Proceedings of the Annual Meeting of the Cognitive Science Society

We present a neural dynamic process model of an intentional agent that carries out compositionally structured action plans in a simulated robotic environment. The model is inspired by proposals for a shared neural and structural basis of language and action. Building on neural process accounts of intentionality we propose a neural representation of the conceptual structure of actions at a symbolic level. The conceptual structure binds actions to objects at which they are directed. In addition, it captures the compositional structure of action sequences in an action plan by representing sequential order between elementary actions. We show how such a neural system can steer motor behavior toward objects by forming neural attractor states that interface with lower-level motor representations, perceptual systems and scene working memory. Selection decisions in the conceptual structure enables the generation of action sequences that adheres to a memorized action plan.