Simulating Realistic Social and Individual Behavior in Agent Societies By: Kenny Wong and Robin Jha Advisor: Yu Zhang Why? -Simulations becoming a useful tool for Social Scientists -Reproducing human decision making can help reproduce human behavior -Simulations would allow scientists to investigate situations that would be difficult to reproduce or require humans to be put in dangerous situations Results Random Dynamic Cluster Analysis For Measuring Social Convention -used to group objects, persons or concepts into homogeneous classes on the basis of their similarities -Local Clustering Coefficient -of a vertex in a graph quantifies how close its neighbors are to being a complete -Global Clustering Coefficient -gives an indication of the clustering in the whole network Memory -An agents memory is made up of all the interactions its been in. -We define the set of memories M by M=I*. -I* denotes a string of elements of I. Future Work -Use a new metric based on cluster analysis to predict social convergence in social systems -Use cluster analysis to predict if the system has reached a stable state. 70 75 80 85 90 95 100 0 50 100 150 200 250 300 350 400 450 500 Update Frequency Number of Cooperations 50 60 70 80 90 100 0 100 200 300 400 500 Random Small World Scale Free 50 60 70 80 90 100 0 100 200 300 400 500 Random Small World Scale Free 50 60 70 80 90 100 0 50 100 150 200 250 300 350 400 450 500 Restart Frequency Number of Cooperations 50 60 70 80 90 100 0 100 200 300 400 500 Random Small World Scale Free 60 70 80 90 100 0 50 100 150 200 250 300 350 400 450 500 Series1