**aamas : track area chairs' papers and get a general feel of what ppaers they want there
**parcel eaamo paper out in peieces -> is it 2-3 papers in one? -> if so, which parts are the break points where we can divide it into multiple papers.
** place of this paper in the arch of all papers -> should ppaer with aditya come first?
**aamas if we wanna submit then how to re-write to fit aamas?
**aistats also is there and also the paper with aditya can be submitted there or if aamas is no good then can we submit the eaamo paper here as well?
**what route are we taking to re-write things? (no relevance did not come up in eaamo, was it because there were no economists in the reviewers again? do we need to go the ruin route that we discussed to get more AI-ish experimental/simulation results?)
***
AAMAS : track pc papers and get a general feel of what papers they want there

Papers that maybe relevant

From Harko Verhagen

[GAM on! Six ways to explore social complexity by combining games and agent-based models](https://www.tandfonline.com/doi/full/10.1080/13645579.2022.2050119)

[Trustworthy AI for people?](https://dl.acm.org/doi/abs/10.1145/3461702.3462470)

[Advances in Social Simulations](https://books.google.com/books?hl=en&lr=&id=f8fgDwAAQBAJ&oi=fnd&pg=PR5&dq=info:zQ2xvTjuYAwJ:scholar.google.com&ots=2zrre7x6EY&sig=V8OyFo4isFwrnvS9NtHJwCSLanU#v=onepage&q&f=false)

Being Transparent about Transparency in the Context of Artificial Intelligence

From Micheal Lees

[Complex agent networks: An emerging approach for modeling complex systems](https://www.sciencedirect.com/science/article/pii/S1568494615005049)

[Survey-based socio-economic data from slums in Bangalore, India](https://www.nature.com/articles/sdata2017200)

[Game Theoretic Modeling of Helping Behavior in Emergency Evacuations](https://www.sbscommunity.nl/resources/uploads/2021/02/Kwak.pdf)

From the previous conferences (2022,2023)

[Agent-Based Modeling of Human Decision-makers Under Uncertain Information During Supply Chain Shortages](https://www.southampton.ac.uk/~eg/AAMAS2023/pdfs/p1886.pdf)
Nutchanon Yongsatianchot *(Northeastern University)*
Noah Chicoine *(Northeastern University)*
Jacqueline Griffin *(Northeastern University)*
Ozlem Ergun *(Northeastern University)*
Stacy Marsella *(Northeastern University)*

In recent years, product shortages caused by supply chain disruptions have generated problems for consumers worldwide. In supply chains, multiple decision-makers act on uncertain information they receive from others, often leading to sub-optimal decisions that propagate the effects of supply chain disruptions to other stakeholders. Therefore, understanding how humans learn to interpret information from others and how it influences their decision-making is key to alleviating supply chain shortages. In this work, we investigated how downstream supply chain echelons, health centers in pharmaceutical supply chains, interpret and use manufacturers’ estimated resupply date (ERD) information during drug shortages. We formulated a computational model of a health center based on a partially observable Markov decision process that learns a manufacturer’s information sharing tendencies through an observation function. To investigate the model and important factors influencing decisions and perceptions of ERD, we conducted a human experiment to study where subjects played the role of a health center during a drug shortage. They received ERDs from a manufacturer on a weekly basis and decided whether or not to switch to an alternative product (and pay additional costs) to avoid running out of stock. The results show that different manufacturers’ sequences of ERDs and the accuracy of ERDs could impact subjects’ decisions, beliefs, performance, and perception of the manufacturer. We also found that the subjective belief of ERDs is the best predictor of subjects’ switching decisions. Lastly, we fit the observation function’s learning rate and show that the model can predict subjects’ decisions better than other baseline models in most conditions.

[Modelling Agent Decision Making in Agent-based Simulation - Analysis Using an Economic Technology Uptake Model](https://www.southampton.ac.uk/~eg/AAMAS2023/pdfs/p1903.pdf) (Page 1903)
Franziska Kl�gl *(�rebro University)*
Hildegunn Kyvik Nord�s *(�rebro University)*

Agent-based Simulation Modelling focuses on the agents’ decision making in their individual context. The decision making details may substantially affect the simulation outcome, and therefore need to be carefully designed. In this paper we contrast two decision making architectures: a process oriented approach in which agents generate expectations and a reinforcement-learning based architecture inspired by evolutionary game theory. We exemplify those architectures using a technology uptake model in which agents decide about adopting automation software. We find that the end result is the same with both decision making processes, but the path towards full adoption of software differs. Both sets of simulations are robust, explainable and credible. The paper ends with a discussion what is actually gained from replacing behaviour description by learning

If we do interventions,

[Using Agent-Based Simulator to Assess Interventions Against COVID-19 in a Small Community Generated from Map Data](https://www.ifaamas.org/Proceedings/aamas2022/pdfs/p1.pdf)

​ Mitsuteru Abe *(University of Tsukuba)*
​ Fabio Tanaka *(University of Tsukuba)*
​ Jair Pereira Junior *(University of Tsukuba)*
​ Anna Bogdanova *(University of Tsukuba)*
​ Tetsuya Sakurai *(University of Tsukuba)*
​ Claus Aranha *(University of Tsukuba)*

