Knowledge Media Institute, The Open University,
Milton Keynes, MK7 6AA, United Kingdom
Tel: +44 (0)1908 659834
Email: Charlie Hou
I am currently a full-time PhD student at the Knowledge Media Institute, The Open University, UK. My research interests include intelligent agents and multi-agent systems. My PhD research concerns online negotiation agents. I am exploring how learning techniques can be applied in order to predict an opponent's negotiation tactic.
Negotiation is a process of joint decision making between two or more parties in an effort to resolve their conflicting demands. Negotiation is central to the activities in e-commerce and distributed systems, here agents can be used to automate the negotiation process. My research focuses on the study of two-party negotiation where the two negotiation agents play opposing roles, such as a buyer and seller.
Agents require appropriate tactics in order to achieve an satisfactory outcome. A tactic is the decision policy for choosing actions in different states. Because negotiation is a strategic interaction, a negotiation outcome is determined not just by an agent’s own decision criteria but also by the other agent’s choices. This characteristic makes it difficult for an agent to find an optimal tactic.
I have proposed a learning approach to model an opponent's tactic online in order to achieve an optimal outcome. The learning mechanism applies nonlinear regression analysis to model a negotiation agent’s behaviour based only on the agent's previous offers. The behaviour of a negotiation agent in this study is determined by their tactics in the form of decision functions. Heuristics, based on estimates of an agent’s tactics, are drawn from a series of experiments. By applying nonlinear regression and the obtained heuristic knowledge, an agent can improve their overall performance by predicting the other agent’s deadline and reservation value, terminating pointless negotiation, and avoiding negotiation breakdown.
The findings of this study show that this learning approach can enable an agent to obtain better deals than the previously proposed decision function tactics. The learning mechanism can be used online, without any prior knowledge about the other agent and, therefore, is very useful in open systems where agents have little or no information about each other.
Example Publication:
Predicting Agents Tactics in Automated Negotiation, in Proceedings of the 2004 IEEE/WIC/ACM International Conference on Intelligent Agent Technology, IEEE Computer Society Press, 2004, pp.127-133 Modelling Agents Behaviour in Automated Negotiation, Technical Report KMI-TR-144. Knowledge Media Institute, The Open University, UK. May 2004.Example Presentation: Predicting Agents Tactics in Automated Negotiation, Knowledge Media Institute Internal Seminar. 19th April, 2004.
Last updated: 4 Nov, 2004