Mind, Mechanism, and Materialism: The Case Against the Computational Theory of Mind and Artificial General Intelligence, #5.

“Homo Machina (Machine Man),” by Fritz Kahn (Redbubble, 2025)

TABLE OF CONTENTS

1. Introduction

2. The Present Limits of AI: Empirical Considerations

3. Philosophical Arguments Against Artificial General Intelligence

4. Robert Hanna’s Systematic Challenge to Computational Mechanism

5. Neuroscientific Evidence Against Digital Computationism

6. Leading Theories of Consciousness: A Critical Analysis of Their Limitations

7. Quantum Mechanics and Consciousness

8. Conclusion

The essay below will be published in six installments; this installment, the fifth, contains section 6.

But you can also download and read or share a .pdf of the complete text of this essay, including the REFERENCES, by scrolling down to the bottom of this post and clicking on the Download tab.

6. Leading Theories of Consciousness: A Critical Analysis of Their Limitations

The scientific study of consciousness has proliferated numerous theoretical frameworks attempting to explain how subjective experience arises from neural activity. While each theory offers valuable insights, we will argue that the current leading theories, Global Workspace Theory, Integrated Information Theory, Higher-Order Thought theories, Predictive Processing approaches, Attention Schema Theory, and Quantum Mechanical theories of mind, all suffer from significant problems that prevent them from providing comprehensive explanations of consciousness. These limitations include computational intractability, explanatory gaps, definitional ambiguities, empirical inadequacies, and philosophical problems that collectively suggest the field remains at best in its theoretical infancy; at worst, untenable.

6.1 Global Workspace Theory and Global Neuronal Workspace Theory

Theoretical Framework

Global Workspace Theory (GWT), originally proposed by Bernard Baars (1988), and its neurobiological implementation, Global Neuronal Workspace Theory (GNWT) developed by Dehaene and Changeux (2011), represent the most empirically supported theories of consciousness. Consciousness occurs when certain neural signals dominate, spreading across the brain, becoming system-wide available.

According to GNWT, consciousness arises through sensory information obtaining access to a network of interconnected cortical areas, resulting in global integration and broadcasting of information. This produces distinct neural signatures, such as late-onset cortical activity (>300ms), gamma-band synchronization, and long-distance cortical connectivity patterns.

Empirical Support

GNWT has received substantial empirical validation through studies of binocular rivalry, masking paradigms, and attentional blink experiments. Empirical successes include various neural correlates of consciousness, including the P3b component in EEG, late cortical activity in fMRI, and patterns of frontoparietal activation during conscious perception.

Fundamental Limitations

Despite its empirical success, GWT/GNWT faces several fundamental limitations:

The Broadcast Problem: As argued by Ned Block (Block, 2019), the theory fails to explain why global broadcasting should necessarily give rise to subjective experience rather than merely sophisticated information integration. A zombie system, having no subjective experience at all, could theoretically exhibit all the neural signatures of global workspace activity without any accompanying phenomenal consciousness (Block, 2019).

Definitional Circularity: GWT defines consciousness in terms of global accessibility, but this creates a definitional problem of circularity. GWT assumes that global access equals consciousness, without providing independent criteria for determining when information integration becomes subjective experience, thus assuming what it is required to demonstrate.

Ignition Threshold Issues: The theory proposes that consciousness emerges when neural activity crosses an “ignition threshold.” However, GWT provides no principled account of what determines this threshold, or why crossing it should produce subjective experience rather than merely stronger neural responses, failing once more to account for conscious experience.

Graded Consciousness Problem: While GWT acknowledges that consciousness may be “graded,” it does not adequately account for partial or ambiguous conscious states. The all-or-nothing “ignition process” conflicts with phenomenological evidence for gradual transitions between conscious and unconscious processing; this evidence exists and cannot be explained away; therefore, GWT is shipwrecked on the problem of the vagueness and fuzzy boundaries of conscious states.

6.2 Integrated Information Theory

Theoretical Framework

Integrated Information Theory (IIT), was developed by Giulio Tononi (1999, 2015). To paraphrase Tononi’s own account for accurate characterization (Tononi, 1999), IIT proposes that consciousness corresponds to integrated information, with such information generated by a system being above and beyond its constituent parts. IIT quantifies consciousness using the measure Φ (phi), where a system with Φ > 0 possesses a degree of consciousness relative to its phi value.

IIT rests on five axioms corresponding to fundamental properties of conscious experience: (1) information, (2) integration, (3) exclusion, (4) intrinsic existence, and (5) composition (Tononi and Koch, 2015). The theory then postulates that physical systems underlying consciousness must satisfy these same properties mathematically.

IIT offers several theoretical advantages: it provides mathematical precision, addresses the hard problem directly by grounding consciousness in information-theoretic properties, and makes specific predictions about consciousness in various systems, including artificial agents and patients with disorders of consciousness.

Severe Limitations

IIT faces perhaps the most serious theoretical problems among theories of consciousness, as follows.

Computational Intractability: Computing Φ for realistic, non-trivial systems becomes computationally impossible. This limits the theory’s practical applicability and empirical testability (Doerig et al., 2021).

