Truthful and Quality Conscious Query Incentive Networks

Abstract

Query incentive networks capture the role of incentives in extracting information from decentralized information networks such as a social network. Several game theoretic models of query incentive networks have been proposed in the literature to study and characterize the dependence, of the monetary reward required to extract the answer for a query, on various factors such as the structure of the network, the level of difficulty of the query, and the required success probability. None of the existing models, however, captures the practical and important factor of quality of answers. In this paper, we develop a complete mechanism design based framework to incorporate the quality of answers, in the monetization of query incentive networks. First, we extend the model of Kleinberg and Raghavan [2] to allow the nodes to modulate the incentive on the basis of the quality of the answer they receive. For this quality conscious model, we show the existence of a unique Nash equilibrium and study the impact of quality of answers on the growth rate of the initial reward, with respect to the branching factor of the network. Next, we present two mechanisms, the direct comparison mechanism and the peer prediction mechanism, for truthful elicitation of quality from the agents. These mechanisms are based on scoring rules and cover different scenarios which may arise in query incentive networks. We show that the proposed quality elicitation mechanisms are incentive compatible and ex-ante budget balanced. We also derive conditions under which ex-post budget balance can be achieved by these mechanisms.