Learning the structure of utility graphs used in multi-issue negotiation through collaborative filtering

Graphical utility models represent powerful formalisms for modeling complex agent decisions involving multiple issues [2]. In the context of negotiation, it has been shown [10] that using utility graphs enables reaching Pareto-efficient agreements with a limited number of negotiation steps, even for high-dimensional negotiations involving complex complementarity/ substitutability dependencies between multiple issues. This paper considerably extends the results of [10], by proposing a method for constructing the utility graphs of buyers automatically, based on previous negotiation data. Our method is based on techniques inspired from item-based collaborative filtering, used in online recommendation algorithms. Experimental results show that our approach is able to retrieve the structure of utility graphs online, with a high degree of accuracy, even for highly non-linear settings and even if a relatively small amount of data about concluded negotiations is available