The Graining Problem: Critics argue that one can slice a system into parts in infinite ways, and IIT gives no clear rule for which divide counts when working out Φ. Different grainings can therefore yield radically different Φ values for the same system (Doerig et al., 2019). This raises an issue of internal consistency and coherence.

Circularity Concerns: IIT derives physical postulates from phenomenological axioms, then uses these postulates to explain phenomenology. This is logically circular: the theory assumes what it claims to explain. While Tononi argues this reflects IIT’s foundational approach, critics contend it undermines the theory’s explanatory power, and that Tononi in his response is begging the question.

Counterintuitive Implications: IIT implies consciousness in simple systems (photodiodes, networks of XOR gates), while potentially denying it in complex systems like feedforward neural networks. These implications conflict strongly with intuitive judgments about consciousness (Aaronson, 2014).

Pseudoscience Allegations: Over 100 prominent consciousness researchers signed an open letter in 2023 characterizing IIT as “pseudoscience,” arguing that its panpsychist commitments, consciousness in simple systems like photodiodes and lack of empirical testability, place it outside legitimate scientific inquiry (Doerig et al., 2021). The letter’s signatories argued that:

IIT makes untestable metaphysical commitments.

The theory’s panpsychist implications are scientifically problematic.

The theory lacks sufficient empirical validation.

IIT’s claims extend beyond what can be scientifically verified.

The IIT pseudoscience controversy represents a significant moment in consciousness studies, exhibiting fundamental problems, as follows.

Scientific Standards: What constitutes acceptable scientific theorizing in consciousness research?

Testability: Whether mathematical theories of consciousness must be immediately empirically testable?

Metaphysical Commitments: How much metaphysical speculation is acceptable in scientific theories?

Definitional Issues: What counts as consciousness and who determines this?

The Unfolding Problem (Doerig et al., 2021): The theory’s mathematical formalism can be “unfolded” to show that seemingly integrated systems may have zero phi, raising questions about the theory’s internal consistency.

While IIT faces legitimate technical and conceptual challenges, the pseudoscience label appears unjustified to many in the field, including us. The controversy ultimately reflects the immature state of consciousness science, where fundamental questions about methodology, definitions, and standards remain unresolved. As we see it, in this field metaphysical assumptions will be unavoidable. The real issue is whether these can be philosophically justified.

This debate serves as a case study in the sociology of science, demonstrating how theoretical disagreements can snowball into broader disputes about scientific legitimacy and professional authority in scientific fields, and about the role of philosophy, epistemology and metaphysics in these areas.

6.3 Higher-Order Thought Theories

Theoretical Framework

Higher-Order Thought (HOT) theories, developed primarily by David Rosenthal (2005), propose that consciousness arises when mental states become the objects of higher-order thoughts. Rosenthal holds that a mental state is a conscious one, if and only if it is accompanied by a suitable “higher-order thought” representing oneself as being in that state. Such thoughts are a type of “meta-thought” about thoughts. Relatedly, Higher-Order Perception (HOP) theories, developed by William Lycan (1996), offer a variant on HOT theories by proposing that consciousness requires higher-order perceptual monitoring rather than conceptual thoughts.

Theoretical Advantages

HOT theories provide a naturalistic account of consciousness that avoids mysterious non-physical properties. They explain the reflexive nature of consciousness and offer clear criteria for determining when mental states become conscious.

Critical Limitations

The Problem of Misrepresentation: According to critics of HOT, you could end up being “aware” of experiences you don’t really have, if your mind generates a mistaken higher-order thought. This leads to the counterintuitive conclusion that consciousness can exist without corresponding first-order mental states, a free-floating consciousness.

Infinite Regress Concerns: If consciousness requires higher-order thoughts, what makes those higher-order thoughts conscious? The theory seems to generate an infinite regress or must arbitrarily stop at unconscious higher-order representations, and in either case, fails to adequately account for conscious experience.

Empirical Inadequacy: Neuroimaging studies fail to consistently identify the kind of higher-order monitoring processes that HOT theories predict. The required metacognitive operations appear neither necessary nor sufficient for conscious experience.

Phenomenological Problems: Careful introspection reveals that conscious experiences don’t typically involve explicit higher-order thoughts about one’s mental states. The theory appears to intellectualize consciousness in ways that conflict with immediate phenomenology.

The Problem of Cognitive Sophistication: HOT theories seem to require conceptual sophistication that may be absent in animals, young children, or patients with certain brain lesions, leading to implausibly restrictive conclusions about the distribution of consciousness.

6.4 Predictive Processing Theories

Theoretical Framework

Predictive Processing (PP) theories, influenced by the work of Andy Clark (2013), Anil Seth (2014), and Jakob Hohwy (2013), propose that the brain is a prediction machine. Predictive processing views the brain as running layered simulations, constantly guessing incoming data, and revising itself when predictions fail.

Within this framework, consciousness corresponds to high-level predictions that achieve optimal precision-weighted prediction error minimization. Some variants propose that consciousness requires specific types of prediction error related to self-models or counterfactual processing.

