O. Khanolkar2, G. An1 1University Of Chicago,Surgery,Chicago, IL, USA 2University Of Illinois At Chicago,College Of Medicine,Chicago, IL, USA
Introduction: In 2016 Chicago experienced a near-historic level of intentional violence, and level that has unfortunately persisted; extensive data analysis of the various factors thought to affect violence was unable to explain why this occurred. This failure is due in part to the lack of a formal unifying framework that can integrate these multiple factors in a dynamic fashion. The fields of Game Theory and Behavioral Economics have provided frameworks to understand how population level phenomena arise from the behavior of individual actors. A classic Game Theoretic model is the Iterated Prisoner’s Dilemma (IPD), used to examine the evolution of cooperation and cheating (proxy for violence). A strategy termed “generous tit-for-tat” is the most evolutionarily successful approach to the IPD, which incorporates a parameter that essentially consists of forgiveness. We parameterize this model to incorporate socio-economic-racial factors into a conflict propensity/resolution matrix and implemented in a virtual population through agent-based modeling.
Methods: Computational agents utilized the “generous tit-for-tat” IPD strategy where the “generosity” parameter” (GP) adapted based on prior encounters. The IPD reward matrix varied across an environmental scarcity metric corresponding to economic development, and GP varied on “racial” group identification. Two classes of agents represented community members and police. All simulations were initialized with the same GP, run to dynamic equilibrium, and the following metrics collected: GP population distributions, # of cheating/conflicts. An additional set of experiments simulated conflict resolution behavioral modification by boosting the GP among targeted agents (both random individuals and among police).
Results: There were 4 main findings: 1) Scarcity drove adaptation to decreased GP and increased cheating/conflict; 2) Scarcity drove convergence of decreased GP between the community and police; 3) “Racial” group identification exacerbated convergence of decreased GP between the community and police; and 4) Behavioral Modification had a temporary effect and required re-application to have a global effect, consistent with reported clinical findings.
Conclusion: This initial abstract Game Theoretic computational model demonstrated plausible behavior consistent with real world observations. The diverging GP based on scarcity is consistent with the concept that “rational” behavior in the real world needs to be appropriately contextualized; the formal demonstration of the baseline similarity of the agents offers potential that such a framework can be used to build empathy for disadvantaged communities, and between those communities and police. Formal integrative methods such as this prototype offer the possibility of developing evaluative frameworks that can better parse the generative factors leading to violence and aid in the design, development and optimization of potential interventions.