Neurodynamical Account for Altered Awareness of Action in Schizophrenia: A Synthetic Neuro-Robotic Study

Abstract

We hypothesize that altered awareness of action in schizophrenia may arise from disturbance of the forward model originating in functional disconnection in a hierarchical neural network. The proposed idea was tested through a neuro-robotic experiment using a hierarchical neural network model connected to a humanoid robot interacting with a physical environment. The results demonstrate that not only top-down forward dynamics, but also bottom-up regression processes driven by prediction error are important mechanisms for flexible adaptation to unpredictable changes in environment. In the simulated functional disconnection, in contrast to the normal condition, it turns out that this bottom-up regression process generates unnecessary modulatory signals which may induce altered awareness of action in patients. These results suggest that the proposed hypothesis may provide novel insight for understanding the pathological mechanisms of schizophrenia.