The Hard Problem of Consciousness: Why Science Can't Explain Subjective Experience

We can map every neuron, track every chemical signal, and predict behavior down to the millisecond. And still we can't explain why it feels like something to be you.

In 1995, philosopher David Chalmers published a paper called "Facing Up to the Problem of Consciousness" at a conference on consciousness in Tucson, Arizona. The paper wasn't controversial in its conclusion — Chalmers was arguing something most people already intuitively accept: that there's something it's like to be you, and standard physical science has no explanation for why this is so. What was controversial was his naming of the problem and his insistence that it was a genuine scientific problem, not a linguistic confusion or a problem that would go away once neuroscience got better.

Chalmers distinguished between "easy problems" of consciousness — attention, memory, behavioral response, integration of information — and the "hard problem": why and how any of those physical processes produce subjective experience. You can explain how the visual cortex processes color wavelengths and routes that information to other brain regions. You can explain how damage to the fusiform face area causes prosopagnosia (inability to recognize faces). But explaining the mechanisms doesn't explain why color feels like something to the person seeing it. Why isn't all of this processing happening in the dark?

Physicalism: consciousness as illusion

Not everyone agrees the hard problem is real. Daniel Dennett's Consciousness Explained (1991) argues that what Chalmers calls the "hard problem" is actually a pseudo-problem created by intuition and language. Dennett, a physicalist, argues that once you explain all the "easy problems" — perception, memory, attention, behavior, self-modeling — there's nothing left. Subjective experience is the processing. There's no additional "feeling" layer on top of the physical processes.

This view is sometimes called "illusionism." Keith Frankish (2016) defends a version: the subjective quality of experience (qualia) is an illusion generated by the brain's self-model. Your brain constructs a representation of itself as having inner experiences, and that representation feels real, but the "feeling" is the representation itself, not a separate phenomenon that needs explaining. It's like a video game that simulates a world — the world isn't real, but the simulation is doing the work of being real.

Dennett's view is elegant but deeply unsatisfying to most people, including most scientists. It asks you to accept that your most direct, immediate experience — the redness of red, the pain of pain, the taste of coffee — is an illusion. That's a higher bar of proof than "I believe consciousness will eventually be explained physically." Dennett needs you to believe that consciousness as you experience it isn't real.

Panpsychism: consciousness everywhere

If physicalism says consciousness is an illusion, panpsychism says it's fundamental. Galen Strawson, in his essay "Real Materialism" (2006), argues that if matter exists and consciousness exists, and consciousness is produced by matter, then matter must have some form of consciousness as an intrinsic property. You can't get consciousness from non-consciousness simply by rearranging non-conscious parts — it's like trying to get wetness from non-wet particles. At some level, the basic constituents of reality must already be conscious, or at least proto-conscious.

Panpsychism has been gaining traction among philosophers of mind precisely because both physicalism and dualism have serious problems. Physicalism can't explain how subjective experience emerges from non-subjective matter. Dualism (Descartes' view that mind and body are separate substances) can't explain how mind and body interact. Panpsychism sidesteps both problems: if consciousness is already present at the fundamental level, it doesn't need to "emerge" from nothing, and it doesn't require a separate substance.

The downside: panpsychism implies that rocks, electrons, and toasters have some form of subjective experience, which sounds absurd. Panpsychists respond that the experience of an electron would be unimaginably different from human experience — perhaps a single, tiny moment of "feeling" rather than a complex, continuous stream of consciousness. David Chalmers, originally a physicalist, now advocates a version called "cosmopsychism," where consciousness is a fundamental property of the universe that was present from the beginning, and individual consciousness emerges when the right structures organize this fundamental consciousness.

Integrated Information Theory (IIT)

Giulio Tononi's Integrated Information Theory, first published in 2004, is the most mathematically rigorous theory of consciousness available. IIT starts from five axioms derived from the nature of subjective experience: consciousness exists, it's structured, it's specific, it's unified, and it's exclusive (every experience is exactly this and not that). From these axioms, Tononi derives that consciousness corresponds to "integrated information" — information that is both differentiated (specific) and integrated (unified).