Theoretical Strengths

PP theories integrate consciousness with broader frameworks in cognitive science and neuroscience. They provide mechanistic accounts of various conscious phenomena including perception, attention, and self-awareness, while offering potential explanations for altered states of consciousness and psychiatric conditions.

Fundamental Limitations

The Prediction-Experience Gap: PP theories, according to critics, fail to explain why prediction error minimization should give rise to subjective experience rather than merely accurate behavioral responses. The explanatory gap between computational processes and phenomenal properties remains intact.

Specificity Problems: All biological systems minimize some form of prediction error, yet consciousness appears only in certain complex systems. PP theories provide no principled criteria for determining which prediction-minimizing systems are conscious.

Circularity Issues: Many PP accounts of consciousness invoke concepts like “higher-order predictions” or “meta-cognitive predictions” that appear to presuppose the very consciousness they aim to explain.

Empirical Underdetermination: The PP framework is compatible with multiple mutually inconsistent theories of consciousness, making it difficult to derive specific testable predictions about conscious experience.

The Problem of Precision: PP theories rely heavily on precision-weighting mechanisms, but provide no account of why precision-weighted predictions should be experienced rather than merely computed.

6.5 Attention Schema Theory

Theoretical Framework

Attention Schema Theory (AST), developed by Michael Graziano (2013), proposes that the brain builds a simplified internal model of where its attention is directed, and that this modelling is what we call consciousness. Thus, according to AST, the brain constructs simplified, accessible models of complex attentional mechanisms, and consciousness corresponds to the brain’s access to these attention schemas.

The theory suggests that consciousness serves a social function, allowing organisms to model not only their own attention but also the attention of others, facilitating complex social interactions and communication.

Theoretical Advantages

AST offers a deflationary approach to consciousness that avoids mysterious properties while explaining why consciousness seems special from the first-person perspective. The theory connects consciousness to well-understood mechanisms of attention and provides an evolutionary account of the adaptive function of consciousness.

Significant Limitations

The Schema-Experience Problem: Even if the brain constructs schemas of attention, this doesn’t explain why having access to these schemas should involve subjective experience rather than mere information processing.

Reductionist Oversimplification: AST essentially claims that consciousness is an illusion, being a mere schema. This conflicts with the intuitive certainty that consciousness involves genuine qualitative experience, not merely information about experience.
Consciousness and Attention Dissociations: Empirical evidence suggests that attention and consciousness can dissociate, with attention possible without consciousness and vice versa. This undermines AST’s identification of consciousness with attention modelling.

The Problem of Qualia: AST provides no account of why attention schemas should involve specific qualitative properties like the redness of red or the painfulness of pain, rather than mere abstract informational content.

Social Function Limitations: While consciousness may serve social functions, this doesn’t explain why these functions require subjective experience rather than sophisticated information processing and behavioral responses.

6.6 The Hard Problem Remains Unsolved

Despite their diversity, all major theories of consciousness still leave us with major problems. They do not show how brain processes turn into experience; they do not agree on what consciousness even is, and their tests are highly uncertain. Even the more mathematical approaches become unworkable beyond simple examples.

Definitional Confusion

All theories of consciousness encounter fundamental disagreements about what consciousness is and what any adequate theory should explain. Some theories focus on access consciousness, others on phenomenal consciousness, and still others conflate or dismiss this distinction. This definitional confusion makes theoretical evaluation and comparison extremely difficult.

Empirical Inadequacy

All theories of consciousness rely heavily on neural correlates of consciousness (NCCs), but do not establish causal relationships between proposed mechanisms and conscious experience. The correlation-causation problem remains largely unresolved, and many theories make predictions that are difficult or impossible to test empirically.

The Problem of Other Minds

All theories of consciousness face fundamental challenges in determining consciousness in systems other than the theorist’s own mind. Whether assessing consciousness in animals, patients with brain damage, or artificial systems, theories provide conflicting predictions and lack independent verification methods.

Computational and Mathematical Limitations

Theories of consciousness attempting mathematical precision (particularly IIT), encounter computational intractability, while more qualitative theories lack the specificity needed for rigorous testing. This creates a dilemma between precision and practicability.

As the science of consciousness matures, its research agenda has expanded beyond neural correlates to focus on theoretical development, but this expansion has revealed the inadequacy of current approaches rather than resolving fundamental questions. Contemporary theories of consciousness, despite their individual insights and empirical contributions, all face severe limitations that prevent them from providing comprehensive explanations of conscious experience. These limitations include:

Explanatory gaps between proposed mechanisms and subjective experience.

Definitional ambiguities about the nature and boundaries of consciousness.

Empirical inadequacies in testing and verification procedures.

Computational intractabilities in formal approaches.

Philosophical problems regarding the mind-body relationship.

Therefore, the current state of consciousness science suggests epistemic humility rather than confidence in theoretical understanding. While each theory contributes valuable insights, none approaches a complete solution to the mystery of conscious experience. This seeming impasse has led some theorists to turn to quantum mechanics in order to formulate alternative theories of consciousness. In the next section, we will discuss the leading quantum mechanical theories of consciousness, and show their limits, but also how these positions, while also flawed, cut the epistemic ground from under physicalist reductionism, and mechanism.

Download