3/23/2023: Jordan Kemp

[GabeNicholson]

Comment below with a well-developed question or comment about the reading for this week's workshop.

If you would really like to ask your question in person, please place two exclamation points before your question to signal that you really want to ask it.

Please post your question by Tuesday of the coming week, at 11:59 PM. We will also ask you all to upvote questions that you think were particularly good. There may be prizes for top question-askers.

[JoeHelbing]

Super interesting work Jordan Kemp,

I didn't see it in the paper and I'm curious what was the framework you used to code up the agent model? What, if any was the trial and error process in getting the desired behavior in the model?

Thanks

[sdbaier]

Jordan, thank you for sharing your work. Instead of informing me, it has left me feeling confused.

What are signals? Given their interchangeable use in the paper, are subjective signals, private signals and opportunities the same? If so, doesn’t the assumption that each agent samples identically distributed signals violate the notion of subjectivity?

What are events? Are environmental outcomes a subset of events?

What are investments? Are investments into the environment and investments into events the same?

I have some familiarity with topics in population ecology and approaches to computational simulations, but in its current form, your paper is simply not accessible to me. Using simpler language, more accessible examples or analogies, or providing a richer background would aid my understanding. I am looking forward to your talk, particularly as I hope to (better) understand your research then.

[yuzhouw313]

Hello Jordan,

Thank you for sharing your research on using stochastic methods to approach the problem of dynamic social inequality regarding wealth accumulation and decision-making. I look forward to hearing more about your study on Thursday.

As a sociology student interested in social stratification and inequality, I believe that your focus on optimal learning based on information-driven growth explains the fast accumulation of wealth of the wealthy population. Could you elaborate on how populations with less wealth can utilize "signals" and other optimal decision-making strategies?

Additionally, as you mentioned in your conclusion, the process of learning and decision-making is not uniform or consistent across populations or time. How do you plan to address the problem of the almost instantaneous and stable decisions made by the agent/model?

[AlexBWilliamson]

Thank you for sharing your research with us Jordan! I was very interested in seeing you create a model for how inequality grows, and how learning can prevent it. I thought that all of the simulations that you did were especially interesting.

I was wondering if you tried applying real world data to your model, in addition to the simulations. It might not be necessary for the paper you are writing, but if you were able to calibrate your model with real world data I think it would be a very interesting application of your work.

[Sirius2713]

Hi Jordan! Thanks for sharing your research with us. You made some assumptions and simplifications to make the real-world situation close to statistical theories. I'm wondering how these assumptions and simplifications would impact your conclusions and whether there could be some ways to lax the assumptions.

[secorey]

Hi Jordan, You mention being able to "mechanistically control variances in growth rates across a society." Though you were able to achieve this in your model, what do you think can be done in a real-world environment? Thanks for your presentation, Sam

[beilrz]

Hello. Thanks for the inspiring paper. My biggest concern was the assumption of individual sampling of identical signals, without considering the underlying social network, which you addressed in the conclusion section. I also want to ask do you plan to empirically examine the finding of this paper in real life scenarios? If so, how would you plan to do it? as it is probably very difficult to observe the signal space available to individuals.

[zihua-uc]

Hi Jordan, Thank you for sharing this interesting framework! As you mentioned in further applications of the model, learning could be made heterogeneous by making inference rates dynamic over time or different across populations. It would be interesting to see how agents' optimal parameters respond to these extended models.

[jinyz1220]

Thank you for sharing your work! You mentioned in the conclusion that the learning rate can be made dynamic and heterogeneous to provide a more realistic model by, for example, introducing SES into the model. I would love to know that how to interpret the model with higher complexity, and what suggestions it can imply to reduce wealth and information inequality?

