Effectiveness of Query Types and Policies for Preference Elicitation in Combinatorial Auctions

Combinatorial auctions, where agents can bid on bundles of items (resources, tasks, etc.), are desirable because the agents can express complementarity and substitutability among the items. However, expressing one’s preferences can require bidding on all bundles. We evaluate an approach known as incremental preference elicitation [3] and show that as the number of items increases, the amount of information required to clear the auction is a vanishing fraction of the information collected in direct revelation mechanisms. Most of the elicitors also maintain the benefit as the number of agents increases. We prove that randomization helps, in that no deterministic elicitor is a universal revelation reducer. Finally, we present a new query type that allows agents to use anytime algorithms to give approximate answers that are refined only as needed.

• Type: Conference
• Year: 2004
• Where: ATAL
• Authors: Benoît Hudson, Tuomas Sandholm
• Keywords: ATAL 2004, Direct Revelation Mechanisms, Incremental Preference Elicitation, Universal Revelation Reducer