Human-Agent Negotiations: The Impact Agents’ Concession Schedule and Task Complexity on Agreements

Abstract

Employment of software agents for conducting negotiations with online customers promises to increase the flexibility and reach of the exchange mechanism and reduce transaction costs. Past research had suggested different negotiation tactics for the agents, and had used them in experimental settings against human negotiators. This work explores the interaction between negotiation strategies and the complexity of the negotiation task as represented by the number of negotiation issues. Including more issues in a negotiation potentially allows the parties more space to maneuver and, thus, promises higher likelihood of agreement. In practice, the consideration of more issues requires higher cognitive effort, which could have a negative effect on reaching an agreement. The results of human–agent negotiation experiments conducted at a major Canadian university revealed that there is an interaction between chosen strategy and task complexity. Also, when competitive strategy was employed, the agents' utility was the highest. Because competitive strategy resulted in fewer agreements the average utility per agent was the highest in the compromising–competitive strategy.