On the commitments and precommitments of limited agents

Abstract

Rationality is an important concept in Artificial Intelligence and Philosophy. When artificial systems are considered to be intelligent or autonomous, it is almost obligatory to attribute intentions and beliefs to them. The currently dominant view of intentions sees them as involving commitments on the part of the agents who have them. But the notion of commitment seems to clash with the notion of rationality. It is argued that this need not be so. Commitments are only appropriate for agents with a limited capacity to reason. A treatment of commitment has been previously proposed that reconciles them with rationality. Here further motivations for the commitments of limited agents are discussed. This analysis is extended to account for the so-called precommitments, which have been excluded by others as introducing too much complexity.