Item

Abstract

This talk introduces a method to investigate perceptual phenomena that is inspired by the neurophenomenological and enactive approach by Varela et al. (Varela, Thompson, Rosch 1991; Di Paolo, Rohde De Jaegher, forthcoming). It combines behavioural experiments and evolutionary robotics simulations of these experiments (Rohde 2008). The study of perceptual crossing and agency detection in virtual environments (e.g., Auvray, Lenay Stewart 2009; Di Paolo, Rohde Iizuka 2008) is presented as an example problem. The experimental data shows that agency detection in the given circumstances is a phenomenon that results from global interaction dynamics, not from individual recognition capacities, an interpretation that is confirmed, extended and enriched by results from the complementary robotic modeling. This research gives an example of how even simple embodied models can contribute to explaining human level cognitive capacities. The study is evaluated in the wider context of embodied perception research and the ongoing dynamic and embodied turn in Cognitive Science. In favour of this turn, it is argued that approaches focusing on simple parsing of structural patterns in the inputs are blind to processes of dynamical self-organisation of sensorimotor invariants, such as those underlying the results presented.