcm. Is a tool that helps us to experiment systematically and thereby. because we have it fully under control, unlike social reality.

466
01:21:29.170 --> 01:21:52.570
Andreas Flache: to develop a very solid understanding. And I want to end with this a statement here from our paper, which partly has been inspired by John Luca. so a realistic Ibm, implementing a theory is a complex system in itself, and it first needs to be understood well before we can use it for understanding the social reality. It is supposed to model.

467
01:21:52.970 --> 01:21:54.939
Andreas Flache: Well, thank you for your attention.

468
01:21:55.130 --> 01:22:00.950
Andreas Flache: I want to give some credits here to people having been involved in this and related research.

469
01:22:01.090 --> 01:22:03.899
Andreas Flache: and I'm looking forward to your comments and questions.

470
01:22:17.000 --> 01:22:20.819
Gianluca Manzo: Alright, thank you very much. And for this very inspiring

471
01:22:20.880 --> 01:22:32.800
Gianluca Manzo: talk I have lots of question myself. But I I I will not raise questions. Now I see Clements. you you you want to ask question Clements. Right?

472
01:22:34.040 --> 01:22:56.369
Clemens Kroneberg: Yeah, thank you for me. So yeah, thank you. Andreas great to see you again after such some time. And it's a great, great piece of work, I think, also very valuable in in, you know, showing our very advanced students right to go beyond teaching the shedding model over and over again. And so I want that. So I was very convinced by the idea how to understand

473
01:22:56.570 --> 01:23:01.130
Clemens Kroneberg: and do this as a methodology, also an increasingly realistic

474
01:23:01.150 --> 01:23:07.760
Clemens Kroneberg: agent based model. And of course, and one always one does how this thing can be translated to explain social reality.

475
01:23:07.790 --> 01:23:19.140
Clemens Kroneberg: We know that Elizabeth Paul, for example, has done work with a rock mayor on on doing this with observational data. But you know, I was rather thinking now to really have to tight coupling that I find so important.

476
01:23:19.160 --> 01:23:27.719
Clemens Kroneberg: maybe it's best to do the kind of online experiments that Damon, Santola and and Mrs. Organic and others have done right, because in such an environment

477
01:23:27.790 --> 01:23:34.540
Clemens Kroneberg: you could really try to experiment, have the same degree of control that you have in the model, because

478
01:23:35.020 --> 01:23:43.859
Clemens Kroneberg: reality as such, is still so messy and so full of institutional richness that it's really difficult for me to see how such specific implication then could be tested.

479
01:23:44.000 --> 01:23:46.010
Clemens Kroneberg: I'm just curious about your thoughts.

480
01:23:46.200 --> 01:24:02.289
Andreas Flache: Yeah, thank you, Clements, for this comment, and indeed good to see you, too. Mit Ctl, and yeah, I I this gives me an opportunity to just link also to things you said. I really believe, using a broader toolbox of methodologies in combination

481
01:24:02.390 --> 01:24:31.789
Andreas Flache: with computational modeling is what we what we need to do So in in some of our work. And this links also to examine the exams. You just discuss what we, what we try to do is, for example, get a bit of understanding, but on the thing of microprocesses by doing left experiments. So how do people, for example, influence each other when it comes to being exposed to other opinions of the views on on certain issues doesn't how does this change their opinions? And what does it depend, and then

482
01:24:31.790 --> 01:24:38.730
Andreas Flache: use that knowledge that we acquired from the lab experiment and feed it into computational models, and then

483
01:24:38.990 --> 01:25:00.289
Andreas Flache: use the models to come up with testable propositions about what would happen at the micro level? If so, for example, we would change the network structure, change the group size whatever, and then test that again. in a lab experiment. And then if we develop by that better understanding of how micro outcomes may be changed.

484
01:25:00.420 --> 01:25:05.320
Andreas Flache: by changing micro conditions through this mechanism.

