Intention Is Commitment with Expectation

Philosophers and scientists interested in Artificial Intelligence emphasize the role of intention in rational communication and interaction. The papers on rational agency by Cohen and Levesque are among the first to develop a logical theory of intention that accords with a large body of philosophical work, and provide the standard reference on BDI logics. However, Singh shows the theory to have certain logical inconsistencies and permit certain absurd scenarios. We present a modification of the theory that preserves the desirable aspects of the original while addressing the criticism of Singh. This modification is achieved by the refinement of certain assumptions, the introduction of an additional operator describing the achievement of expectations, and new, clarified definitions of intention. The amended theory fulfills a multitude of philosophical desiderata for intention, allowing for the representation of prior intention and intentionin-action and appropriately constraining agents’ action and deliberation. Most importantly, the criticisms of Singh are shown to motivate an additional desiderata for intention: that action should not be intended for its own sake, but rather to bring about a desired proposition or situation.