University of Southern California
Powell Hall of Engineering 516
3737 Watt Way
Los Angeles, CA 90089-0781
Education:
A.B. (2005) degree in Economics from Harvard University.
Short bio:
Jason Tsai is a Ph.D. student at the University of Southern California, before which he completed his undergraduate studies as an Economics major at Harvard University and worked at the management consulting firm Oliver Wyman. He is currently a member of the Teamcore group, led by Prof. Milind Tambe, which focuses on multi-agent systems research with a use-inspired motivation.
His research interests are in interdisciplinary applications of multi-agent techniques, such as in security, transportation, finance, and social psychology. Past projects include game-theoretic resource allocation for the Federal Air Marshal Service(FAMS) as well as for urban network domains. His work for the FAMS, culminating in the IRIS software system, advanced the state-of-the-art in solving massive Stackelberg games and was recognized with a first prize at the third annual Department of Homeland Security research summit in 2009 as well as the Best Industry Track Paper award at the Autonomous Agents and Multi-Agent Systems conference in 2009. After extensive testing and congressional hearings, the IRIS program is now in operational use by the FAMS.
His current work uses social psychology to inform the design of agents that possess realistic emotional and social as well as cognitive components. Placing these agents within multi-agent simulations allows for more accurate predictions of crowd phenomena with current applications in security policy design/evaluation, architectural design/evaluation, and advertising. He continues to seek new applications of multi-agent techniques to create impactful, immediate solutions to real-world problems.
Honors
Best Paper Award, IVA-11, September 2011. (SIM) CREATE DHS Fellowship 2010-2013. Finalist, Excellence in Practice Award (EEPA), EURO-10, July 2010. (FAMS) Best Paper Award, AAMAS-09, Industry Track, May 2009. (FAMS) Best Student Poster Award, Third Annual Department of Homeland Security University Summit, March 2009. (FAMS) USC Annenberg Graduate Fellowship 2008-2010.
Current projects
Multi-Agent Simulation (SIM)
In evacuation simulation, existing work typically focuses on higher-level metrics and correctness. We hope to examine low-level behaviors using psychological frameworks to accurate model individual-level people in simulation. This, combined with compelling visualizations will allow for an enhanced training and policy-making tool for evacuation authorities.
We are also exploring other applications of Multi-Agent Simulations, such as in building design. Other ideas/collaborations welcome!
Past projects
Federal Air Marshal Service (FAMS)
An extension of the ARMOR project done for LAX, we apply the same principles of intelligent randomization to the assignment of Federal Air Marshals to flights. Research challenges are primarily focused around the complex scheduling constraints of the transportation network and much larger scale of the problem.
Randomized Security in Graph-Based Domains (GAPS)
Another extention of the ARMOR project done for LAX, we apply the fundamental principles used in ARMOR in a graph-based domain, such as an urban area. In such a setting, the roads can be seen as edges in a graph (with intersections being nodes). An adversary will seek to attack nodes in the graph by traversing from specified sources. Our (the defender) goal will be to place randomized personnel on edges (e.g., road blocks, checkpoints) to catch the adversary as he traverses the network. Research challenges focus on handling the increased size of the game (exponential in the size of the graph).