485
01:25:05.490 --> 01:25:16.840
Andreas Flache: I think it's time is ripe to look around outside of the lap. like in online social media. whether we find their places where we can test or

486
01:25:17.140 --> 01:25:18.860
Andreas Flache: that, the theories then

487
01:25:22.630 --> 01:25:24.690
Andreas Flache: hope that answers you the question.

488
01:25:25.000 --> 01:25:26.529
Clemens Kroneberg: Yeah, thank you. Yeah.

489
01:25:26.900 --> 01:25:31.510
Gianluca Manzo: Thank contrace. Is there any question in the room? Masa?

490
01:25:33.840 --> 01:25:37.130
you have any question.

491
01:25:42.750 --> 01:25:45.070
Gianluca Manzo: Okay?

492
01:25:45.080 --> 01:25:48.340
Gianluca Manzo: Okay? Sure. So

493
01:25:48.690 --> 01:25:51.540
Gianluca Manzo: well, I would love to

494
01:25:51.580 --> 01:26:10.819
Gianluca Manzo: to continue the discussion. At the same time I feel a little, but not to show you the the video record by the Daniel Liter, because he he he spent time to record the. So I I suggest that we well, I share with you the Daniel Video, which is about 40 min, and then we will have

495
01:26:11.080 --> 01:26:13.539
Gianluca Manzo: 7, 8 min to continue discussion.

496
01:26:13.660 --> 01:26:16.259
Gianluca Manzo: Would you agree with the with with this?

497
01:26:17.480 --> 01:26:22.029
Gianluca Manzo: Okay, okay. So let me let me see.

498
01:26:22.060 --> 01:26:27.740
Gianluca Manzo: if I can share my screen. Yes.

499
01:26:28.150 --> 01:26:33.619
Gianluca Manzo:  So, did you see my screen?

500
01:26:34.790 --> 01:26:35.570
Yes.

501
01:26:35.710 --> 01:26:40.400
Gianluca Manzo: yeah. Okay, so. No, I I I I start

502
01:26:45.930 --> 01:26:47.260
Gianluca Manzo: let me see.

503
01:27:04.950 --> 01:27:05.630
and so

504
01:27:10.030 --> 01:27:12.000
we can't hear the sound.

505
01:27:12.810 --> 01:27:16.250
So General Kar, do, do you? A mute?

506
01:27:19.620 --> 01:27:22.470
You should mute to yourself?

507
01:27:26.650 --> 01:27:29.490
Gianluca Manzo: I I'm sorry. Mezza. There's a problem.

508
01:27:30.240 --> 01:27:33.520
Yes. but we cannot hear, boys.

509
01:27:35.580 --> 01:27:39.360
Gianluca Manzo: Oh, probably it's my! It's me. It's me. I, I mute.

510
01:27:40.110 --> 01:27:40.840
Yeah.

511
01:27:47.970 --> 01:27:48.720
to me.

512
01:28:02.810 --> 01:28:04.850
Gianluca Manzo: is it not good enough, mother?

513
01:28:05.300 --> 01:28:16.999
If the option your your source you can choose the is currently that follows us should be a microphone, and you can just your in terms of

514
01:28:20.600 --> 01:28:23.670
Gianluca Manzo: I'm sorry I'm I'm not sure I

515
01:28:26.040 --> 01:28:29.440
Gianluca Manzo: I understand what you suggest.

516
01:28:32.030 --> 01:28:37.309
Gianluca Manzo: I don't know if the problem comes from my laptop or from from the room.

517
01:28:38.540 --> 01:28:44.580
Lucas sage: Yeah, I think it's it's in the sharing options you have to share. Also the the sound

518
01:28:45.560 --> 01:28:48.050
Lucas sage:  on your laptop.

519
01:28:50.040 --> 01:28:52.620
Gianluca Manzo: You have to share this

520
01:28:54.820 --> 01:28:59.930
Gianluca Manzo:  from where

521
01:29:01.950 --> 01:29:03.110
Gianluca Manzo: from the

522
01:29:03.550 --> 01:29:06.940
Gianluca Manzo:  here.

523
01:29:07.790 --> 01:29:13.199
Gianluca Manzo: I think it might be under more so down right. This one you think.

