03/03: Epstein

[ehuppert]

Comment below with a well-developed group question about the reading for this week's workshop. Please collaborate with your groups on Hypothesis (via the Canvas page) to develop your question.

One person can submit on the group's behalf and put the Group Name in the submission for credit. Your group only needs to post on assigned week (rotating every other week).

Please post your question by Wednesday 11:59 PM, and upvote at least three of your peers' comments on Thursday prior to the workshop. Everyone in the group needs to upvote! You need to use 'thumbs-up' for your reactions to count towards 'top comments,' but you can use other emojis on top of the thumbs up.

[Raychanan]

Group 1C: Rui Chen, Yawei Li, Val Alvern Cueco Ligo, Yutai Li, Max Kramer

Hi Professor Epstein, many thanks for sharing your work with us!

For the lack of previous experience that I have, I find it difficult to comprehend the framework. Using neurocognitive science as a foundation, you present Agent_Zero as an agent that can outperform the Homo economicus model in instances like genocide. You developed some interpretations in order to illustrate its generative power, where slaughter represents the darkest example. One of the central human paradoxes is precisely that we do see another side: the same Germany that produced Hitler also produced Einstein. My question is, following the Agent Zero model, whether we will see a higher probability of "Hitler" over "Einstein" following Agent Zero? Or vice versa?

[JadeBenson]

Group 1K: Jade Benson, Joseph Helbing, Hazel Chui, Isabella Duan

We think incorporating affective and social components to agent-based modeling that expands beyond just reason is philosophically exciting! Your discussion of fear as a motivator is compelling and applies to many circumstances we can think of (pandemics, politics). However, we wonder what other types of affects we should adjust for and if there are any that are more compelling than fear? Does the social component account for the role that love and passion play in our lives? How do we account for innate interest and curiosity that go beyond fear and shape our beliefs and actions?

[Tanzi11]

Group 1G: Tanzima Chowdhury, Yulun Han, Tian Chen, Qihui Lei

Looking forward to learning more about your work tomorrow, Professor Epstein!

Admittedly, in our group we are not familiar with the mechanisms of cognitive neuroscience, but there is a poetic quality to the idea of Agent Zero that we find appealing; so please forgive any misunderstandings of your work on our part.
When constructing the theoretical model, was there a specific society that you had in mind and would it be applicable then for any future scenario? If we understand correctly, can fear and our reactions to it be only considered from the perspective of the amygdala? How much of societal conditioning and culture have an impact and how can these be taken into account when studying neurocognitive processes? You do reference that “unconscious conditioned fear may be modified both by conscious deliberation and (often unconscious) social influence” —by this it is presumed you mean the act of un-conditioning one’s self out of fear, but what of acquiring that fear? Also, philosophically, can a model really be sensitive to affectivity and human emotion? Many thanks!

[borlasekn]

Group 1E: Kaya Borlase, Xin Li, Zoey Jiao, Shuyi Yang

Thank you for sharing your expertise. Our group had a couple of questions:

1. You mention that fear conditioning can occur unconsciously. In what ways do you think this could occur with over-exposure to media in the digital world? Would any changes here be captured by your agent modeling?
2. There are 3 modules: emotional, cognitive, and social factors in the agent_zero model. When obtaining the individual's, the three components seems to be considered separately, and then they were combined to compare with the threshold. I'm wondering that how could we separate the three components clearly? Like in come cognitive component, there might also be mixed with emotional behavior, since in my viewpoint, it's hard to distinguish how much of the behavior is out of reason, and how much is out of passion. So how to consider the situation?

[hsinkengling]

Group 1H: Yuetong Bai, Boya Fu, Zhiyun Hu, Hsin-Keng Ling

Thank you Professor Epstein for bringing us such an interesting topic on agent-based modeling!
We have prepared a number of questions:

1. In the Minimalism part, you mentioned that you have hand-coded three agents to control the smallest possible model that could apply to general cases, we are wondering why did you choose the three agents rather than a more or fewer number? Besides, I could imagine how the model could apply to other fields such as network science, but could you please explain more on the interpretations for economics and law?

