Cleeremans’s Learned Consciousness via Radical Plasticity

Cognitive psychologist Axel Cleeremans proposes that consciousness is not an intrinsic property of certain neural states but something the brain learns to do. His Radical Plasticity Thesis takes consciousness not as an intrinsic feature of particular neural states, but as an acquired capacity: the brain learns to represent, evaluate, and predict the consequences of its own representational activity. Through plasticity, first-order neural states become the targets of learned meta-representations that characterize their strength, stability, distinctiveness, reliability, and emotional significance. Phenomenal consciousness thus arises when a cognitive system has learned both to know that it occupies certain internal states and to care about those states (Cleeremans, 2008, 2011).

From the premise that neural activity is inherently unconscious, Cleeremans argues that the brain continuously and unconsciously learns to redescribe its own activity to itself, developing these metarepresentations that characterize and qualify their first-order targets; these learned redescriptions, weighted by acquired emotional value, constitute conscious experience. Phenomenal experience is therefore developmental rather than intrinsic: experiences occur only in experiencers that have learned they possess certain first-order states and have learned to care more about some than others. The mature statement, the self-organizing metarepresentational account, casts consciousness as the brain's unconscious, embodied, non-conceptual theory about itself.

Neural Activity Is Inherently Unconscious

Cleeremans begins where most theories end, from the deliberately strong assumption that neural activity is inherently unconscious. The explanatory question is therefore not which neurons are intrinsically conscious, but how an initially unconscious system acquires knowledge of its own processing. Brains continuously learn predictive relations among actions, environmental consequences, other agents, and activity in different neural regions. In doing so, they construct second-order redescriptions of first-order representations—a nonconceptual, implicit “theory” of their own internal states (Cleeremans, 2011).

In other words, rather than asking which neural states are conscious, he asks how any neural state could come to be conscious at all, treating nonconsciousness as the default condition of neural activity (Cleeremans, 2008; Cleeremans, 2011). The question becomes developmental and mechanistic: what must a system acquire before there is something it is like to be it? What requires explanation is then not a property switching on above some threshold of integration or activation, but a capacity built over time through the same learning mechanisms that produce skill and expertise elsewhere in cognition. This is the sense in which the thesis is radical: plasticity is not a modulator of consciousness but its generative condition.

Representational Redescription and Metarepresentation

The core mechanism is generalized from developmental psychology: representational redescription, whereby implicit knowledge embedded in a procedure is recoded into a format available for further operations (Karmiloff-Smith, 1992; Clark and Karmiloff-Smith, 1993). Cleeremans applies this inward. Networks learn not only about the world but about their own internal states, so that an agent comes to know something of the geography of its own representations. This can be modeled by a second network taking both the input and the internal representations of a first-order network as its own input, learning to predict and qualify them (Cleeremans, Timmermans and Pasquali, 2007; Pasquali, Timmermans and Cleeremans, 2010; Timmermans et al., 2012). Such metarepresentations also support anticipation: if the brain learns that supplementary motor activity reliably precedes action, it can represent that relation explicitly, and the result is experienced as intention.

Quality of Representation and Availability to Experience

Representations are graded along dimensions Cleeremans specifies as strength, stability in time, and distinctiveness. These determine both influence on behavior and availability to experience, yielding a three-way taxonomy: implicit representations, too weak for awareness or control yet able to influence behavior; explicit representations, strong enough for awareness and control, and the locus of subjective experience; and automatic representations, so strong that they escape control while remaining available to awareness. Because representational quality improves with learning, availability to consciousness has a developmental trajectory. Emotional value is constitutive rather than additive: metarepresentations are enriched by the value a system has learned to attach to its own states, which is why experiencers must have learned to care differentially about what they represent.