524
01:29:19.610 --> 01:29:21.909
Gianluca Manzo: Now I share so.

525
01:29:24.390 --> 01:29:25.670
Gianluca Manzo: and

526
01:29:26.630 --> 01:29:37.279
Lucas sage: any good sociological account must provide. Yes, it is based on a description of the mechanisms that give rise to those outcomes.

527
01:29:38.300 --> 01:29:50.689
Second, the mechanisms need to be identified at the level of the actors who make up those outcomes. Third, the researcher should always strive for realism in assumptions about

528
01:29:51.050 --> 01:30:05.859
mechanisms, hypotheses about social processes, and so forth. Fourth, Analytical sociology should be pluralistic about the theories of the actor which it makes use of not be constrained simply to

529
01:30:06.150 --> 01:30:23.510
rational choice models or or other narrow economic rationality. Models of actor decision making in fifth explanations should always take the form of an ascending explanation from individuals to social outcomes.

530
01:30:24.680 --> 01:30:51.249
So the test case that I have in mind is contentious politics, and this topic might seem to be an ideal subject for treatment by the methods of analytical sociology, because it is superficially plausible that we could account for a great deal of the phenomena of riots, rebellion, civil wars, by building upwards from the grievances that exist among actors, the political incentives of political actors.

531
01:30:51.250 --> 01:31:09.700
and then the processes of aggregation through which these motivations lead to collective action. Doug Mcadam treats the Us. Civil rights movement in his book Political Process and the development of Black Insurgency and his model of explanation does not take that structure.

532
01:31:10.630 --> 01:31:26.460
Macadam notes the significant increase in civil rights activism in the 1950 s. American South and its rapid spread to new locations. Jim Crow, segregation and repression were suddenly the object of organized resistance across the South.

533
01:31:26.530 --> 01:31:42.310
This was a complex and decentralized process. with multiple actors and local populations that were involved. There were the black churches, the civil rights organizations, multiple influential leaders in many different places.

534
01:31:42.460 --> 01:31:56.409
There were local activist groups in different cities, organizing around specific issues. And of course there were millions of ordinary black people who were expressing their grievances and joining in demonstrations

535
01:31:56.820 --> 01:32:07.290
the 3 large causal factors which Doug Mcadam emphasizes and kind of organizes his analysis around include these first

536
01:32:07.320 --> 01:32:17.650
successful, contentious action. Collective action is most likely in a time of expanding political opportunities.

537
01:32:17.800 --> 01:32:28.599
Second. in order for contentious politics to succeed, there need to be existing mobilizing structures, including organizations.

538
01:32:28.830 --> 01:32:43.589
And, third, turning to the actor, there is a recognition on Doug Mcadam's part that actors are not static, fixed once and for all individuals. Instead, they are meaning bearing agents.

539
01:32:43.730 --> 01:33:03.719
topics, and issues and grievances need to be framed. They need to arrive at their own meanings, to attribute to the circumstances. They find themselves in so expanding political opportunities, existing, mobilizing structures and the processes of framing and meaning creation. These are the 3 main ideas.

540
01:33:04.570 --> 01:33:15.990
So the question for me is whether Mcadam's method of analysis of the Civil Rights movement can be reconciled with the analytical sociology manifesto. and

541
01:33:16.540 --> 01:33:28.219
I'm inclined to what is for me anyway, kind of a surprising conclusion. It is the possibility that almost all of Mcadam's key theoretical statements, and his analytical statements

542
01:33:28.370 --> 01:33:40.710
and the methodology that he pursues can be reformulated in analytical sociology terms. So we might think macadam was perhaps an analytical sociologist before his time.

543
01:33:41.110 --> 01:33:52.590
and it turns out that that is almost true, but with important provisos which are of interest to analytical sociologists. So first, the points of compatibility.

544
01:33:52.800 --> 01:34:17.800
When you read through Mcadam's book, you're really struck by how clear his exposition is. He has a knack for stating both what was happening on the ground and his own interpretation of the causal factors leading to those those results. Secondly, he makes extensive use of available evidence, both macro-evident statistical evidence, quantitative and qualitative evidence.

