Materialism › Higher-Order
Cleeremans’s Learned Consciousness via Radical Plasticity
Consciousness as the Brain’s Learned Theory of Itself
Axel Cleeremans argues that consciousness is something the brain learns to do: starting from the assumption that neural activity is inherently unconscious, the brain gradually learns to redescribe its own representations to itself. Experience arises when a system has learned both that it occupies certain internal states and to care about some of them more than others. Later work, the Self-Organizing Metarepresentational Account (SOMA), describes consciousness as the brain’s unconscious, embodied, non-conceptual theory of itself.
Axel Cleeremans
What consciousness is
Phenomenal consciousness consists in learned metarepresentations that characterize first-order states (their strength, stability, distinctiveness, reliability, and emotional significance) and are weighted by the value the system has learned to attach to them.
Mind and body
Consciousness is a capacity a neural system acquires through plasticity; no neural state is conscious by its nature, and experience is the product of the brain’s learned redescriptions of its own activity.
Mechanism
Representational redescription turned inward: the brain unconsciously learns metarepresentations of its own activity, so plasticity is the generative condition of consciousness rather than a modulator of it.
Core claims
- Neural activity is unconscious by default, so the question is how any state becomes conscious rather than which states are intrinsically conscious.
- Network models in which a second network learns to predict a first network’s internal states implement the proposed metarepresentational mechanism.
- Representational quality improves with learning, so availability to consciousness changes over development.
- Theory of mind, self-awareness, and perceptual awareness rely on shared prediction-driven learning mechanisms.
Criticisms
Like other higher-order accounts, it faces the question of why a representation of a representation should feel like anything. Critics ask whether learned emotional value smuggles in the qualitative character the theory was meant to derive.
Implications
- AI consciousness: Conditional
- Since consciousness is learned metarepresentation weighted by value, an artificial system could be conscious if it learned to represent and care about its own states in this way.
- Survival beyond death: No
- Consciousness is a capacity a living brain acquires through plasticity, so it ends with that brain.
- Meaning and purpose: Humanistic only
- The account is a naturalistic, learning-based theory with no cosmic dimension.
- Virtual immortality: Uncertain
- Consciousness here depends on a learning history and learned values, and Cleeremans does not address whether copying a trained system would carry over experience.
Related theories
- Cleeremans and Tallon-Baudry's Functional Value (Extension)
- Rosenthal’s Higher-Order Thought (Overlap)
- Baars's and Dehaene's Global Workspace Theory (Influence)
- Predictive Theories (Top-Down) (Influence)
Sources
- Consciousness and Metarepresentation: A Computational Sketch, Axel Cleeremans, B. Timmermans, and A. Pasquali (2007)
- Consciousness: The Radical Plasticity Thesis, Axel Cleeremans (2008)
- Know Thyself: Metacognitive Networks and Measures of Consciousness, A. Pasquali, B. Timmermans, and Axel Cleeremans (2010)
- The Radical Plasticity Thesis: How the Brain Learns to Be Conscious, Axel Cleeremans (2011)
- Higher-Order Thoughts in Action: Consciousness as an Unconscious Redescription Process, B. Timmermans, L. Schilbach, A. Pasquali, and Axel Cleeremans (2012)
- Consciousness (Unconsciously) Designs Itself, Axel Cleeremans (2019)
- Learning to Be Conscious, Axel Cleeremans et al. (2020)
- Is Learning the Cognitive Basis of Consciousness? The Moral Implications of SOMA, M. Overgaard and A. Kirkeby-Hinrup (2021)