Leveled commitment contracts with myopic and strategic agents

Abstract

In automated negotiation systems consisting of self-interested 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 commitment contracting protocol which allows the agents to decommit from contracts by paying a monetary penalty to the contracting partner. The efficiency of such protocols depends heavily on how the penalties are decided. In this paper, different leveled commitment protocols and their parameterizations are experimentally compared in sequences of multiple contracts. In the different experiments, the agents are of different types: self-interested or social welfare maximizing, and they can carry out game-theoretic lookahead or be myopic. Several meeting technologies, ways of setting the contract price, and ways of setting and increasing the penalties are compared. Surprisingly, self-interested myopic agents reach a higher social welfare quicker than cooperative myopic agents when decommitment penalties are low. The social welfare in settings with agents that perform lookahead does not vary as much with the decommitment penalty as the social welfare in settings that consist of myopic agents. In all of the settings, the best way to set the decommitment penalties is to choose low penalties, but ones that are greater than zero. This indicates that leveled commitment contracting protocols outperform both full commitment protocols and commitment free protocols. 1 1 An early version of this paper appeared at the National Conference on Artificial Intelligence (AAAI) ( Andersson and Sandholm, 1998b).