545
01:34:18.180 --> 01:34:33.420
A third, he is realistic about the assumptions and descriptions that he puts forward of the social processes involved. As a historical sociologist, he is committed to the idea of providing realistic descriptions.

546
01:34:34.060 --> 01:34:51.860
Fourth, he is committed to providing an account of mechanisms leading to the outcomes of interest. This is a shared commitment with other historical sociologists like Cho Tilly, but their view of explanation is entirely focused on the idea of providing accounts of the mechanisms and processes.

547
01:34:52.220 --> 01:35:06.269
and finally, there is a substantial attention given throughout this book to the circumstances and actions of individual actors within the movement. So this, too, is shared with analytical sociology.

548
01:35:06.590 --> 01:35:32.769
So what about points of incompatibility and tension? Macadam? And this is the most important point of incompatibility. I would say macadam gives explanatory importance to dynamic meso-level entities in the explanations that he offers. In particular, he highlights the role that organizations played in facilitating and shaping contentious action throughout the South.

549
01:35:32.850 --> 01:35:42.879
and that those that we we cannot explain the development of collective action without referring to these organizations.

550
01:35:43.010 --> 01:35:59.180
A second point of divergence is, the macadam has a somewhat different view of the actor. in that he puts forward a relational and dynamic conception of the actor. The actor's intentionality and beliefs, and so forth, are not fixed

551
01:35:59.460 --> 01:36:12.519
throughout the process. Instead, they change their ideas and aspirations and grievances, and the definition of their situation through the relationships that they have with other actors and with organizations.

552
01:36:12.640 --> 01:36:32.180
And then, third, a point that I can't argue for here today, but have argued for elsewhere. There are no plausible agent based models of contention that have anything like the explanatory plausibility of Mcadam's multi level approach. Abm approaches to social contention have not proven to be very illuminating.

553
01:36:33.060 --> 01:36:48.610
So let's look at measure-level organizations for a second. But this plays an absolutely critical role in Mcadam's book organizations like the Council on racial equality, the and noble Acp, the student, nonviolent, coordinating committee, and so forth.

554
01:36:48.640 --> 01:37:02.210
and the networks of activists and lower level leaders that they drew around them in different social locations. These organizations play key causal moles in the account the macadam offers

555
01:37:02.410 --> 01:37:15.910
throughout the South. These organizations, furthermore, changed dynamically through the course of the civil rights movement. and they were located at different levels, some national, some regional, and some local.

556
01:37:16.780 --> 01:37:35.530
So the social movement, the civil rights movement that Mcadam describes cannot be analyzed in a way that ignores the activity and coordinating capabilities of organizations like these. You can't make his arguments without making essential reference to these kinds of dynamic organizations of

557
01:37:35.630 --> 01:37:51.090
civil civil right activism. These organizations, furthermore, cannot be treated as fixed and unchanging conditions. They cannot be exogenized. Instead, they are dynamic parts of the process.

558
01:37:52.000 --> 01:38:02.030
So Mcadam disagrees with the idea that causal mechanisms exist only at the micro level. But this is a bedrock assumption for people like Hedstrom.

559
01:38:02.070 --> 01:38:16.109
the Luca, the Mulneri, and Beerman Schultzer. Macadam's view is that dynamic organizations are essential to understanding the civil rights movement grievances at the individual level are not sufficient.

560
01:38:16.350 --> 01:38:29.609
But analytical sociology has a hard time incorporating organizations like the student nonviolent coordinating committee or the black Church into explanations of activism in mobilization.

561
01:38:30.000 --> 01:38:47.260
This leads me to an important conclusion that this finding about organizations requires abandoning the view that mechanisms exist only at the micro level. That core belief of analytical sociology is no longer adequate.

