New Constructions of Mechanisms with Verification

A social choice function A is implementable with verification if there exists a payment scheme P such that (A,P ) is a truthful mechanism for verifiable agents [Nisan and Ronen, STOC 99]. We give a simple sufficient condition for a social choice function to be implementable with verification for comparable types. Comparable types are a generalization of the well-studied one-parameter agents. Based on this characterization, we show that a large class of objective functions μ admit social choice functions that are implementable with verification and minimize (or maximize) μ. We then focus on the well-studied case of oneparameter agents. We give a general technique for constructing efficiently computable social choice functions that minimize or approximately minimize objective functions that are non-increasing and neutral (these are functions that do not depend on the valuations of agents that have no work assigned to them). As a corollary we obtain efficient online and offline mechanisms with verification for some hard scheduling problems on related machines.