Leveled Commitment Contracts with Myopic and Strategic Agents

In automated negotiation systems consisting of selfinterested agents, contracts have traditionally been binding, i. e. , impossible to breach. Such contracts do not allow the agents to efficiently deal with future events. This deficiency can be tackled by using a leveled commitmentcontracting protocol which allows the agents to decommitfrom contracts by paying a monetary penalty to the contracting partner. Theefficiency of such protocols depends heavily on howthe penalties are decided. In this paper, different leveled commitment protocols and their parameterizations are empirically comparedto each other and to several full commitment protocols. In the different experiments, the agents are of different types: serf-interested or cooperative, and they can perform different levels of lookahead. Surprisingly, self-interested myopicagents reach a higher social welfare quicker than cooperative myopic agents when decommitment penalties are low. The social welfare in settings with agents that performed lookahead did not vary as muchwith the decommitmentpenalty as the social welfare in settings that consisted of myopicagents. For a short range of values of the decommitmentpenalty, myopic agents performed almost as well as agents that performed lookahead. In all of the settings studied, the best wayto set the decommitment penalties was to choose low penalties, but ones that were greater than zero. This indicates that leveled commitmentcontracting protocols outperform both full commitmentprotocols and commitmentfree protocols.

Authors

• Martin R. Andersson Tuomas W. Sandholm

Keywords

No keywords are indexed for this paper.