562
01:38:47.940 --> 01:39:12.259
Turn now to the relational theory of the actor. Mcadam's account of individual actors is admittedly sketchy, but it seems clear that his perspective does not attempt to explain mobilization simply as the aggregation of isolated, fixed, individual actors. Instead, he favors the relational actor theory, the idea that individuals develop through

563
01:39:12.260 --> 01:39:28.299
the struggles through the periods of activity that they are engaged in. They make decisions about action and inaction, in part through the ongoing relationships they have with other individuals, neighbors, co-workers, other people on the bus.

564
01:39:29.110 --> 01:39:41.330
the actors, motivations believe cemental frameworks take shape in relation to those of other actors, and they continue to change as a result of interaction. This is much harder to model.

565
01:39:41.370 --> 01:39:58.349
but it is more realistic. And again, it's a lesson which I think analytical sociology ought to pay attention to. There are promising approaches to social movements and contentious politics within analytical sociology, but there is still a lot of progress that needs to be made

566
01:39:58.550 --> 01:40:20.320
stethosc. Calabas has attempted to provide an analysis of the mechanisms through which individuals are brought to support group violence. But the only reference here in this account is reference to the organization. The only reference to organizations is to political parties, and these are treated statically and exogenously.

567
01:40:20.460 --> 01:40:33.989
Second example, Roberto Fjosi offers a causal theory of the temporal and spatial patterns of lynchings in the American South, and this account conforms to the as framework, and it is a a

568
01:40:34.060 --> 01:40:52.130
approach to empirical explanation of the pattern of lynching. But this account also addresses only the question of individual motivation of perpetrators. Neither essay provides a basis for understanding the broad mass movements the macadam considers within the civil rights movement.

569
01:40:52.610 --> 01:41:08.889
Delia Baldassari provides the most developed treatment of group political action within analytical sociology. Her approach gives primary attention to the motivations of individual participants. She comes out of the collective action

570
01:41:09.090 --> 01:41:43.250
problem framework, but she provides good descriptions of key ideas from the contentious politics literature. She pays attention also to the processes leading to group identity and shared interest formation, the kind of symbolic and meaningful part of the formation of the action orientation of the individuals. She acknowledges the causal importance of social networks, as indeed all or almost all analytical sociologists do within the process of group identity formation.

571
01:41:43.370 --> 01:41:59.520
However, the the account does not offer a convincing description of the workings of organizations in the creation of collective action, and her account remains excessively bound to the bottom up, or aggregative model of explanation and causal influence.

572
01:42:00.420 --> 01:42:14.910
So what's the scorecard? How did the test go out. In most respects I would say that Mcadam's treatment of the Civil rights movement is compatible with analytical sociology. The generata of good sociology, research.

573
01:42:15.400 --> 01:42:29.880
analytical sociology must confront the reality of Meso level structures and causal mechanisms, and in particular, analytical sociology needs a much more nuanced and dynamic theory of organizations.

574
01:42:31.480 --> 01:42:59.629
Further, analytical sociology should welcome even greater pluralism on the subject of the features of the actor who are involved in collective action and social movements, and finally, analytical sociology should abandon the dogmatic requirement that only ascending explanatory arguments from individual to macro outcome are acceptable. I think that Mcadam's book demonstrates the

575
01:42:59.800 --> 01:43:09.290
utility of a mixed-level analysis where measure level entities like organizations play a causal role. So thank you.

576
01:43:17.130 --> 01:43:23.729
Gianluca Manzo: Okay, thank you, Daniel. So I I I want to show you

577
01:43:24.210 --> 01:43:26.829
Gianluca Manzo: Daniel's presentation, because I I think

578
01:43:26.920 --> 01:43:52.820
Gianluca Manzo: it contains some important messages about about the realism of analytics, sociology, and more generally about the realism of of sociological exponential. I I I thought it was important to to show also his presentation. well, we have now about 10 min to Have a more general discussion.

579
01:43:52.940 --> 01:44:03.489
Gianluca Manzo: So if there is any any questions remaining, we can raise. there is in the room apparently a questions.

580
01:44:03.760 --> 01:44:05.120
This.

581
01:44:05.420 --> 01:44:06.110
Gianluca Manzo: yeah.

