#113 — Consciousness and the Self

Episode Summary

Executive Summary: The conversation with Anil Seth explores consciousness from first principles: what it is, how it relates to perception, selfhood, sleep, anesthesia, dreams, animals, and theories like predictive processing and IIT. Seth argues for scientifically mapping phenomenology to mechanisms rather than trying to solve the hard problem head-on, and frames perception as controlled hallucination shaped by top-down inference.

Main Topics: Seth’s background and path into consciousness science (Priority: 5/5): He traces his route from physics to experimental psychology, then computer science/AI, artificial life, and finally consciousness research in San Diego with Gerald Edelman before returning to Sussex. Defining consciousness and the hard problem (Priority: 5/5): The discussion contrasts Nagel’s ‘what it is like’ definition with Chalmers’ hard problem, including skepticism about zombie arguments and concern that scientific theories may be asked for too much intuitive closure. Levels of consciousness: wakefulness, sleep, dreams, anesthesia (Priority: 5/5): Seth separates conscious level from wakefulness, noting dreaming, serial awakenings, deep sleep, general anesthesia, and anesthesia awareness as key cases for neuroscience. Perception as controlled hallucination (Priority: 5/5): He argues perception is not bottom-up recording but inference: the brain constructs experience by combining sensory input with prior expectations, yielding ordinary perception, hallucinations, and dreams along a continuum. Different contents of consciousness: objecthood, interoception, emotion (Priority: 5/5): Vision is object-based and action-oriented, while bodily feelings and emotions arise from interoceptive inference and allostatic regulation, explaining why inner experience feels less object-like. Selfhood, pure consciousness, and meditation (Priority: 4/5): The dialogue distinguishes self-consciousness from consciousness itself, explores purported contentless awareness in meditation and psychedelics, and questions how such states fit with information-based theories. Integrated Information Theory (IIT) and limits of current models (Priority: 4/5): They examine IIT’s claims that consciousness is integrated information, but Seth raises objections about testability, ontology of information, and whether the theory can account for phenomenology without becoming too abstract.

Key Arguments: Consciousness should be defined minimally as subjective experience or ‘what it is like’ to be a system, but technical definitions may evolve with science rather than precede it. The hard problem is powerful rhetorically, but zombie conceivability arguments are weak and become less persuasive as we understand mechanisms more deeply. A better scientific strategy is the ‘mapping problem’: relate specific phenomenological features to specific biological and computational mechanisms rather than trying to explain why experience exists at all. Conscious level is distinct from wakefulness: sleep, dreaming, anesthesia, and disorders of consciousness show that arousal and experience can dissociate. Perception is a controlled hallucination: brains infer causes of sensory input using top-down predictions; what we perceive is the best guess, not raw data. Visual objecthood depends on action contingencies and counterfactual predictions about how sensory input would change if we moved or interacted with objects. Interoception is the brain’s inference about internal bodily state; emotion is rooted in predictions about the allostatic consequences of events and actions. The self is not identical to consciousness; self-related experiences are only one subset of conscious contents and can vary or disappear without consciousness disappearing. IIT usefully emphasizes informational and integrative structure, but may struggle with empirical testability and with explaining phenomenology if it relies on highly abstract identity claims. Scientific accounts of consciousness should aim for explanatory bridges and predictive success, not necessarily a magical feeling of intuitive completeness.

Data Points: Research papers published: More than 100 - Seth’s publication record across neuroscience, AI, psychology, and consciousness research. Years at Neurosciences Institute in San Diego: About 6 years - He worked there after his PhD with Gerald Edelman before returning to Sussex. Years back at Sussex leading a lab: Last 10 years - He has been running the Sackler Center for Consciousness Science at Sussex. Consciousness research emphasis: 3 distinct dimensions - Seth discusses level of consciousness, contents of consciousness, and selfhood as separable heuristically useful targets. Axioms in current IIT: 5 - He notes the current version of Integrated Information Theory has five axioms/postulates. Time to lose awareness under anesthesia: Seconds to single-digit count from 10 - He describes counting backward and then instantly being back after general anesthesia. Sleep and awakening studies: Serial awakenings - He cites experiments waking sleepers to report contents of mind, showing even non-REM sleep can involve experience.

Pivotal Quotes: "Consciousness is the presence of any kind of subjective experience whatsoever." — Anil Seth: Used while defining consciousness in a broad, folk-level sense. "Perception in general works more from the top down or the inside out rather than from the bottom up or the outside in." — Anil Seth: Explaining predictive processing and the brain’s inferential role in constructing experience. "What we call reality is just when we all agree about our hallucinations." — Baba Brinkman (quoted by Anil Seth): Illustrating the controlled hallucination view of perception and shared reality.

Implications: The episode suggests consciousness science is moving toward testable links between brain mechanisms and experience, with major implications for anesthesia, psychiatry, AI, animal ethics, and meditation research.