Negotiating Flexible Agreements by Combining Distributive and Integrative Negotiation
Quoc Bao Vo,Lin Padgham and Lawrence Cavedon
School of Computer Science and Information Technology
RMIT University,GPO Box 2476V, Melbourne,VIC 3001,Australia
This paper presents an approach to automated negotiation between agents which attempts to combine the advantages of a co-operative value adding approach,with the reality that negotiating agents are also competing.We use the concept of a trusted mediator to facilitate openness regarding what one values,without disadvantaging oneself by revealing sensitive information (such as a reserve price) to the other party.Social science and management literature deals with negotiation between people,and so can be both more complex,and less well defined than automated negotiation between software agents.We take inspiration from the social science literature and develop a computational framework to support negotiating software agents. The framework includes recognition that agents are self interested,and therefore will manipulate the system to their advantage if possible.We include mechanisms to discourage this kind of manipulation in the form of a transaction cost associated with making only small concessions, and a bias in dividing the pie which is the gain from trade which favours the agent who is most “honest”in making offers.
Many decisions in our everyday life are increasingly being made by computers in a relatively autonomous fashion. For instance,the routing of telephone calls and data packets in telecommunication networks has been controlled and decided autonomously by computers.Computers make decisions and control over how loads in electrical grids will be balanced at times of peak demand.Similarly research is being done on how computers can react to, and control, automative and airplane traffic in real time.In many cases,these decisions are generated as concerted efforts between several machines.