During the COVID-19 pandemic, governments have struggled to devise strategies to slow down the spread of the virus. This struggle happens because pandemics are complex scenarios with many unknown variables. In this context, simulated models are used to evaluate strategies for mitigating this and future pandemics. This paper proposes a simulator that analyses small communities by using real geographical data to model the road interactions and the agent’s behaviors. Our simulator consists of three different modules: Environment, Mobility, and Infection module. The environment module recreates an area based on map data, including houses, restaurants, and roads. The mobility module determines the agents’ movement in the map based on their work schedule and needs, such as eating at restaurants, doing groceries, and going to work. The infection module simulates four cases of infection: on the road, at home, at a building, and off the map. We simulate the surrounding areas of the University of Tsukuba and design three intervention strategies, comparing them to a scenario without any intervention. The interventions are: 1) PCR testing and self-isolation if positive; 2) applying lockdown measures to restaurants and barbershops 3) closing grocery stores and restaurants and providing delivery instead. For all scenarios, we observe two areas where most infection happens: hubs, where people from different occupations can meet (e.g., restaurants), and non-hubs, where people with the same occupation meet (e.g., offices). The simulations show that most interventions reduce the total number of infected agents by a large margin. We observed that interventions targeting hubs (2-4) did not impact the infection at non-hubs. In addition, the intervention targeting people’s behavior (1) ended up creating a cluster at the testing center.

[Fairly Dividing Mixtures of Goods and Chores under Lexicographic Preferences](https://www.southampton.ac.uk/~eg/AAMAS2023/pdfs/p152.pdf)
Hadi Hosseini *(Pennsylvania State University)*
Sujoy Sikdar *(Binghamton University)*
Rohit Vaish *(Indian Institute of Technology Delhi)*
Lirong Xia *(Rensselaer Polytechnic Institute)*

[Approximation Algorithm for Computing Budget-Feasible EF1 Allocations](https://www.southampton.ac.uk/~eg/AAMAS2023/pdfs/p170.pdf)
Jiarui Gan *(University of Oxford)*
Bo Li *(Hong Kong Polytechnic University)*
Xiaowei Wu *(University of Macau)*

[Efficient Nearly-Fair Division with Capacity Constraints](https://www.southampton.ac.uk/~eg/AAMAS2023/pdfs/p206.pdf)
Hila Shoshan *(Ariel University)*
Noam Hazon *(Ariel University)*
Erel Segal-Halevi *(Ariel University)*

***

Parceling EAAMO Paper:

**Conclusion of the discussion:** Given that each component by itself does not make sense, and the ruin route would not really give much weight in a AI conference such as AAMAS where the reviewers being economists interested in the concept of ruin low, we decided that rather than re-parceling the paper, re-organizing it based on how the previously accepted papers were packaged would be the most economical approach.

What do we do in the paper?

- Evaluate existing models and their limitations
- Introduce a new model to capture realistic behavioral features
- Simulate the new model and analyze the temporal behavior and the effects on precarity

One possible way we can re-parcel this paper is to break it into two. First, group the first two parts and introduce the new model as a possible model to develop simulations. In terms of validation, we might need to figure out cases where the behavior of existing methods are incorrect and then try to showcase how our model behaves in those cases as a comparative study thus validating the simulation model. I don't really know the feasibility of this though.

The follow-up paper would be what we have been discussing where we do the current experiments plus the AIES style experiments. In this, we will be exploring the introduced model. The problem with this is just by itself, it might not be a proper paper. Maybe the better thing to do is to parcel the first part out as a conference paper and then combine the new experiments and maybe extend it to a journal paper or something if that is possible. With the new experiments, we should have enough changes for the resubmission to the journal to be acceptable. We can even try to extend the limitations and math sections if possible for the journal part.

***

Place in the line of things:

**Conclusion of the discussion:** This paper by itself, as long as we re-organize things and re-package it, should be good for AAMAS. If we try to make the paper with Aditya the one before this, then it would constrain us to make sure things are linked and because we have not really started anything proper in that, we do not currently have a proper idea where that might lead. So, it might be better to try this for AAMAS.

If we are using existing methods and exploring precarity in an algorithmic manner the paper with Aditya might need to come in front in the hierarchy. The merit from that paper to this one would be if we manage to link things to precarity and ruin, wherein we can say that precarity and ruin are interesting phenomenon that require depth of modeling. Without proper models that capture this it might be hard to make valid decisions. On the other hand, this does constrain us to make sure we do link things to precairty since unless we do that it will be hard to use that paper to justify this.

In regards to what fits where, AAMAS should be fine for either paper. But I honestly don't think this paper itself is a good fit for AISTATS. We can market the paper with Aditya for AISTATS if needed since it would have the theoretical framework of online learning + algorithms. A pure simulation paper might be more suitable for a pure AI venue or a more broad aspect venue.

***

Ruin agenda:

**Conclusion of the discussion:** The ruin agenda does not really matter unless the conference is going to have economists as reviewers. Given AAMAS is not specailized in that way, we decided not to pursue this.

