Creating incentives to prevent execution failures: an extension of VCG mechanisms

When information or control in a multiagent system is private to the agents, they may misreport this information or refuse to execute an agreed outcome, in order to change the resulting end state of such a system to their benefit. In some domains this may result in an execution failure. We show that in such settings VCG mechanisms lose truthfulness, and that the utility of truthful agents can become negative when using VCG payments (i.e., VCG is not strongly individually rational). To deal with this problem, we introduce an extended payment structure which takes into account the actual execution of the promised outcome. We show that this extended mechanism can guarantee a nonnegative utility and is (i) incentive compatible in a Nash equilibrium, and (ii) incentive compatible in dominant strategies if and only if all agents can be verified during execution.