582
01:44:09.240 --> 01:44:11.969
Gianluca Manzo: yes. Hi, everybody.

583
01:44:12.580 --> 01:44:23.100
I have a question for Clement the You. You showed that there was new developments on the.

584
01:44:23.720 --> 01:44:33.350
So we have the

585
01:44:34.250 --> 01:44:39.069
developments on the theory side. So we have new

586
01:44:39.550 --> 01:44:46.600
greatest theories in the analytical sociology that we had when it's a so.

587
01:44:51.480 --> 01:44:53.469
Clemens Kroneberg: yeah, thank you. Yeah. And good to see you.

588
01:44:53.540 --> 01:44:56.310
Clemens Kroneberg:  So

589
01:44:56.470 --> 01:45:09.909
Clemens Kroneberg: let me put it this way. For example, I mean, Andreas, just showed that. Of course, there's theoretical progress in certain areas, right that we improve the on on existing models and try to add more realism.

590
01:45:09.970 --> 01:45:20.420
Clemens Kroneberg: And this is true, and also more monuments and complexity, so actually quiet along the lines that any little also called for. Right? Yeah, even though, of course.

591
01:45:20.500 --> 01:45:36.850
Clemens Kroneberg: organizations are for the issue. I agree, we can talk about that separately. So I would think that there is progress. But it doesn't happen at the same speed as the advancements of methods. And with regard to the availability of of new data, right? So it's difficult to.

592
01:45:36.980 --> 01:45:45.160
Clemens Kroneberg: And why that methods and data race ahead in a way. So in my own field of ethnicity and social networks.

593
01:45:45.290 --> 01:45:48.510
Clemens Kroneberg: and they are also, you know, people love to apply

594
01:45:49.559 --> 01:46:03.479
Clemens Kroneberg: more and more complex network models that to to new large data sources. But the theoretical arguments are often only borrowed from source of psychology right? And there is no really attempt to to come up with the convincing, theoretical.

595
01:46:04.020 --> 01:46:17.329
Clemens Kroneberg: formalized model. It allows us to to derive implications. Right? So I I would think that no way we we need to to try to keep up with these data and method developments.

596
01:46:18.309 --> 01:46:19.600
Clemens Kroneberg: It's a challenge.

597
01:46:22.340 --> 01:46:23.010
the

598
01:46:23.889 --> 01:46:25.150
Gianluca Manzo: thank you, Clement.

599
01:46:26.600 --> 01:46:31.860
Gianluca Manzo: There is a question from from Lucas. if there is no other question

600
01:46:31.950 --> 01:46:33.230
Gianluca Manzo: the room.

601
01:46:34.020 --> 01:46:38.430
Lucas sage: Yes. Oh. can I go? Yes.

602
01:46:39.090 --> 01:46:43.990
Lucas sage: So I have a a question for for on the yes

603
01:46:44.000 --> 01:47:05.049
Lucas sage: actually a short question. But so the shelling model was a. It's a the first papers that they died the date back to 1970 seventies. Let's say so. It's been the more than 50 years, and you just showed in your presentation that even for such a simple model

604
01:47:05.049 --> 01:47:16.630
Lucas sage: we are still discovering, exploring the the some properties, modifying a little bit this will produce this aggregate effect and stuff. So

605
01:47:17.029 --> 01:47:40.209
Lucas sage: my question is, do you think it's a bit provocative? But do you think like when we we go towards the really realistic model. And do you think it's going to be possible one day to have like a model of segregation that will integrate also all other parameters of the like

606
01:47:40.250 --> 01:47:49.550
Lucas sage: relevant to to actually how people select where they they are going to leave because I I I think you understood my my point.

607
01:47:50.400 --> 01:47:53.249
Andreas Flache: Yeah. Well, let me say that.

608
01:47:53.490 --> 01:48:01.870
Andreas Flache: I mean the reason we we chose the shelling model as an example for this paper was also that was meant to be a sort of introductory paper for people

609
01:48:01.879 --> 01:48:18.960
Andreas Flache: who have not seen been exposed a lot to age based modeling in analytical sociology.  I guess if you, if you collect everything that has been done on this model, you will find that there are many more advances, 3

610
01:48:19.059 --> 01:48:24.750
Andreas Flache: already in in understanding it and its implications, and making it moreistic, and so on.

