As was mentioned on the opening page, this project was started to attempt to simulate the realities on the ground in Summit County through Agent Based Modeling to reveal nonobvious patterns and trends that are happening in Summit County, and work as an analytic tool for comparison to the realities of Summit County as part of the Healthy Summit Research Project Dr. Castellani and Dr. Hafferty are undertaking. The work displayed and discussed on this website served as a foundation for the computer model that we would later develop. Below is a brief summary of my work and how it helped advance our research.
Originally, we had chosen to use a computerized version of Schellings original experiment (for more information on Schelling and his work, click on the Schelling link above) as a start point for our research. This lead to my researching on Schelling and his work, and eventually the discovery of his original paper and coin model. Utilizing the basic concept of that model, I attempted to use the modifications we wanted to make and tried them out in paper form to see the results. They were astounding, and proved that our model our modifications produced meaningful results. Three agent types, rich, middle class, and working class/poor were arranged on an 8x8 grid and both middle class and working class/poor seeking rich, while rich seeking rich. After showing my results to the class in a brief presentation, we decided to continue with this avenue of research.
As a group we tried running the model in class to see what we might take from it. This lead to some methodological changes that improved the speed of executing the paper model, which more importantly lead to further insights on how the model worked and how we might modify it further to increase its rigor and representation of the realities in Summit County. The most important change was that we limited the movement of middle class and working class/poor, which lead to an increase in the interaction between the two groups. Another key change was the inclusion of a 50% movement/nonmovement rule as modeled in the fourth model on this website. This again increased the dynamics between working class/poor and middle class, and between each of these agent types within their own grouping.
We were feeling pretty confident after we explored this that we onto something. It seemed apparent now that we were modeling social mobility and its effect on the different agent types ability to increase their health and wellbeing. One final test involved my run of the modified model 4. This model brought many of the ideas that I observed from the different models together in a cohesive form that would go on to the foundation for the computer model; the movement limitations for middle class and working class/poor, the 50/50 movement/nonmovement rule, working class/poor seeking middle class, not rich, rich have only clustering rules (not movement rules in reaction to middle class/working class/poor encroachment), and the absolute movement rule.
It also raised a few questions that would go on to be answered by the computer model. One that had troubled me considerably during my research was how far middle class and working class/poor could "see" in order to seek rich or middle class. This was answered simply by giving middle class and working class/poor random movement when they were able to move, and thus sight was no longer a problem. Having them move in a random direction when their happiness rules are not met did not impede the clustering affects I had witnessed in the paper model and also comes closer to the trial and error approach agents take, as Robert Axelrod has discussed (The Complexity of Cooperation), as a simpler and just as robust and insightful way of modeling Agent Based Modeling.
This research, both its findings the questions raised provided a foundation for the computer simulation model of Summit County; which not only incorporated many of the findings, it also improved on the model as was noted in the above paragraph.
For more information on our computer model, please visit: http://www.personal.kent.edu/~mdball/Summit-Sim.htm