AAMAS 2012 - June 4-9, 2012 - Valencia, Spain

Doctorial Consortium Abstracts

DC_9

People's cultural background has been shown to act the way they reach agreements in negotiation and how they fulfil these agreements. This paper presents a novel agent design for negotiating with people from different cultures. Our setting involved an alternating-offer protocol that allowed parties to choose the extent to which they kept each of their agreements during the negotiation. A challenge to designing agents for such setting is to predict how people reciprocate their actions over time despite the scarcity of prior data of their behavior across different cultures. Our methodology addresses this challenge by combining a decision theoretic model with classical machine learning techniques to predict how people respond to offers, and the extent to which they fulfil agreements. The agent based its initial strategy on a general model of the population in each culture, and adapted its behavior to its particular partner over time. This agent was evaluated empirically by playing with 157 people in three countries (Lebanon, the U.S., and Israel) in which people are known to vary widely in their negotiation behavior. The agent was able to outperform people in all countries under conditions that varied how parties depended on each other at the onset of the negotiation. This is the first work to show that a computer agent can learn to outperform people when negotiating in three countries representing different cultures. A Cultural Sensitive Agent for Human-Computer Negotiation Galit Haimhas 2 papers

DC_22

Stackelberg games have garnered signicant attention in recent years given their deployment for real world security, such as ARMOR, IRIS and GUARDS. Most of these systems have adopted the standard game-theoretical assumption that adversaries are perfectly rational, which may not hold in real-world security problems due to the bounded rationality of human adversaries and could potentially reduce the eectiveness of these systems. My thesis focuses on relaxing the assumption of perfectly rational adversaries in Stackelberg security games. In particular, I aim at developing new adversary models incorporating their bounded rationality and building new algorithms for eciently computing a defender's best response against these new models. To that end, I have developed a new adversary model using quantal response (QR) and a new ecient algorithm (Pasaq) to compute a defender's strategy against such a model in massive real-world security games. Experimental results with human subjects show that this new model gives signicantly better defender strategies than the previous leading contender. Furthermore, Pasaq has been deployed in a real-world security application, PROTECT, by the U.S. Coast Guards at the port of Boston. Recently, I started extending the model to incorporate features from more complicated games, including Network Security games and Bayesian Stackelberg games. Designing Better Resource Allocation Strategy against Human Adversaries in Security Games Rong Yanghas 5 papers

DC_23

There has been significant recent research interest in utilizing leader-follower Stackelberg game in security applications. Indeed, Stackelberg games are seen at many deployed applications: ARMOR at Los Angeles International Airport, IRIS for Federal Air Marshals Service, GUARDS for the Transportation Security Administration, and TRUSTS for the Los Angeles Metro Rail System (under evaluation). The foundational assumption for using Stackelberg games is that security forces (leaders), acting first, commit to a randomized strategy; while their adversaries (followers) choose their best response after surveillance of this randomized strategy. Due to the adversarial environment and the nature of law enforcement activities, many types of uncertainty, such as execution, observation, and preference uncertainty, must be taken into account in game-theoretic modeling for practical security applications. To that end, focusing on security games I explicitly model the aforementioned uncertainty and present theoretical analysis and novel algorithms for computing robust solutions. Furthermore, as the cornerstone in providing real world evaluations of my robust solution techniques, I propose TRUSTS, a compact game-theoretic formulation, for fare evasion deterrence in the Los Angeles Metro Rail system. In my future research, I will extend TRUSTS to address real world uncertainty and evaluate the solutions within the LA Metro system. Addressing Uncertainty in Stackelberg Games for Security: Models and Algorithms Zhengyu Yinhas 4 papers

 


Copyright © 2012 IFAAMAS