[sushanz]

Thank you for sharing Jordan. Really interesting topics. You mentioned that stochastic multiplicative dynamics always occur in different natural phenomena, such as wealth distribution and evolution. I wasn't aware that population heterogeneity in stochastic growth rates could be identified as a critical factor. The argument you put forth in the paper was based on mathematical and statistical models and empirical evidence as well. However, I wonder how you think of those uncertainties in the real environment as it is hard to capture the real-world social and biological phenomenon. You and your research partners also discussed that further discussions and researches are needed in this field to confirm the theory's applicabilities and effectiveness in the real and different contexts. How would you suggest future researchers to contribute and develop in this area?

[LynetteDang]

Hi Jordan,

Thank you so much for sharing your work with us. I am wondering the reason that you chose sequential Bayesian inference instead of variational inference or other statistical techniques.

[JunoWuu]

Hi Jordan!

You mentioned that agents consider the cost and benefits of learning and that teaching is costly. However, I wonder as information becomes more and more available, and forms of teaching is becoming more diverse, how that would affect the agents' way of gaining information? and how that would affect your model?

[Dededon]

Hi there! It's a refreshing research topic to combine ABM simulation with information theory. I'm interested in the part of parameter selections in agent based models. Model behaviours could be very volatile under different sets of parameters. As you defined the information learning and growth rates with math-intensive methods, I wonder what are the previous literatures and theories that helped you come up with those formulas. Also, what are your process and your experiences of doing the parameter tuning? I'm interested in how you resolved with managing such complexity and difficulties in accomplishing those tasks.

[YLHan97]

Hi Jordan,

Thank you so much for sharing your work with us. I have the following questions:

1. How does learning contribute to reducing inequality in shared noisy environments according to the research?
2. What implications does the study have for designing educational policies aimed at promoting growth and reducing inequality?

[Yuxin-Ji]

Hi Jordan, thanks for sharing your work with us. It is interesting to learn about approaching growth and inequality through information theory. What are some potential applications of this proposed theory? How generalizable is it on different populations and environments?

[XTang685]

Hi Jordan, thank you for your work! I am curious what is the methodology used in the study to simulate shared noisy environments and measure the effects of learning on growth and inequality? Thank you!

[bhavyapan]

Hi Jordan, thank you for sharing your work! The framework leveraging the application of information theory to growth and inequality is super interesting, and was a totally new avenue of research for me. I'm wondering what the implications of the assumptions can be vis-a-vis the question of designing policies and studying socio-economic indicators in various kinds of environments. It would be very illuminating for my understanding if you could illustrate an elaboration of the applications of such frameworks. Best wishes for the talk, looking forward to it!

[kuitaiw]

Dear Jordan, I would like to ask how you plan to deal with heterogeneity. We know that even under the same conditions the effect of the same thing may vary from person to person. So I would like to ask how we should eliminate heterogeneity if we bring the model in this article into the data in the real world?

[taizeyu]

Dear Jordan, Because the model is a simplified representation of complex natural phenomena, it may not capture all the nuances of real-world scenarios. Whether future research will focus on this?

Thanks

[y8script]

Hi Jordan, thank you for sharing this work with us! I would like to learn more about the implementation details of this model. In your opinion, how to better connect theories from cognitive psychology with agent-based dynamic modeling? Could more complex cognitive psychology phenomenons be quantified with this modeling method?

[yhchou0904]

Hi Jordan, Thank you for sharing your idea with us. There are many presumptions about the underlying distribution of learning outcomes and signals, just like in the aforementioned comments and the discussion in your work. Do you have any theories in mind for the direction of the effects, even though you mentioned that these should be looked into further?

[tangn121]

Hello Jordan, Thank you for sharing your work. You propose an approach to growth phenomena in heterogeneous populations based on agent decision-making in a noisy environment with many assumptions, I am curious about your plans to align it more closely with real-world settings. Could you please elaborate on your future plans for this research? Thank you!

