My goal is to help understand and assist decision-making. In particular, I am interested in situations where multiple (human, software or robotic) agents interact in dynamic and uncertain domains. My research spans many areas of multi-agent systems including:
Game-Theoretic and Data-Driven Human Behavior Modeling
This involves application and generation of mathematical models of rational behavior from data of actual human behavior in dynamic uncertain multi-agent environments where people may be cooperating or competing. Examples include analysis of the Social Ultimatum Game where we were able to identify issues such as reciprocity, multiple behavior types, the impact of moral values and societal stability. In theoretical domains, I have shown how one can create market mechanisms that can, in a decentralized manner, optimally allocate a divisible good, e.g., computational resources such as bandwidth and processing.
Spatiotemporal Pattern Recognition and Anomaly Detection
We have an ever-increasing ability to track and monitor spatiotemporal movements of agents from location-based tweets on a worldwide scale, to traffic patterns in a citywide scale and even sports data on a fine-grained scale. We investigate discovering patterns of behavior in this data, predicting outcomes from these patterns and detecting anomalous events before they emerge.
Automated Decision-Support, Tracking and Inference
This topic involves creating agents, tools and visualizations that help humans or human teams make better decisions in dynamic uncertain environments. For example, we were able to develop methods that significantly outperformed both state-of-the-art approaches as well as human teams in the DARPA Coordinators Program. This is aided by our agents' abilities to monitor data and infer states of the world releasing human operators from the need of feeding in information or performing lower-level reasoning operations.
Stochastic Decision Theory and Dynamic Decision-Making
I am interested in developing theoretical models that support decision-making in dynamic uncertain environments, employing techniques such as Partially Observable Markov Decision Processes and Graphical-Game-Based Approaches.
Interactive Multiplayer Games
Games are a rich and engaging research tool enabling data gathering for human behavior in web-based and mobile environments and for analyzing and creating automated decision-making approaches. They also are an increasingly valuable platform for serving as an experimental testbed for merging computational and social sciences.
STEM Education
I am involved is several efforts to use games, and alternate reality games in particular, to help engage students in science, technology, engineering and mathematics.