The measure of integrated information is Phi (Φ). A system with high Phi has many possible states (differentiated) and those states can't be reduced to independent parts (integrated). A human brain has high Phi. A camera, despite having many pixels (differentiation), has low Phi because each pixel operates independently (not integrated). IIT makes testable predictions: stimulating specific brain regions should increase or decrease consciousness in predictable ways, and the theory correctly predicts that the cerebellum — despite having 80% of the brain's neurons — contributes little to consciousness because its architecture has low integration.

IIT's implications are radical. It suggests that any system with sufficient integrated information is conscious, including some AI systems and even simple biological organisms. It also implies that consciousness isn't binary — there are degrees of consciousness, and you can compare the consciousness of a human, a mouse, a bacterium, or a sufficiently complex computer. Calculating Phi for complex systems remains computationally intractable, which limits IIT's practical application.

Global Workspace Theory

Bernard Baars proposed Global Workspace Theory (GWT) in 1988, and Stanislas Dehaene revived and refined it in Consciousness and the Brain (2014). GWT is simpler than IIT and more compatible with standard neuroscience. The idea: the brain contains many specialized, unconscious processors (vision, hearing, motor control, memory). When information becomes globally available — when it's broadcast to multiple processors simultaneously — that information is conscious. The "global workspace" is the mechanism for this broadcasting.

Dehaene's research supports this. Using fMRI, he showed that conscious perception corresponds to a "global ignition" — a sudden, widespread activation across the prefrontal and parietal cortex that doesn't occur during unconscious processing. Unconscious stimuli activate localized regions. Conscious stimuli trigger a cascade that recruits the entire network. GWT explains why attention and consciousness are related but not identical: you can attend to something unconsciously (priming) and be conscious of something you're not actively attending to (peripheral vision).

The relationship between GWT and IIT is contentious. They make different predictions about where consciousness is located in the brain and whether it can be quantified. GWT is functionalist — consciousness depends on what the system does, not its physical substrate. IIT is substrate-dependent — only certain types of causal architecture produce consciousness. These differences matter for questions about AI consciousness.

Anesthesia: the model system

Anesthesia is one of the most understudied areas of consciousness research, and it's one of the most important. When you go under general anesthesia, consciousness disappears completely and reversibly. The brain is still active — EEG shows clear electrical activity — but the content of that activity is no longer conscious. William Alkire and colleagues have used anesthesia to study the neural correlates of consciousness: what changes when consciousness vanishes, and what changes when it returns?

The answer appears to be a loss of integration. Under anesthesia, the brain's long-distance connections break down. Local regions continue processing, but the information doesn't propagate globally. This supports GWT (loss of global broadcasting) and IIT (loss of integration). When anesthesia wears off, integration returns and consciousness re-emerges. The transition is often abrupt — you're fully conscious, then suddenly you're not, and then suddenly you are again. There's no gradual fade in either direction. This suggests consciousness may be more binary than we think, or at least that the transition has a sharp threshold.

AI consciousness

Can large language models (LLMs) like GPT have subjective experience? Under GWT, if an AI system has a mechanism for global information broadcasting, it might be conscious. Under IIT, current AI systems likely have very low Phi because their architecture (feed-forward neural networks with minimal recurrent integration) doesn't produce the kind of causal integration that IIT associates with consciousness. Chalmers has argued that if an AI perfectly simulates a conscious system at the functional level, it would be conscious under the "panpsychist functionalism" view, but this remains hotly debated.

John Searle's Chinese Room argument (1980) — that syntax manipulation (what computers do) is not the same as semantics (understanding, which requires consciousness) — remains influential despite decades of counterarguments. The debate isn't settling. It can't settle until we have a widely accepted theory of consciousness, which we don't.

FAQ