Learning to Act Optimally in Partially Observable Multiagent Settings (Doctoral Consortium)

My research is focused on modeling optimal decision making in partially observable multiagent environments. I began with an investigation into the cognitive biases that induce subnormative behavior in humans playing games online in multiagent settings, leveraging well-known computational psychology approaches in modeling humans playing a strategic, sequential game. My subsequent work was in a scalable extension to Monte Carlo exploring starts for POMDPs (MCES-P), where I expanded the theory and algorithm to the multiagent setting. I first introduced a straightforward application with probably approximately correct guarantees (MCESP+PAC), and then introduced a more sample efficient partially model-based framework (MCESIP+PAC) that explicitly modeled the opponent.