611
01:48:24.870 --> 01:48:40.499
Andreas Flache:  erez agmoni at the same time. I think this. This may also reflect that very often in in research, let's say, advancing the model, trying to apply it to, to, to to explain a pyramid

612
01:48:40.639 --> 01:48:45.650
Andreas Flache: people would focus more on well, how can we get it to fit to the data

613
01:48:46.030 --> 01:49:02.790
Andreas Flache: and a little bit less on how can we really understand what happens? if we change certain things in the model, and how generalizable would that be? if we look beyond the particular data pattern we want to want to

614
01:49:03.480 --> 01:49:04.610
Andreas Flache: explain

615
01:49:04.830 --> 01:49:19.809
Andreas Flache:  so so I would. I would call for, let's say, going forward, let's say on in in different lines, but always in interaction, and communication with each other. So try to advance this 150

616
01:49:19.860 --> 01:49:36.360
Andreas Flache: E. C.

617
01:49:36.600 --> 01:49:51.659
Andreas Flache: and use different tools in our toolbox. not only engine best modeling not only fitting models to data, but also feeding and insights from experiment, qualitative research about people, how people really make them moving decisions

618
01:49:51.690 --> 01:49:53.480
Andreas Flache: and so on. That would be my.

619
01:49:53.710 --> 01:50:04.700
Andreas Flache: it might take. And just one small remark, if you allow. I I I would propose to in the future refer to this model as the model of shelling, and so.

620
01:50:04.760 --> 01:50:14.120
Andreas Flache: in the spirit of rhino hex, one's wonderful paper, and the recent piece in the New York Times about so coda no longer being forgotten.

621
01:50:19.210 --> 01:50:40.020
Gianluca Manzo: Thank you, Andreas. I I I had myself a question if if I can. and then I would like to raise a a last point to my question more on an institutional institutional level. So the question is a question for on race, and it's part to relate to Lucas questions.

622
01:50:40.440 --> 01:50:53.580
Gianluca Manzo: I I I obviously agree on on on the main message of of of your of your presentation and of of the chapters you refer to, which is essentially that we we Yup.

623
01:50:53.750 --> 01:51:01.700
Gianluca Manzo: we don't have to forget, the importance of understanding to why and how

624
01:51:01.730 --> 01:51:12.569
Gianluca Manzo: the model that would be produces what the model produces. So the the idea is the good model is the model that we can understand, and that I I I perfectly agree on this

625
01:51:12.790 --> 01:51:15.009
Gianluca Manzo: at the same time.

626
01:51:15.130 --> 01:51:15.950
Gianluca Manzo: and

627
01:51:17.180 --> 01:51:21.500
Gianluca Manzo: it. It seems as if you you you, you you present

628
01:51:21.750 --> 01:51:39.440
Gianluca Manzo: do do to really, really, really with dimension, and to understand the dimension as 2 dimensions that are partly in in conflict, or as if there was a tension between the 2 dimensions. And I I I was wondering if to some extent

629
01:51:39.670 --> 01:51:50.500
Gianluca Manzo:  a certain understanding of realism could be a solution to The fact that

630
01:51:51.000 --> 01:51:54.039
Gianluca Manzo: these models in particular are so flexible

631
01:51:54.120 --> 01:51:56.210
Gianluca Manzo: in the sense that

632
01:51:56.580 --> 01:51:59.080
Gianluca Manzo: if you know empirically

633
01:51:59.110 --> 01:52:05.959
Gianluca Manzo: that some aspects of your model are eggs rather than why?

634
01:52:06.090 --> 01:52:12.590
Gianluca Manzo:  you don't have to explore this aspect of your model.

635
01:52:12.740 --> 01:52:27.349
Gianluca Manzo: and so you don't have to say, Well, but if I use this network topology if I use these for for the physical neighbors, if and so on. So if you know that in the reality.

