The Syntax and Semantics of Goals in Cognitive and Computer Science

Summary

The paper treats goals in cognitive science and computer science as compositional cognitive representations that combine with world knowledge to organize purposeful behavior. Their content is linked to rational action, making goals comparable to representations studied through the syntax-semantics interface in linguistics and logic. The authors frame goal representation as a design space with at least three central dimensions: expressivity, design, and efficiency. A goal language must be expressive enough to describe behaviors of interest, while different representations may or may not encode the same behavior. This shifts attention from goals merely as fixed constraints on desirable behavior to constraints on the representations themselves. The paper synthesizes work on the properties of alternative goal representations and discusses how separating their form from their meaning can expose implicit assumptions about goals. It also argues that this distinction can clarify interactions between higher-level cognition and motivation, while providing axes for comparing different conceptions of goals.