On Honesty in Sovereign Information Sharing

We study the following problem in a sovereign informationsharing setting: How to ensure that the individual participants, driven solely by self-interest, will behave honestly, even though they can benefit from cheating. This benefit comes from learning more than necessary private information of others or from preventing others from learning the necessary information. We take a game-theoretic approach and design a game (strategies and payoffs) that models this kind of interactions. We show that if nobody is punished for cheating, rational participants will not behave honestly. Observing this, our game includes an auditing device that periodically checks the actions of the participants and penalizes inappropriate behavior. In this game we give conditions under which there exists a unique equilibrium (stable rational behavior) in which every participant provides truthful information. The auditing device preserves the privacy of the data of the individual participants. We also quantify the relationship between the frequency of auditing and the amount of punishment in terms of gains and losses from cheating.