The pre-reflective experience of being a body is a fundamental component of consciousness and selfhood. A leading account from the perspective of active inference is Seth and Tsakiris’ (2018, see also Seth, 2021, 2024) proposal that this experience arises from instrumental interoceptive inference (III), namely the nervous system’s interoceptive predictions about the sensory consequences of autonomic adjustments that continuously sustain physiological integrity. However, this approach does not distinguish between the anticipatory autonomic regulation that keeps us alive even during unconscious states (e.g., dreamless sleep) and the interoceptive predictions that ground the conscious feeling of being a body that arguably allows a distinctly conscious form of adaptive behavior. We therefore suggest two types of III: conscious and unconscious instrumental interoceptive inference (C-III and U-III). After rejecting two potential ways to resist this distinction, and building on recent literature on active inference and consciousness, we propose that C-III can be distinguished from U-III neurocomputationally: C-III might additionally involve in-context, real-time modulation of the precision weighting of interoceptive prediction errors, coupled with temporally and counterfactually deep self-models enabling basic future-oriented adaptive behaviors. We conclude that differentiating C-III from U-III clarifies the border between felt bodily selfhood and insentient allostasis, and provides a tractable framework for future neurobiological and computational investigations on the dividing line between adaptive, minimal forms of self-consciousness, and behaviorally rigid self-regulatory unconscious states.
Ignacio Cea (Sat,) studied this question.