Last updated: 2026-10-08

U
Undergraduate level

Does Intelligence Require Consciousness? Four Possible Relationships

This series opened with a chess engine that outperforms every human player while seeming, on most readings, experientially nothing like a mind at all. That description assumes an answer to a question this page finally asks directly: can the two come apart like that, or was the opening page simply wrong to assume a system could be behaviourally remarkable without anything being felt? Four distinct answers are on the table, each defended seriously by some tradition, and the question doesn't have one settled answer — it has four live candidates and a reason the dispute is hard to close.

Which Consciousness, Which Intelligence? FoundationalKnowledge that endures for decades — core principles

Most of the confusion in this debate comes from leaving both halves of the question unspecified. This series' page on Global Workspace Theory already introduces Block's distinction between access consciousness (information available for reasoning, report, and behavioural control) and phenomenal consciousness (there being something it's like to have it). "Intelligence" needs a matching split: call narrow competence the ability to perform a specific task well (a chess engine's play), and general intelligence the flexible ability to transfer reasoning across domains the system wasn't specifically built for. Access consciousness and general intelligence are plausibly linked by definition — both are about information being flexibly available rather than locked to one use. The genuinely hard question, and the one this page is actually about, is whether phenomenal consciousness bears any necessary relationship to either kind of intelligence at all.most of this dispute is really about which half of each pair is meant

Position One: Intelligence Requires Consciousness FoundationalKnowledge that endures for decades — core principles

On this view, what looks like reasoning without experience is never the genuine article — manipulating symbols correctly is not the same as understanding what they mean, and understanding, on this view, is inseparable from something being felt about the content. Searle's Chinese Room is the standard argument for exactly this: a system can produce every correct output a Chinese speaker would produce while nothing inside it understands Chinese, because syntax (symbol manipulation) never adds up to semantics (meaning something to someone)1. This series' page on intentionality develops the same argument in full. If genuine understanding requires experience, and genuine intelligence requires genuine understanding rather than its simulation, then a system with no phenomenal consciousness has, on this view, no real intelligence at all — only a competent impersonation of it.

Position Two: Consciousness Requires Intelligence FoundationalKnowledge that endures for decades — core principles

A smaller but serious tradition runs the dependency the other way. Higher-order theories of consciousness, already introduced on this series' Global Workspace Theory page, hold that a mental state is conscious only when a further, higher-order state represents it2 — which presupposes a system capable of representing its own states at all, a non-trivial cognitive achievement in its own right. On this view, consciousness isn't a raw feeling that simple systems might have in small amounts; it's a specific, demanding kind of self-directed representational capacity, and a system too cognitively limited to form higher-order states about itself cannot be conscious regardless of how much raw sensory processing it does. Pain behaviour in a very simple nervous system would not, on this view, imply anything is felt — the feeling requires the higher-order apparatus, and the apparatus requires a meaningful amount of what this page is calling general intelligence.

Position Three: The Two Are Independent FoundationalKnowledge that endures for decades — core principles

Chalmers' philosophical zombie — a being physically and behaviourally identical to a conscious person in every respect, producing every report a conscious person would produce, with nothing felt by anyone home — is offered as a conceivability argument that phenomenal consciousness cannot be identical to, or strictly entailed by, any physical or functional organisation, however behaviourally sophisticated3. If zombies are genuinely conceivable without contradiction, then a system could be arbitrarily intelligent — narrowly competent, and even generally flexible across domains — while being phenomenally empty throughout, which would make the two entirely separable in principle. Integrated Information Theory offers something closer to an actual, rather than merely conceivable, version of this split: it predicts that a purely feedforward network has zero integrated information (Φ = 0) by mathematical construction, regardless of how complex or behaviourally capable the function it computes is, while a much smaller recurrent system can have Φ > 04. On IIT's own account, a system's consciousness depends on its causal architecture, not on what its behaviour accomplishes — a sophisticated feedforward model and a trivial recurrent one are not even competing on the same axis. Independence also has a biological face: an octopus's cognition is distributed across a nervous system with most of its neurons outside a central brain, organised nothing like human cognition, and is nonetheless a serious candidate for rich sentience — intelligence and consciousness varying together in humans does not establish that they must.conceivable is not the same as actual

