Using and evaluating adaptive agents in electronic commerce negotiation

Abstract

Agent technology has been applied to the Electronic Commerce domain, giving birth to what is known as agent-mediated Electronic Commerce. Current real-world applications refer only to the delegation of product or merchant brokering tasks to software agents. Automated negotiation is a less explored stage in this field, since it implies the trust of bargaining power to software agents. We here present SMACE, a layered platform for agentmediated Electronic Commerce, supporting multilateral and multi-issue automated negotiations. In this system, the negotiation infrastructure through which the software agents interact is independent from their negotiation strategies. SMACE has been used to test several negotiation strategies. The system includes agents that are capable of increasing their performance with their own experience, by adapting to the market conditions. This adaptation is reached through the use of Reinforcement Learning techniques. In order to test the agents' adaptation process, several different experiments have been tried out, and the respective results are here reported. These results allow us to conclude that it is possible to build negotiation strategies that can outperform others in some environments. In fact, knowledge gath