636
01:52:27.620 --> 01:52:40.539
Gianluca Manzo: The distribution of preferences is this one, if you know that the most realistic net or fetus are this one. So this is a way to reduce the degree of freedom of your of your model.

637
01:52:40.600 --> 01:52:50.219
Gianluca Manzo: So realism. It's not necessarily an obstacle to understand your model. It's something to help

638
01:52:50.230 --> 01:52:58.150
Gianluca Manzo: not to have too much, too many details to explore, because one of the risk is that

639
01:52:58.990 --> 01:53:04.080
Gianluca Manzo: you? Well, you you you never go out from from the model. Yeah.

640
01:53:04.640 --> 01:53:18.440
sorry. We have another session on side here. So could you please nap up very shortly, because the time 20 is almost us.

641
01:53:18.940 --> 01:53:22.630
Gianluca Manzo: Okay, yes, yes, I I, I I'm done. I'm done

642
01:53:22.770 --> 01:53:24.410
really like

643
01:53:24.440 --> 01:53:25.819
to the next system.

644
01:53:26.620 --> 01:53:40.020
Andreas Flache: Okay, I could give a one sentence answer, with which is, I, I agree. But empirical knowledge often is, of course, also context bounded. So very often, if we say, Okay, we know this is

645
01:53:40.340 --> 01:53:51.699
Andreas Flache: true, and therefore it rules out lots of possibilities that we don't need to explore them anymore in the modeling exercise that made true be true for certain contexts for which we have the empirical knowledge.

646
01:53:52.400 --> 01:53:54.219
Gianluca Manzo: I I I agree. Yes.

647
01:53:54.440 --> 01:54:08.110
Gianluca Manzo: very quickly. Maza just 30 s Andreas wanted to say something about rationality and and and society

648
01:54:08.200 --> 01:54:16.620
Gianluca Manzo: journal so very quickly, Andreas, please go go go ahead

649
01:54:16.740 --> 01:54:30.039
Andreas Flache: I just wanted to to raise your attend, or ask for your attention, for this journals of which I'm now, the editor in chief succeeding Dr.

650
01:54:30.290 --> 01:54:47.650
Andreas Flache: And just well, this is the Rc. 45 rational choice, though we have just critically discussed rational choice. And this journal really is not only devoted to rational choice. my perspective is, Well, of course, there must be clear link

651
01:54:47.650 --> 01:55:07.879
Andreas Flache: between your work and rational, action-based theory if you publish there. But it can also be a critical link. It can be a link like, Okay, we use a different behavioral model. This is the difference to the rational action approach, and we come to different insights, or we have a certain critique on the model, that's all valued and fine.

652
01:55:07.880 --> 01:55:11.159
Andreas Flache: I just want to also say,

653
01:55:11.220 --> 01:55:33.099
Andreas Flache: well, we have worked hard in recent years to increase the turnaround time of decisions. We are now at about 2 weeks for initial decisions. So if you submit something, after 2 weeks you will know whether it's desk rejected or goes into review typically in the last one or one year are in our decisions. Are we taken within about 2 months?

654
01:55:33.220 --> 01:55:57.410
Andreas Flache:  and if you can go further in the process. typically, you can get very close editorial guidance, which is not meant to be scary, but which is meant to be that I at least try to help authors to then develop their paper even better, and make it a nice contribution to the journal. So please consider submitting your work to this channel.

655
01:55:57.600 --> 01:55:59.169
Andreas Flache: Thanks a lot.

656
01:56:03.920 --> 01:56:10.950
Gianluca Manzo: Thank you very much, Andreas, and thank you all for your very, very nice and reaching talks I.

657
01:56:10.970 --> 01:56:15.689
Gianluca Manzo: So it was very, very nice, nice sessions. And so thank you. And

658
01:56:15.940 --> 01:56:24.470
Gianluca Manzo: see you hopefully. in the real life, in the close future. master, do you have anything to add.

659
01:56:24.630 --> 01:56:30.000
Okay, thank you very much. So this will close that session. Good.