Final Project Topics
Below is a list of six possible final projects. You can, of course, decide to do another project, but please discuss it with us first. Any of the projects below should be doable with what you have learned this quarter. You can do your final projects in groups of two or three. When you turn in your project by email be sure that all of your names are included. The final project will include (1) Your project written using MASON and (2) a write up of the project. The write up should include: (A) an introduction that includes background information on your problem and what you investigated with your model; (B) a description of the model including (i) the space, (ii) agents, (iii) rules for how agents interact, (iv) key variables (e.g., presented in a table), (v) initial conditions for your simulations, and (vi) how you systematically investigated your model; (C) your results should include a written summary of what you found plus figures or tables illustrating your results; (D) a discussion of your results focusing on how they may shed theoretical light on the problems and issues you posed in the introduction. There are no specific page limits just include the requested points.
1. Evolution of Ethnocentrism
Reading: Hammond, R. A. and Axelrod, R. (2006). The Evolution of Ethnocentrism. Journal of Conflict Resolution, 50, 926-936.
Ethnocentrism is a syndrome of attitudes that typically include in-group favoritism. Under what conditions does in-group favoritism emerge in populations of social interacting agents? Answers to this question could provide theoretical insight into why ethnocentrism and specifically in-group favoritism is found across cultures. You can develop a model similar (but much better) than Hammond and Axelrod’s model using what you have learned this quarter. Within an evolutionary game-theoretical perspective, you could extend the evolutionary prisoner’s dilemma model developed in class to include (1) “tags” represented by color indicating different ethnocentric groups, (2) include rules for an agent interacting with agents of the same tag and agents of other tags, and (3) determine whether and under what conditions ethnocentrism evolves starting, for example, with all defectors (for both same and others) or all cooperators in a prisoner’s dilemma game.
2. Know when to walk away
Reading: Aktipis, C. A. (2004). Know when to walk away: contingent movement and the
evolution of cooperation. Journal of Theoretical Biology, 231, 249-260.
One classical approach to investigating the evolution of cooperation is to conduct round-robin tournaments with different cognitive strategies (e.g., tit-for-tat, TFT, when an agent always cooperates on the first encounter but then plays the strategy of its opponent on subsequent encounters). Aktipis showed that a behavioral strategy, walk away, can perform better than cognitive strategies such as TFT or PAVLOV under certain conditions. You could implement walk away with and without aggregation and analyze how cooperation could evolve in a prisoner’s dilemma game. You could also implement TFT and PAVLOV. Another possibility to consider is what if defectors walked away from other defectors?
3. Farming landscape patterns of use
Reading: Cailault, S. et al. (2012). Influence of incentive networks on landscape changes: A simple agent-based simulation approach. Environmental Modelling & Software, 45, 64-73.
This model aimed to investigate how different levels of organization affect diversity of landscape use in farming. In their model, they considered three types of landscape use numerically represented by 1, 2, 3 (e.g., corn, soybeans, alfalfa). In a more general model, one would want to be able to specify 2 or more. Farms occupy each cell in a grid, so a 50 x 50 space would have 2500 farms. Each farmer is associated with a single (though this could be changed to allow farmers to have multiple farms). Farmers decide how to use their land in a specific way depending on (1) the farm’s current use, (2) what neighboring farmers have done, (3) what is recommended locally, and (4) what is recommended globally. The global network (i.e., agent assumed to be composed of all farmers), recommends the least frequent landscape use. If there is a tie for the least used, no recommendation is made. The local network (i.e., the Moore neighborhood with a search radius of 1) recommends the most frequent land use. No recommendation is made for ties. The larger social network, also recommends the dominant land use, except when there are ties in which case no recommendation is made. Each land use type can only be used for x years before a farmer must switch to another land use type. A farmer gets the recommendations from the global, local, and social networks, and then can make a decision for the up coming year. If the current land use type has not exceeded the maximum age x, then all land use types are in the list of possible landscape uses otherwise the list is all of the possible land type uses minus the current. If the list came back (1, 2, 2), then the farmer would choose 2 if it is in the current list of possible landscape uses otherwise the farmer would continue with the current land use if it is not too old or choose randomly from the list of possible landscape uses. If there is no dominant land use type or no recommendation at all, then the farmer would continue with the current land use if it is not too old or choose randomly from the list of possible landscape uses. The main observational measure is the Shannon Diversity Index (SDI)
Where pi is the number of spatial units (farms) with land use type i and S is the total number of spatial units (farms).
4. Predator-prey dynamics
Reading: Thierry, H. et al. (2015). From the Lotka–Volterra model to a spatialised population-driven individual-based model. Ecological Modelling, 306, 287-293.
The authors used an agent based model to determine population dynamics similar to those described by the Lotka-Volterra model (coupled differential equations) of predator-prey dynamics. You could implement a somewhat different model in which predators and prey randomly move discretely in an environment. Prey could eat food items place in their environment or they could just be given a constant amount of food each round of play. Predators could search for prey within a search radius and have a probability, p, of eating a prey item. Predators and prey can reproduce if they have enough resources. You could include a cost of living parameter such that with each time step, they lose resources and can eventually die if food is not obtained. In short, you could build a predator-prey model and investigate whether predator-prey populations oscillate and under what conditions stable populations numbers are achieved over time.
5. Coupled Contagion Dynamics of Fear and Disease
Reading: Epstein, J. M., Parker, J., Cummings, D., Hammond, R. A. (2008). Coupled Contagion Dynamics of Fear and Disease: Mathematical and Computational Explorations. PLOS One, http://dx.doi.org/10.1371/journal.pone.0003955
A little over half way through the paper in Part II: Part II: Spatial Propagation in the Agent-Based Computational Model with Flight, the authors describe a simple model of contagion and fear with three types of agents (fleers, hiders, and ignorers) that respond differently to fear. You could implement this model to see if you can replicate their findings or modify their model as you think appropriate.
6. Social Identity Dynamics
Reading: Smaldino, P., Pickett, C. Sherman, J., and Schank, J. (2012). An Agent-Based Model of Social Identity Dynamics. Journal of Artificial Societies and Social Simulation, 15 (4) 7 http://jasss.soc.surrey.ac.uk/15/4/7.html
This is a simple model of optimal distinctiveness theory in which agents attempt to optimize their social identity. This model should not be too hard to implement and attempt to replicate the author’s results. It might also be interesting to look at agents that aggregate or avoid agents as a function of social identity.
7. Mate Choice
Reading: Smaldino, P. E. and Schank, J. C. (2011) Human mate choice is a complex system. Complexity, 17, 11-22.
You could further develop the Mate Choice model that we developed in class. For example, you could add agents with personalities. Personalities could be bold and shy. Bold individuals could move a lot to find dates while shy individual hardly move at all. Or shy individual might have a low probability of attempting to find a date. You could probably think of other personality types and model how personality affects behavior.