Introduction
Knowing the parts of a position is one thing. Assembling them is another. This reading builds one phenomenologically informed position statement, step by step, on a question that's widely debated: could a large language model be conscious? The example demonstrates a method. It isn't a verdict on the question, and other positions, built the same way, are equally defensible. This content reflects the field as of September 2026.
Locating the Study
The walkthrough draws on three sources for its map of positions. Each has a free route.
- The 2023 report by Butlin and colleagues, on AI consciousness and computational functionalism, is free on arXiv (Butlin et al. 2023).
- Anil Seth's biological view is set out in Being You (Seth 2021). The free SEP entry "The Neuroscience of Consciousness" gives background.
- Evan Thompson's enactive view is set out in Mind in Life (Thompson 2007). The free SEP entry "Embodied Cognition" gives background.
The Starting Point
Begin with the question as it's usually asked:
Could AI be conscious?
As stated, the question can't be answered. "AI" covers thermostats, chess programs, and chatbots. "Conscious" could mean responsive, intelligent, or experiencing. The steps below turn it into a position.
Walking Through the Method
Step 1: Sharpen the question
Fix the scope and the key term. Which systems? Current large language models: systems trained to predict text, running on digital hardware, without a body. Which sense of "conscious"? Phenomenal consciousness: whether there's something it is like to be the system.
Could a current large language model have phenomenal consciousness?
Notice what's been set aside: future systems, robots, and the question of whether such models are intelligent. Each is a separate question.
Step 2: Map the rival positions
Choose the two or three answers that actually compete, each at its strongest.
| Position | Answer | Main reason |
|---|
| Computational functionalist | Possibly, if it has the right computational organization | Consciousness depends on computation, whatever the substrate (Butlin et al. 2023) |
| Biological | No | Consciousness depends on a living body regulating itself (Seth 2021) |
| Enactive | No | Mind is continuous with life: a self-producing system for which things matter (Thompson 2007) |
The crux is whether consciousness depends only on what a system computes, or also on what it is: a living, self-maintaining body.
Step 3: State a claim
The worked example takes the enactive side. State the claim so that it could be wrong.
A current large language model doesn't have phenomenal consciousness, because phenomenal consciousness, as it's lived, is structured by a body engaged in sense-making, and such a model has no body whose continued existence depends on its own activity.
Check falsifiability. What would count against the claim? A convincing description of experience that doesn't involve bodily perspective or mattering. Or evidence that a system lacking any self-maintaining activity has the structures that the grounding names. The claim is exposed.
Step 4: Name the phenomenological grounding
Say which structures of experience the claim relies on, and link each to the claim.
- The lived body. Experience is given from a bodily here. Things are near or far, reachable or not, and the body is felt from within. A language model has inputs and outputs, but no here from which a world is given.
- Sense-making. Things show up as mattering to a living system that must keep itself going. A language model's outputs don't affect its own continued existence, so nothing is at stake for it.
The link: if these structures belong to phenomenal consciousness as it's lived, a system that lacks them lacks at least the kind of consciousness we know.
Step 5: State the strongest objection and a reply
Steelman the functionalist.
Objection. The claim confuses the conditions under which consciousness arose in animals with what consciousness requires. Living bodies may be one way to implement the relevant organization, but a different system could implement it without being alive.
Reply. Concede part of the objection: the claim doesn't show that no artificial system could be conscious. A future system with a self-maintaining, precarious body might qualify. But a current language model doesn't implement the organization the grounding names. The functionalist must say which computations suffice. Without that, the objection shows only that the question is open for other systems, not for this one.
Step 6: State the limits
Say what the position doesn't claim.
- It doesn't claim that no artificial system could ever be conscious.
- It doesn't solve the hard problem, or claim that the lived body explains why experience exists.
- It concerns current language models, as of 2026.
- It doesn't claim certainty. It gives reasons that others can test.
Key Considerations
Terms.
| Term | Role in the statement |
|---|
| Phenomenal consciousness | Fixes the sense of "conscious" (Step 1) |
| Computational functionalism | The main rival (Steps 2 and 5) |
| Lived body | Grounding: bodily perspective (Step 4) |
| Sense-making | Grounding: mattering to a self-maintaining system (Step 4) |
A common pitfall. The most common pitfall is a claim so hedged that nothing could count against it.
- Example. "Language models may lack some aspects of embodied experience in some sense." No evidence could show this wrong, so it says almost nothing.
- Test. Ask what would count against your claim. If you can't name anything, sharpen it.
- Hedging isn't the same as limits. State your claim plainly, then say what it doesn't cover.
Summary
The method moves from a loose question to a sharpened one, maps the rivals, states a falsifiable claim, names its grounding, meets the strongest objection, and states its limits. Here is the annotated model position statement:
Could a current large language model have phenomenal consciousness?[1] Computational functionalists say it could, if it has the right organization; biological and enactive views say no.[2] I hold that it doesn't, because phenomenal consciousness, as it's lived, is structured by a body engaged in sense-making, and such a model has no body whose continued existence depends on its own activity.[3] Experience is given from a bodily here, and things show up as mattering to a system that must keep itself going; neither structure has a counterpart in the model.[4] The functionalist replies that life is only one way to implement the relevant organization; I grant that a future self-maintaining system might qualify, but a current model doesn't implement what the grounding names.[5] I don't claim that no artificial system could be conscious, or that this solves the hard problem.[6]
- Precise question. Scope (current LLMs) and sense (phenomenal consciousness) are fixed.
- Map. Three rival positions, each with its answer.
- Falsifiable claim. A plain claim with a stated reason; evidence of experience without bodily perspective would count against it.
- Grounding. Two named structures, the lived body and sense-making, linked to the claim.
- Objection and reply. The strongest objection, steelmanned, with a partial concession.
- Limits. What the position leaves open.
This is one defensible position, not the right answer. A functionalist or biological position statement built by the same six steps would be just as good an example of the method.
References
- Butlin, Patrick, Robert Long, Eric Elmoznino, Yoshua Bengio, Jonathan Birch, Axel Constant, George Deane, et al. 2023. "Consciousness in Artificial Intelligence: Insights from the Science of Consciousness." arXiv:2308.08708.
- Seth, Anil. 2021. Being You: A New Science of Consciousness. New York: Dutton.
- Thompson, Evan. 2007. Mind in Life: Biology, Phenomenology, and the Sciences of Mind. Cambridge, MA: Belknap Press of Harvard University Press.