[hsinkengling]

Hi Prof. Kemp,

Thank you for sharing your work with us. I'm interested in the model's external validity considering the sheer scale of existing inequalities. Given the Matthew effect (rich gets richer, poor gets poorer), the actual conditions of social inequality seem to be very different from that one individual learning, investing, and gaining, and rather mostly based on inheritance and scaling. Yet your model also kind of make sense for individuals with similar economic standing. How do you think your model fits within the bigger picture of existing wealth disparities?

[borlasekn]

Hi Jordan, thanks for sharing your work! You mention that coupling learning rates with sociodemographic factors (such as SES, in your work) would alter the learning trajectories of individuals. I am wondering how this sort of coupling might occur overtime, accounting for things such as the weathering effect (i.e. that repeated exposure to systemic barriers can cause increasing negative effects over time). Can you account for increasing negative effects on learning over time through these types of models? Thanks!

[zhiyun0707]

Hi Jordan, thanks for sharing your work with us! My question is that you mainly elaborated the theory and modeling mechanisms behind the information-based growth in the paper, could you please talk more on the applications?

[mdvadillo]

Hi Jordan, really interesting work! I was wondering if you had tested if your results are robust to different activation schemes, and if not, do you think there is something to be gained by that?

[mingxuan-he]

Hi Jordan, I'm wondering since the model assumed that each agent samples signals individually, would it make sense to add a utility function which is heterogeneous among agents to reflect the possibility that the preferences of agents might differ?

[xin2006]

Hi Jordan! Thanks for sharing the research. The paper provides a valuable insight into the potential benefits of learning in a shared environment with uncertain information. However, I am curious about the assumption that learning always leads to increased productivity and economic growth. In reality, the relationship between education and economic outcomes can be complex. How did you consider the real-world factors when employing the simplified models in the research?

[zoeyjiao1104]

Hi Jordan, thank you for sharing your work. I'm really curious about the application of your research to real world cases. In addition, could you elaborate a bit more on the external validity of your research? Thank you!

[mintaow]

Hi Jordan, thanks for sharing your work with us. Super interesting work! I am particularly curious about jow did you select the agents and environments used in your simulations. And if you could describe a bit more about the process of developing and testing your model of agent decision-making in noisy environments, that would be great. Thanks. Looking forward to the presentation.

[ShiyangLai]

Hi Jordan. Thanks for sharing your papers with us. I was wondering how might the findings on the role of mutual information and sequential Bayesian inference in attenuating growth rate disparities be applied to design more equitable social and economic policies that reduce wealth inequality in real-world settings?

[javad-e]

Thank you so much for presenting your work at our workshop, Jordan! The model is very interesting and can explain the trends we observe in the real world. Of course, one limitation of many agent-based models is not examining the behavior of outliers. What are some potential factors that might result in outlying residents not following this pattern?

[awaidyasin]

Professor, thank you for sharing your work.

I see that agents have individual endowments, receive private signals, and then invest in events e, but I think there is no communication/learning between the agents (i.e., they go about their processes independently without learning from the signals of others). Is this a necessary simplifying assumption of your work, or can this be relaxed to study more realistic societies/setups?

[Coco-Jiachen-Yu]

Hello Jordan,

Thank you so much for sharing you work with us. I look forward to your presentation tomorrow! The topic is interesting and I'm wondering if the models will induce algorithmic biases towards certain populations. How can you identify or prevent potential biases?

[iefis]

Thank you for presenting your work with us. It seems to be an interesting baseline model for understanding the theoretical foundation of wealth dynamics. I'm wondering how could we further develop the current model to incorporate the potential information complementarities often seen in real world. Thanks!

[hazelchc]

Hello Jordan, I'm interested in the policy implications of your research, which appears to be highly theoretical. I'm wondering if you anticipate any discrepancies between the real-world situations and the model you've developed. Thank you for your time!

[HongzhangXie]

Dear Jordan

Thank you very much for sharing this interesting paper. I am interested in the hypothesis and research method differences between physics, computer science, biological, and social sciences in analysing the question about growth, population heterogeneity and wealth inequality. How does the model help us to make connections between different fields research models?