2. My question is about weighing consequences and costs. (Apologies in advance if this question is already covered elsewhere in the book. I only read the recommended parts) In the conflict/guerilla warfare simulation, it seems that actors are assumed to decide to kill or not kill only based on one's disposition value crossing the action threshold.

However, it seems intuitive that actors also need to consider the consequences and costs of their actions (cost of energy to kill, consequences of counterattacks). Hence, it's more common to see soldiers combat in grouped units rather than alone due to strength in numbers. On a broader level, do you think the idea of collective action alignment should be included in your account of disposition (that you can't talk about disposition spreading without talking about collective alignment)? or do you think that collective alignment is a separate issue and should be modeled as a separate, additional condition?

3. When building agent-based models, it seems essential to keep the simplicity of rules for agents. However, We need extensions for the model to simulate a more complex real-life social situation. What may be the principle for choosing the right extension rules while preventing the model to be too complex? What level of complexity do you think the model should have when it can simulate a 'real' situation?

[DehongUChi]

Group 1N: Henry Lin, Dehong Lu, Alfred Chao, Naiyu Jiang, Qiuyu Li

Thanks for coming and sharing with us your exciting works, the idea of Agent Zero is truly fascinating. You discussed fear as a motivator in many circumstances, but are there other constructs that we could use in combination with fear? (or perhaps you have used other constructs in other models, I have only read the recommended part)

Another question is that we understand when building an agent-based model, keeping the rules simple is important. However, I assume in order to simulate more complex real-life scenarios, there often inevitably needs to be an increase in the level of complexity of the rules. What ways do you think are the best to improve the models without making them "too complex"?

Thanks!

[FrederickZhengHe]

Group 1A: Angelica Bosko, Chongyu Fang, Frederick Zheng He, Yier Ling

Dear Professor Epstein,

Thank you so much for presenting your research today. Our questions are: Are there any alternative models other than the Rescorla-Wagner model? If there are, will these models be better than the Rescorla-Wagner model, and will the results of these models be significantly different from the results generated from the Rescorla-Wagner model?

[Toushirow1]

Group 1D: Zixu Chen, Mengchen Chung, Yujing Huang, Feihong Lei, Zhihan Xiong

1. According to the theory of delayed feeling and my experience, conscious fearing-inducing and physiological responses seem to happen concurrently. Since the human brain better controls the expression of complex conscious physiological stimuli, does it mean we can autonomously suppress the feeling of fear if we do the opposite of our physiological response? Also, how can we design the experiment to verify the fear is stimulated after physiological responses?

2. How is Agent Zero different from the traditional definition of a rational economic individual?

3. How to quantify an agent’s emotion variable(s) and how such variable(s) impact individual behavior?

[fiofiofiona]

Group 1F
Thank you for sharing your research and models with us. It is interesting to start with the mechanism of fear conditioning, and I found the “low road” and “high road” really intriguing. However, I wonder how the “fight or flight” phenomenon can be explained by these two routes — are they conscious decisions that went through the high road and being made after conscious calculation? Or are they different kinds of unconscious reactions wired to us?
Another question came to my mind when I was reading the modeling part. As shown in the figures, when the overall disposition of blue agent and orange patches reaches threshold, an action is taken, and damage is made within certain radius. What is the role of fear conditioning in this algorithm? When the blue agent feels fear, does this emotion decrease overall disposition and therefore he is less likely to take the action? Or does unconscious reaction change little to the disposition?