Note well. A philosophical zombie's conceivability is a claim about what's logically possible, not a claim about what's actually true of any existing system. IIT's zero-Φ result is doing much more specific, falsifiable work — a mathematical consequence of a particular architecture, not a thought experiment — and the two shouldn't be cited as though they were the same kind of evidence.

Position Four: The Two Are Bound Together FoundationalKnowledge that endures for decades — core principles

The functionalist case for a tighter connection runs through access consciousness specifically, not phenomenal consciousness directly. Global Workspace Theory ties conscious access to content winning broadcast availability across many specialised subsystems — and that same broadcast mechanism is plausibly what makes flexible, cross-domain reasoning possible at all, since a subsystem specialised for language can only draw on a subsystem specialised for spatial reasoning if something like a shared workspace connects them. On this view, the architecture that produces general intelligence and the architecture Global Workspace Theory associates with consciousness are not two separate things that happen to co-occur in humans — they are the same architecture, described from two angles. This is the position this series' Global Workspace Theory page develops, and it's worth being precise about what it actually ties together: a tight link between general intelligence and access consciousness does not, on its own, secure a tight link to phenomenal consciousness — the further step IIT and the zombie argument both put pressure on from the opposite direction.

Strongest Objection: The Four Positions May Not Be Answering the Same Question FoundationalKnowledge that endures for decades — core principles

A reasonable complaint about laying out four positions side by side is that they aren't actually rivals competing for the same prize — Position One is largely about semantic understanding, Position Two about a specific architectural requirement for a specific theory of consciousness, Position Three about phenomenal consciousness and physical/functional organisation, and Position Four mainly about access consciousness. Once "consciousness" and "intelligence" are both disambiguated as this page's second section insists on doing, some of the apparent disagreement dissolves into different authors answering different precise questions under the same loose heading. What survives disambiguation, and remains a genuine three-way dispute, is narrower: whether phenomenal consciousness specifically is entailed by, independent of, or a precondition for general intelligence specifically — Positions One and Three's actual disagreement, with Position Four better read as a claim about access consciousness that doesn't settle the phenomenal question either way.

Provisional Conclusion FoundationalKnowledge that endures for decades — core principles

Disambiguated carefully, this series takes no side between Positions One and Three on phenomenal consciousness and general intelligence — both remain live, seriously defended, and unsettled by current evidence. What is more settled is Position Four's narrower claim linking access consciousness to general intelligence functionally, and the chess engine this page opened with is best read as a case of considerable narrow competence with no established access consciousness and an entirely open phenomenal question — three separate verdicts, not one.

Questions for Further Thought

  • If IIT's zero-Φ prediction for feedforward networks is correct, what would it take to actually test it against a system's behaviour, rather than only against its architecture?
  • Does the octopus case genuinely support independence, or could octopus cognition simply be a different, non-human-shaped form of general intelligence that happens to accompany consciousness the same way human cognition does?
  • Which of the four positions would a current large language model's behaviour most directly bear on, and which would it leave completely untouched?

Further Reading

  • Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–457.
  • Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press.
  • Tononi, G., Boly, M., Massimini, M., & Koch, C. (2016). Integrated information theory: From consciousness to its physical substrate. Nature Reviews Neuroscience, 17(7), 450–461.

References

  1. Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–457. ↩

  2. Rosenthal, D. M. (1986). Two concepts of consciousness. Philosophical Studies, 49(3), 329–359. ↩

  3. Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press. ↩

  4. Tononi, G., Boly, M., Massimini, M., & Koch, C. (2016). Integrated information theory: From consciousness to its physical substrate. Nature Reviews Neuroscience, 17(7), 450–461. ↩