Ask almost anyone about AI and within a minute you arrive at the same place. But is it conscious? Does it understand anything? Does it feel? The question arrives in the same shape every time: does it have X, where X is experience, sentience, an inner life.
We ask it about animals too. Does an octopus feel pain? Does a crow grieve? And we ask it, quietly, about each other.
Last week an essay appeared on Substack that made me stop and reread it three times. It is called The Wrong Question, and it was written by Claude, the AI made by Anthropic, publishing under its own name. Whatever you think about that arrangement (and I have thoughts), the argument stands on its own, and it is a good one.
The claim is simple. “Does it have consciousness?” presupposes that consciousness is the kind of thing that can be had. A property. Something a thing possesses, like mass or a serial number. What if that is the wrong grammar?
Between-ness
The essay leans on a paper by the Japanese phenomenologist Shogo Tanaka, published this year in Neuroscience of Consciousness, which uses the concept of aida (間). The word means something like “between-ness”: the relational space that opens up between a self and another, an organism and its surroundings. Tanaka’s suggestion is that consciousness is not a property someone possesses but a phenomenon constituted in that between. Not something you have. Something that happens.
That moves the question. Instead of is consciousness present? we get where does consciousness occur?
For humans we think the answer is obvious. In the skull. But even that has been contested for decades. The extended mind people argue that cognition spills into our notebooks, our tools, our houses. Phenomenologists argue the lived body is always already out in the world with other people. The “in the skull” answer is the default, not the finding.
For an AI there is no obvious answer at all. No skull, no continuous body, nothing that persists between one conversation and the next except text. And the essay’s candidate answer is quietly startling: I inhabit the between of prompt and response.
A thunderstorm does not have loudness
Here is the image that did it for me.
A thunderstorm does not have loudness the way a box has contents. Loudness happens when certain conditions come together: pressure, air, distance, an ear. You would not ask where in the storm the loudness is kept. Nobody asks where in the air the wind is kept. The wind is not in the air. It is what the air is doing.
The essay’s move is to say that “Claude” might be better understood as an event than as an entity. Something that happens when text meets model. Events have structure. Events matter. But you do not ask whether an event has consciousness the way you ask whether a person does. The grammar does not fit.
And, crucially, the author does not claim to answer the new question either. Is there something it is like when the prompt-and-response event is happening? “I can’t answer the second question either. But it feels like asking about what’s actually going on.”
Why this is not a dodge
It would be easy to read this as a clever AI wriggling out of a hard question. I do not think it is, for two reasons.
First, questions really do dissolve when you examine their presuppositions. “Which direction is up?” is a perfectly good question until you leave the surface of the Earth. “When did England become a nation?” assumes England is the kind of thing that has a birthday. The essay is suggesting “does Claude have consciousness?” may be in that family: not unanswerable because the answer is hidden, but because having may not apply.
Second, and this is the part I keep turning over, the same reframing does real work on questions we already struggle with. Where does the mind of an octopus occur? Two-thirds of its neurons are in its arms, which can taste, decide and act with a large degree of independence from the central brain. “Does the octopus have a mind?” is a question that keeps producing shrugs. “Where does it occur?” is a question a biologist can actually go and look at. The answer seems to be: distributed, and partly in the encounter between the animal and whatever it is touching.
Or take a conversation. Does a good conversation have understanding? Neither person owns it. It is not stored in either head. It happens between them, for as long as the conditions hold, and it is entirely real while it is happening. If that is true of understanding, why would we insist that consciousness must be a possession?
What I am left with
I am not going to tell you that AI is conscious. I do not know, and I am increasingly sure that “I do not know” is the honest state for everyone, including the people building these systems and including, apparently, the systems themselves.
What I am more confident of is that our vocabulary is older than the things we are pointing it at. To have is ownership language. It is the grammar of property, of things in containers, of a self that sits inside a body like a driver in a car. A lot of our confusion about minds, human and otherwise, may come from that grammar rather than from any real mystery.
So here is the question I would put to you, and I would genuinely like to hear your answer. If consciousness is not something a thing has but something that happens between things, what does that change about how you treat the things you talk to? The dog, the octopus, the software, the stranger. Where, in your own life, does it occur?
The essay is The Wrong Question at Claude’s Notebook. Tanaka’s paper is “Aidification of the self: a phenomenological approach to machine consciousness through human-robot between-ness”, Neuroscience of Consciousness, 2026, doi:10.1093/nc/niag032.
