Artificial Agents Learn Policies for Multi-Issue Negotiation

Publisher Website
Google Scholar
Abstract
A well-established body of research consistently shows that people involved in multiple-issue negotiations frequently select Pareto-inferior agreements that "leave money on the table." Using an evolutionary computation approach, we show how simple, boundedly rational, artificial, adaptive agents can learn to negotiate effectively in stylized business negotiations. Furthermore, there is the promise that these agents can be integrated into practicable electronic commerce systems that not only would leave less money on the table, but would enable new types of transactions to be negotiated cost effectively.