[helyap]

Hi Jordan,

Thank you for sharing your work with us. I'm curious if you have tested your approach on existing data and how interaction effects within events are modeled.

[yujing-syj]

Hi Professor Jordan, thanks so much for sharing this interesting topic with us! I am curious about whether there are some other findings related with the connection between social behavior and growth from the point view of information and learning.

[ValAlvernUChic]

Hello Jordan,

Thank you for the paper. I was just wondering if this paper assumes that with access to information, agents will just then be able to optimize their personal wealth. I might be misunderstanding but doesn't this firstly just assumes that all agents will act on this information, secondly that they have the ability to act on this information, and lastly ignores other socio-institutional factors that complicate this dynamic?

[jiehanL]

Hi Jordan,

Thank you for presenting your work! My question about that paper is, what is the impact of population heterogeneity in stochastic growth rates on wealth inequality over long time scales, and how can this be explained by a general statistical theory that accounts for the adaptation of agents to their environment and the subjective signals each agent perceives?

[ChongyuFang]

Dear Jordan, Thanks for presenting the paper! Can you explain the relationship between population heterogeneity and wealth inequality in more detail?

[bningdling]

Hi Jordan, thank you for sharing your work with us. I was wondering what is the significance of incorporating Bayesian learning in managing growth disparities across populations over time, and how does this approach differ from traditional wealth dynamics modeling.

[BaotongZh]

Hi Jordan,

Thank you so much for your work. I was also wondering if you could help me understand how policy plays a role when treated as an information source.

[lguo7]

Hi Jordan, Thank you for sharing your recent research with us. I am very interested in the concepts you present, but limited to the fact that I can only see the draft version of this paper, I cannot go deeper into the subsequent conclusions and discussions. My current question is, how did you measure the subjective signal? Is there a bias between this signal and P(S) in a noisy environment? Can you elaborate on how you deal with such deviations?

[erweinstein]

Hi Jordan,

I agree that sequential Bayesian inference is a very useful approach for modeling the path-dependent/history-dependent manner in which individuals form their sequence of optimal decisions. I do, however, share some of my classmates' skepticism regarding the external validity of your modeling approach. Rather than repeat those points, instead I'll ask a (related) question about science communication: Do you find that when you present for (or just talk to) audiences of people from different disciplines that they identify different angles with which to engage with your work that vary in a way that's correlated with their discipline, or is there no apparent correlation? If so, that might be an example of the system "working" (as it were), which justifies investments in interdisciplinary programs and workshops like this one. Or, in your experience, is it more of an issue of differential misunderstandings, where people get caught up in terminological differences or other meta-level points of disagreement or confusion (or not) in ways that correlate with how far apart their disciplinary background is from yours?

[Thiyaghessan]

Hi Jordan, thank you so much for the wonderful paper! I wondered if you had any thoughts on including structural inequalities in the model. For example, after some time, wealthier agents could work to engineer laws/policies that benefit them, making it harder for social mobility to occur. Additionally, I wondered if it would be helpful to include shocks, e.g. financial crises, that are typical of boom/bust cycles and have the potential to result in wealth destruction and lead to the emergence of a new class of wealthy agents.

[shaangao]

Thanks for sharing your work with us, Jordan! I'm not very familiar with the socio-political science field, and I was wondering if you could share more about how you select the model to use, and how you extend a schematic model by introducing more variables and relationships into it step by step. Thank you!

[Emily-fyeh]

Thanks for sharing your work! I wonder how we adapt and apply the model in some desired real-world settings under your set of assumptions. Also, how would you predict and interpret the signals from different environments? Thanks!

[yjhuang99]

Hi Jordan, thanks a lot for sharing your interesting work with us! I wonder if it is possible for you to share more about how your model fits the real data and the policy implications of your model. Thanks!