From Radical Plasticity to SOMA

The framework was later elaborated as the self-organizing metarepresentational account, drawing selectively on global workspace theory, higher-order theories, social theories of mind, and predictive processing (Cleeremans et al., 2020). SOMA (Self-Organizing Metarepresentational Account of consciousness) proposes that theory of mind, self-awareness, and perceptual awareness share prediction-driven learning mechanisms operating over three loops: an inner loop through which regions learn to predict one another, a perception–action loop, and a self–other loop through which learning about other minds and about one's own are mutually informing. Cleeremans has also pressed the reflexive corollary that consciousness unconsciously designs itself (Cleeremans, 2019). Commentators note that if learning is the cognitive basis of consciousness, substantial moral consequences follow for attributing experience to infants, animals, and artificial systems (Overgaard and Kirkeby-Hinrup, 2021).

Position, Objections, and Distinctness

The account is higher-order in structure but departs from standard formulations: the metarepresentations are learned rather than presupposed, non-conceptual rather than thought-like, and acquired through domain-general plasticity rather than supplied by an inherent, dedicated monitoring faculty. Zstill, there is no escaping the general higher-order difficulty of explaining why a representation of a representation should feel like anything. Critics press whether the account explains phenomenality or relocates it, and whether the appeal to learned emotional value smuggles in the qualitative character it was meant to derive. Cleeremans's later collaborative work on the functional value of phenomenal experience addresses a related but distinct question (Cleeremans and Tallon-Baudry, 2022). See Cleeremans and Tallon-Baudry's Functional Value entry, classified under Materialism: Phylogenetic/Evolutionary.

Assessment

The Radical Plasticity Thesis is unusual in making consciousness an achievement with a learning history, and in supporting that claim with implementable network models rather than argument alone. Its strength is mechanistic specificity and continuity with a large empirical literature on implicit learning and metacognition; its weakness, shared with arguably all theories, is the hard question of whether learned redescription delivers phenomenality or only a functional surrogate for it. The account seems to be developmentally and comparatively predictive in a way few theories are, which could make it unusually testable and unusually vulnerable.

References

Clark, A., and Karmiloff-Smith, A. (1993). The cognizer's innards: A psychological and philosophical perspective on the development of thought. Mind and Language, 8(4), 487–519.

Cleeremans, A. (2008). Consciousness: The radical plasticity thesis. In R. Banerjee and B. K. Chakrabarti (Eds.), Progress in Brain Research, 168, 19–33. Amsterdam: Elsevier.

Cleeremans, A. (2011). The radical plasticity thesis: How the brain learns to be conscious. Frontiers in Psychology, 2, 86.

Cleeremans, A. (2019). Consciousness (unconsciously) designs itself. Journal of Consciousness Studies, 26(3–4), 88–111.

Cleeremans, A., Achoui, D., Beauny, A., Keuninckx, L., Martin, J.-R., Muñoz-Moldes, S., Vuillaume, L., and de Heering, A. (2020). Learning to be conscious. Trends in Cognitive Sciences, 24(2), 112–123.

Cleeremans, A., and Tallon-Baudry, C. (2022). Consciousness matters: Phenomenal experience has functional value. Neuroscience of Consciousness, 2022(1), niac007.

Cleeremans, A., Timmermans, B., and Pasquali, A. (2007). Consciousness and metarepresentation: A computational sketch. Neural Networks, 20(9), 1032–1039.

Karmiloff-Smith, A. (1992). Beyond Modularity: A Developmental Perspective on Cognitive Science. Cambridge, MA: MIT Press.

Overgaard, M., and Kirkeby-Hinrup, A. (2021). Is learning the cognitive basis of consciousness? The moral implications of SOMA. Trends in Cognitive Sciences, 25(1), 8–9.

Pasquali, A., Timmermans, B., and Cleeremans, A. (2010). Know thyself: Metacognitive networks and measures of consciousness. Cognition, 117(2), 182–190.

Timmermans, B., Schilbach, L., Pasquali, A., and Cleeremans, A. (2012). Higher-order thoughts in action: Consciousness as an unconscious redescription process. Philosophical Transactions of the Royal Society B, 367(1594), 1412–1423.