Cognition for action: an architectural account for “grounded interaction”

Abstract

The effects of priming are not limited to semantics but have also been witnessed in visual-motor tasks (Tucker & Ellis, 2001). By generalizing ACT-R’s (Anderson, 2007) existing spreading activation account to include visual representations and broadening the context within which associations are established, we have been able to replicate this small but reliable phenomenon both in simulation and embodied on a humanoid robotic platform. This model illustrates that the effect doesn’t require strict embodiment (e.g., Barsalou, 1999) but can instead be accounted for with abstract representations that are “grounded by interaction ” (Mahon & Caramazza, 2008).