Representation and Affect in Cognitive Science, Universitat de Barcelona

Teleosemantic theories aim to naturalise representational status and content by grounding them in the proper functions of components of cognitive systems. Yet teleosemantics inherits a substantive open question: which notion of proper function can do this grounding work across biological and artificial cases, while accommodating malfunction, dysfunction, and abnormal conditions without making function ascriptions arbitrary? This paper addresses that question by developing adequacy conditions on any function concept suitable for teleosemantics. The conditions are motivated by the canonical role of cognitive representations—tracking relevant environmental conditions and guiding behaviour with respect to them—together with familiar departures from that role (misrepresentation, missed representation, lucky success, failure despite correct representation, and systematic breakdown).

On the basis of these conditions, the paper proposes Etiological Service Functions (ESF). Very roughly, an item has the function to φ for a higher-level system when, in its recent etiological lineage, φ-ings occurred, were difference-making for the system’s successful behaviour, and thereby contributed to the stabilisation of the item (or its lineage). ESF combines a Wright-style “there because it did” idea with an explicit service relation to system-level success, in a way that is neutral between evolutionary selection, learning, and training. The payoff is illustrated via a small set of frog-style toy cases—evolved, learning, and AI systems—which show how ESF handles a wider range of teleosemantic targets than standard selected-effects approaches.