Strategic voting when aggregating partially ordered preferences

Preferences of a single agent are often partially ordered. For example, it may be hard to compare a novel with a biography. In such a situation, the agent may want the novel and the biography to be considered incomparable. We consider here how to aggregate the partially ordered preferences of multiple agents in order to return a set of most preferred outcomes. We define the notion of strategy-proofness for such a scenario. This is when preference aggregation cannot be manipulated. We prove that if there is no dictator, agents can manipulate the result by voting strategically to determine the most preferred outcomes. This extends the well-known theorem by Gibbard and Satterthwaite for total orders. Categories and Subject Descriptors I.2.11 [Computing Methodologies]: Artificial Intelligence-Distributed Artificial Intelligence General Terms Theory Keywords Preference aggregation, strategy-proofness, partial orders, Gibbard Satterthwaite theorem

• Type: Conference
• Year: 2006
• Where: ATAL
• Authors: Maria Silvia Pini, Francesca Rossi, Kristen Brent Venable, Toby Walsh
• Keywords: ATAL 2006, Intelligent Agents, Partially Ordered, Preference Aggregation, Preferred Outcomes