Approximately Optimal Mechanism Design

A social planner wants to optimize the choice of a social alternative vis-a-vis the players’ private information. We introduce a generic mechanism, [[ Moshe: the sentence is unclear ]] this is (almost) the only context such that almost optimal choice is guaranteed in [[Moshe: say strictly? ]] dominant strategies. In addition, this mechanism does not hinge on monetary transfers. We demonstrate the mechanism in two specific contexts – location problems and pricing of digital goods. The design of this mechanism involves a lottery between two mechanisms – With high probability we actuate a mechanism that makes players [[ Moshe: informationally small may be unclear ]] informationally small while being approximately optimal. The informational smallness ensures players cannot profit much by misreporting their true type. With the complementary probability we actuate a ’punishment’ mechanism [[Moshe: I’m not sure that the term punishment is not misleading; people may think about something much less sophisticated]] that provides strong incentives for being truthful.