Active inference unifies intentional and conflict-resolution imperatives of motor control

Antonella Maselli, Pablo Lanillos, Giovanni Pezzulo

Abstract: Author summary Most of our everyday actions are motivated by the "intentional" imperative to achieve some goal. However, several studies reveal that when we experience a multisensory conflict, such as for example during the rubber hand illusion or when we embody an avatar in virtual reality, we execute movements that are not intentional in the traditional sense, but aim to resolve the conflict (most often, without awareness). Nonetheless, the mechanisms of this second, "conflict resolution" imperative of motor control have received less attention so far. In this article, we advance a model of movement control grounded in the formal theory of active inference that integrates both intentional and conflict-resolution imperatives. Our simulations show that the same computational model can reproduce intentional and conflict-resolution components of action, or both simultaneously. Furthermore, these simulations highlight fundamental differences between movements that are guided by the two imperatives, which map to different error-correction mechanisms of active inference and might contribute to shed light on the reasons why movements guided by conflict-resolution are often executed without awareness.