Event-driven Game Theory for Predicting Dynamics of Social Systems
Abstract
The purpose of this project is to develop new formal models to explain interactions between social systems and physical systems (with particular attention to cybernetworks Existing models arc primarily built using game theoretic strategies. A major shortcoming of these models is that they cannot predict human behavior over physical networks accurately, because game theory requires precise predictions of who moves when and full understanding of how the network will unfold. In reality, actors behave stochastically and as a result events often occur when they are, in fact, not expected. Another strategy of detecting such unexpected, anomalous events arc anomaly dectors, which attempt to identify unlikely events under normal behavior. This strategy, however, is also inefficient due to the high frequency of false positives that it generates. To address these limitations of existing formal models of unexpected behavior, the PI will integrate game theory with a stochastic process theory of master equations. This will generate a new class of formal models more amenable to capture difficult-to-predict social behaviors. The models will be validated using existing data testbeds that the PI has gathered in collaboration with researchers at Los Alamos National Laboratory. The Pl will generate a stochastic process theory of master equations that will be integrated into existing game-theoretic frameworks. This approach accounts for the fact that the timing of anticipated human behaviors is difficult to capture and that sequences of human behavior often appear random. Using this system of master equations, the Pl will build a model that can simulate the evolution of human behavior. The model will be validated against existing data testbeds on individual human behavior and refined in the second and third years to account for multiple humans (up to thousands) simultaneously interacting. The strategy for this will be Reduced Order Modeling strategies, adopted from aviation engineering, which can capture dynamical, simultaneous systems interaction. The models will be validated using existing data testbeds that the Pl has gathered in collaboration with researchers at Los Alamos National Laboratory.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 12, 2017
- Source ID
- W911NF1510127
Entities
People
- David Wolpert
Organizations
- Army Contracting Command
- Santa Fe Institute
- United States Army