Problem-solving frameworks for sensible agents in an electronic market

Abstract

The need for responsive, flexible agents is pervasive in the electronic commerce environment due to its complex, dynamic nature. Two critical aspects of agent capabilities are the ability to (1) classify agent behaviors according to autonomy level, and (2) adapt problem-solving roles to various situations during system operation. Sensible Agents, capable of Dynamic Adaptive Autonomy, have been developed to address these issues. A Sensible Agent’s “autonomy level” constitutes a description of the agent’s problemsolving role with respect to a particular goal. Problem-solving roles are defined along a spectrum of autonomy ranging from command-driven, to consensus, to locally autonomous/master. Dynamic Adaptive Autonomy allows Sensible Agents to change autonomy levels during system operation to meet the needs of a particular problem-solving situation. This paper provides an overview of the Sensible Agent Testbed and provides examples showing how this testbed can be used to simulate agent-based problem solving in electronic-commerce environments.