An Efficient Approach for Ordering Outcomes and Making Social Choices with CP-Nets

In classical decision theory, the agents’ preferences are typically modelled with utility functions that form the base for individual and multi-agent decision-making. However, utility-based preference elicitation is often complicated and sometimes not so user-friendly. In this paper, we investigate the theory of CP-nets (conditional preference networks) as a formal model for representing and reasoning with the agents’ preferences. The contribution of this paper is two-fold. First, we propose a tool, called RA-Tree (Relational Assignment Tree), to generate the preference order over the outcome space for an individual agent. Moreover, when multiple agents interact, there is a need to make social choices. But given a large number of possible alternatives, it is impractical to search the collective optimal outcomes from the entire outcome space. Thus, in this paper, we provide a novel procedure to generate the optimal outcome set for multiple agents. The proposed procedure reduces the size of the search space and is computationally efficient.