On the trustworthy fulfillment of commitments

Abstract

An agent that adopts a commitment to another agent should act so as to bring about a state of the world meeting the specifications of the commitment. Thus, by faithfully pursuing a commitment, an agent can be trusted to make sequential decisions that it believes can cause an intended state to arise. In general, though, an agent’s actions will have uncertain outcomes, and thus reaching an intended state cannot be guaranteed. For such sequential decision settings with uncertainty, therefore, commitments can only be probabilistic. We propose a semantics for the trustworthy fulfillment of a probabilistic commitment that hinges on whether the agent followed a policy that would be expected to achieve an intended state with sufficient likelihood, rather than on whether the intended state was actually reached. We have developed and evaluated algorithms that provably operationalize this semantics, with different tradeoffs between responsiveness and computational overhead. We also discuss opportunities and challenges in extending our proposed semantics to richer forms of uncertainty, and to other agent architectures besides the decision-theoretic agents that have been our initial focus of study. Finally, we consider the implications of our semantics on how trust might be established and confirmed in open agent systems.