Researcher at a glance
- Affiliation: Sackler Centre for Consciousness Science, University of Sussex
- Orientation: Predictive processing / Bayesian brain; interoceptive inference; consciousness science
Key contributions
- Developed ‘interoceptive inference’ — emotions and the embodied self as active Bayesian inference on the causes of interoceptive signals (Seth 2013; Seth, Suzuki & Critchley 2011).
- Extended predictive coding from exteroception/action to interoception, with the AIC as comparator and autonomic control as active inference.
- Applied the framework to body ownership (cardiac RHI) and to psychopathology (depersonalization as imprecise interoceptive prediction; ‘conscious presence’).
- With Friston (2016), reformulated the model with the full active-inference/free-energy apparatus, added a cytoarchitectural argument (agranular ‘visceromotor areas’ lacking layer IV) for reading the interoceptive hierarchy as predictive rather than feed-forward, and extended it to autism (interoceptive trait prediction error) and depression/fatigue (dyshomeostasis, allostatic self-efficacy).
Key papers
- Seth (2013) — Seth (2013), Trends in Cognitive Sciences.
- Seth & Friston (2016) — Seth & Friston (2016), Phil. Trans. R. Soc. B.
- Seth, A. K., Suzuki, K., & Critchley, H. D. (2011). An interoceptive predictive coding model of conscious presence. Front. Psychol., 2, 395.
- Seth, A. K., & Critchley, H. D. (2013). Extending predictive processing to the body. Behav. Brain Sci., 36, 227–228.
Connections
Consciousness scientist who imported the Predictive coding / free-energy framework into interoception, producing the wiki’s most recent theoretical synthesis. See Seth (2013).
Central idea: interoceptive inference
Emotions are the integrated content of top-down interoceptive predictions counterflowing with bottom-up interoceptive prediction errors, computed in a salience network on the AIC/ACC. Autonomic regulation is Active inference (predictions enslaving autonomic reflexes, given attenuated precision on interoceptive errors). This generalizes Cognitive appraisal (of emotion) — “interpretation” of arousal becomes Bayesian inference. See Interoceptive inference.
Positioning against predecessors
Seth explicitly distinguishes his top-down/inferential view from the largely bottom-up hierarchies of Damasio, Craig, and Critchley — the crux of inferential?. (Critchley is a frequent collaborator and has moved toward the predictive camp.) His acknowledgements thank Sarah Garfinkel, Jakob Hohwy, Keisuke Suzuki, and — for shaping the ideas — Hugo Critchley and Manos Tsakiris.
Reach
Beyond emotion: body ownership (cardiac RHI with Suzuki), bodily self-consciousness, and a computational account of psychiatric disturbance (anxiety, psychosis, depersonalization) via aberrant/imprecise interoceptive inference. See Embodied selfhood.
Formalized with Friston (2016)
Seth & Friston (2016) restates the model with Karl Friston’s full free-energy formalism (stated at source in Friston 2005), names the AIC/ACC/SGC/OFC complex visceromotor areas (VMAs), and gives the predictive reading of interoception a structural argument (VMAs are agranular, lacking layer IV) rather than only a functional one. It also distinguishes epistemic from instrumental active inference — relevant to Is perceptual inference itself a genuine regulatory mode, or does the act of ‘choosing to let go’ collapse it back into active inference?, since the paper’s own two-mode taxonomy of active inference does not include a passive/perceptual third mode alongside them — and works out clinical extensions to autism (Interoceptive trait prediction error (ITPE)) and depression/fatigue.