[sabinahartnett]

Group 1J: Sabina Hartnett*, Lynette Dang, Silvan Baier, Yingxuan Liu

Really interesting work and a new approach on agent-based modeling! Although this model discussed fear as the primary motivator, we were wondering about confounding factors that influence how and whether fear is detected in certain scenarios. In your paper you reference a snake appearing in one's path - of course, anyone who recognizes a snake and the potential danger they pose would be afraid (although, arguably someone more expert in snake species may know not to fear a garden snake or other not threatening snake), but some fears are specific to individuals (confounded by personal or learned traumas) and wouldn't be universally observed. Perhaps its my own lack of familiarity with agent based modeling but can we assume consistency amongst actors in their perception of fear? How have you handled this?

[xxicheng]

Group 1L: Xi Cheng, Elliot Delahaye, Hongxian Huang, Yutong Li

As a group of sociology/economy students who hold a very amateur understanding of neuroscience, we still find the idea of capturing emotions with computational models inspiring and illuminating. Some sociologists in the field of social networks have long been diving into better understanding or even predicting human emotions and other social interactions and processes that might have been influenced. As MACSS students, we also frequently heard from computer scientists using machine learning, big data, etc., in these studies. Therefore, it is fascinating personally for me to get to know neuroscientific models, equations, etc., which is apparently a fresh angle to think about the production and influence of emotions within brains. Thanks for sharing your work with us.

So our question is: the generation and diminishment process of emotion are always case-specific. "Affect (positive or negative) [in some cases] can remain above the action threshold long after the stimulus has stopped." Therefore, what is your expectation on the generalizability of your model? More specifically, do you think the fearing model is somehow culture-specific and individual-specific? Maybe some groups are more anti-Japanese than others due to the unique cultural and political contexts of the former. Also, how are you and your team dealing with possible hundreds of, if not thousands of, different human emotions besides fearing? I am interested in balancing the population- and individual-level results regarding your findings.

[pranathiiyer]

Group 1B: Pranathi Iyer, Guangyuan Chen, Qishen Fu, Yuxuan Chen
Thanks for your paper professor! Like most of my classmates, I'm not very familiar with agent models but you did mention in your paper that we could look at phenomena such as measuring the probability that an agent might be a terrorist based on whether he is a muslim or not. What could be some of the possible applications and implications of simulations and models like these?

[Jasmine97Huang]

Group 1I: Yu-Hsuan Chou, Bowen Zheng, Jasmine Huang, Jingnan Liu, Yile Chen

Thank you so much for coming to the workshop! The idea of Agent_Zero is particular convincing as an appropriate subject for Agent Based Modeling in social science because of its capacity for emotional responses and the social influence from their fellow agents. However, I am curious how Agent_Zero can be adopted to model socially deviate or anti-social behaviors that do not follow the emotional responses coded into the social conventions in the artificial world?

[j2401]

Group 1I: Yu-Hsuan Chou, Bowen Zheng, Jasmine Huang, Jingnan Liu, Yile Chen

Thank you Professor Epstein for sharing with us on agent-based modeling!

1. The idea of Agent_Zero is particular convincing as an appropriate subject for Agent Based Modeling in social science because of its capacity for emotional responses and the social influence from their fellow agents. However, I am curious how Agent_Zero can be adopted to model socially deviate or anti-social behaviors that do not follow the emotional responses coded into the social conventions in the artificial world?
2. We found that the four cases in Run 4. A Day in the Life of Agent_Zero were particularly interesting and illustrative. Based on my naïve understanding, we think the key ingredient here is the exponential decay of the feelings. It seems that we know this property of our brain before hand so that we could implement it into our model. Our question is that, as we may discover more secrets in our brain and continue to add such properties, do you expect the interactions of these features to be still tractable and interpretable?

[jiehanL]

Group 1M: Jiehan Liu, Partha Kadambi, Peihan Gao, Shiyang Lai, Zhibin Chen
Thank you Prof. Epstein for sharing such novel research! Our question is mainly about the evaluation of the performance of the agent_zero model: as the experimental framework is difficult to produce in agent_zero model, how should we evaluation the model performance, or compare different models? how should policy change when agents’ endogenous differences are taken into consideration?