Information acquisition as a factor in improving agent performance in negotiation and decision making

University Of Maryland, College Park, College Park MD

Investigators

Abstract

The proposed research contributes to the development of autonomous agents capable of operating in real world situations. Agents in multi-agent environments must often choose between maximizing their expected utility according to their current knowledge, and trying to learn more about the world and thereby increase their gains. A negotiator may try to obtain information about its options by asking outside sources, taking actions, or by making offers and counter offers. In order to find the best negotiation strategy, agents' decision-making processes will be modeled. This research will lead the automated agent to reach more beneficial agreements when negotiating with bounded rational agents (human or automated). Strategies and decision-making procedures will be tested in the domains of e-commerce and international crises. The significance of the research lies first in the evaluation of alternate strategies for the acquisition of information by self-interested agents, from the perspective of a fully automated environment, and also by evaluating the behavior of automated agents interacting with human actors in simulated decision making environments. The dynamics of information acquisition in complex decision making structures will be better understood. The research will also contribute to the better understanding of decision making when automated negotiators interact with bounded rational agents.