Agents that acquire negotiation strategies using a game theoretic learning theory
✍ Scribed by Norberto Eiji Nawa
- Publisher
- John Wiley and Sons
- Year
- 2005
- Tongue
- English
- Weight
- 601 KB
- Volume
- 21
- Category
- Article
- ISSN
- 0884-8173
No coin nor oath required. For personal study only.
✦ Synopsis
Automated negotiation systems and real-world negotiation situations have many aspects in common. Time is a relevant factor for all parties; information about preferences is private, and there is no interest in having it disclosed; negotiators learn about the opponents and try to enhance their strategies while interacting with one another. Experiments were performed with computational agents employing a learning algorithm based on the ideas of the Experience-Weighted Attraction theory of learning in games, which has been shown to model well human behavior observed in experimental settings. Negotiation strategies are acquired as the agents play bargaining games against one another. The strategies determine the agents' behaviors: how much they offer to the opponent, when they make offers, and the conditions for accepting an offer. The results show that the learning agents were able to acquire sensible strategies even from the most unstructured and dynamic environments.