Interoceptive inference, emotion, and the embodied self

Anil Seth is a cognitive neuroscientist at the University of Sussex, where he co-founded and now co-directs the Sackler Centre for Consciousness Science with a joint appointment in the School of Engineering and Informatics. He came out of Gerald Edelman's group at the Neurosciences Institute in San Diego in the early 2000s — Edelman's selectionist theory of consciousness is part of his intellectual lineage — and his programme since has applied Karl Friston's free-energy principle and predictive-processing framework to subjective experience and emotion. His 2021 trade book Being You (Faber & Faber / Dutton) extends the argument to a general audience.

Published in Trends in Cognitive Sciences in November 2013, the piece is a theoretical review — a position paper, not a primary report. The argument is interoceptive inference: that subjective feeling states are not direct readouts of body state but rather actively-inferred predictions generated by the brain about what its interoceptive afferents should be saying, with prediction errors driving learning and updating. The frame is Bayesian-brain and predictive-processing throughout, adapted from the perceptual- inference literature (Friston, the Helmholtz lineage) to the body-state-monitoring domain. The conceptual rival is Craig's direct-readout model, in which the right anterior insula re-represents homeostatic afferent state without strong top-down prediction; Seth's frame inverts the directionality and gives top-down predictions the active role.

The piece sits as the founding theoretical statement of the predictive-processing approach to interoception, and is cited by nearly every subsequent paper that takes the inference frame. The empirical follow-up has come from Seth's own group at Sussex, from Critchley and Garfinkel (also at Sussex), and from Klaas Stephan's group at Zurich, who have built computational models of interoceptive inference and tested them against fMRI and behavioural data. The 2018 Roadmap places this frame on equal footing with Craig's in its consensus statement; Barrett and Simmons 2015 is the empirical-anatomical companion in Nature Reviews Neuroscience a year and a half later.

The stake is theoretical. Seth has built a research programme on the inference frame and has skin in the game of its success; the trade book and a widely-watched TED talk (Your brain hallucinates your conscious reality, 2017) give the frame a popular life beyond its empirical base. The TICS paper itself is honest about being a theoretical proposal awaiting empirical test; the popularisation is less circumspect. Read the journal piece for the careful version of the argument; chase the Sussex / Zurich empirical follow-ups for what has and has not been demonstrated.