Situated Neural Representations: Solving the Problems of Content.
Frontiers in neurorobotics January 1, 2022 Gualtiero Piccinini 61 citations
Situated approaches to cognition—emphasizing embodiment, embeddedness, enaction, and affect—are often seen as opposed to computational and representational views. This paper argues they are deeply intertwined: a neurocomputational account relying on neural representations not only is compatible with situated approaches but requires them at its core. Constructing neural representations with semantic content and learning appropriate computational processes demand tight dynamic interaction between nervous system, body, and environment. Situatedness enables neurocognitive systems to build representations with original semantic content, coordinate neural vehicles with content, make content causally efficacious and determinate enough, represent distal stimuli, and misrepresent. This suggests what is needed to build artifacts with basic cognitive capacities.