We want to understand the way in which algorithmic decision making affects individual behavior and for that we need good models that capture consumption behavior and we also need models that help us capture an important part of precarity which is ruin. Building on top of our AIES paper and having gone through the econ literature we have found all these interesting elements we can add to the classic ifp that we'd use. On the same note, we relate this follow up paper on top of our AIES version by saying: to understand precarity we need to understand ruin (as a pre-requisite). This can be done either from a philosophical view point or by numerically saying which precarity index ranges are danger zones (for different income classes, etc.). The novelty of the paper is the design of this new ifp *as a side note* but as the main part, the "simulation as a sandbox" to explore things with different classifiers, populations, fairness measures, consumption profiles for 2 individuals, and even we can do proper ML classifiers (by synthetically assigning features to points and then doing experiments). So the insights we'd get with a ruin point of view of simulation (that we do not know as of now and could not do before with basic ifp) are the main contributions on top of the model design we did (to make the ifp in AIES paper better).

**Agenda:** As of now the main plan should be how to generate plots for ruin. Should it be just asset plots or something else, etc.

***

Change of focus might be necassary, that is:

We might need a title modification: I like this one for example: "Modelling Agent Decision Making in Agent-based Simulation - Analysis Using an Economic Technology Uptake Model" , here, we have keywords like agent decision making, simulation and economic uptake

Change of focus from precarity to agent behavior and decision making, along the same line, we have to change the focus of the abstract and intro. In particular, the first paragraph of the abstract needs to be something similar to the paper mentioned in the title change above, e.g., "Agent-based Simulation Modelling focuses on the agents’ decision
making in their individual context. The decision making details
may substantially affect the simulation outcome, and therefore need
to be carefully designed."

Or this one the first abstract paragraph in "Agent-Based Modeling of Human Decision-makers Under
Uncertain Information During Supply Chain Shortages" is: "In recent years, product shortages caused by supply chain disrup￾tions have generated problems for consumers worldwide. In supply
chains, multiple decision-makers act on uncertain information they
receive from others, often leading to sub-optimal decisions that
propagate the effects of supply chain disruptions to other stakehold￾ers. Therefore, understanding how humans learn to interpret infor￾mation from others and how it influences their decision-making
is key to alleviating supply chain shortages"

Note that, this second paper also has a very nice intorduction initial paragraph were they explain everything in the context of a tangible real world exmaple.

The uptake paper also has a very nice introduction starting point:

"The advantage of agent-based simulation comes from the possibility
to explicitly formulate the individual agents’ decision making in
its local context. This context may consists of a spatial, economic,
organisational or ecologic environment –, and of other agents to
interact with. A modeller takes the perspective of an individual
agent for capturing the agents’ decision making. What does the
agent perceive? What does the agent know at a particular point in
time? How does the agent decide which goal to pursue and which
action to take next? Finally the agent may evolve as well as change
its environment.
One of the challenges of developing agent-based simulation
models concerns the different perspectives that a modeller needs
to match during development: The agent’s ego-perspective versus
the population perspective often in form of aggregate values or
bird’s eye type of observations"

In the related work, we need more on the importance of simulation and less economicss-focused.

They have good pictorial and flow-chart exmaples. Also, less math in the main body like figure 1 in the shortages paper.
They also have some way of validating their propsed model/simulation which at this point I am not sure how to do that for our paper.

***

### Modelling Agent Decision Making in Agent-based Simulation - Analysis Using an Economic Technology Uptake Model

**Observations**:

- Since it is a straightforward modelling question the paper has a simple and straightforward abstract that covers the total paper in a few simple sentences.
- Introduction highlight the motivation properly. Since they are directly talking about agent-based modelling (which is the context of the conference), they do not need to introduce any concepts. They basically delve into why agent-based modelling is important thus motivating the problem. However, in our case we do need to introduce some concepts. One thing we could do is rather than starting with saying precarity is this, start with the motivation and say precarity is what captures this.
- The motivation is followed by the challenges and what is needed. This is similar to us saying to capture precarity we would need these things and etc.
- Clearly define all the features of what they explore.

### Using Agent-Based Simulator to Assess Interventions Against COVID-19 in a Small Community Generated from Map Data

**Observations**:

- Abstract starts with clearly stating the motivation. What they are doing is clearly stated in the abstract.
- Introduction also starts with clearly motivating the problem. And clearly states why the suggested method is a good approach.
- They highlight the limits of the existing systems in the intro itself. This though would be more of a personal choice. But we would also need to emphasize on the fact that there are limitations.

### Agent-Based Modeling of Human Decision-makers Under Uncertain Information During Supply Chain Shortages

**Observations**:

- Abstract and the intro clearly sets up the problem by showing what the motivation for the problem is and then introducing what they are planning to do to solve it. In some way the abstract has similarities to us in the sense they need to first establish the motivation and then state the necessity for the specific approach of understanding more realistic decision making paradigm.
- Related work follows a logical flow of reasoning than just establishing a group of different areas but not giving them a strong link to each other.