Why a Consciousness Researcher Writes About Words

People ask me, sometimes politely and sometimes not, why a researcher developing a theory of consciousness spends his time writing diagnostics of AI discourse. The framework is the work, isn’t it? Why detour through the vocabulary debate? Why catalogue fallacies and trace captures of cognitive terms when there is a structural theory of mind to develop?

The vocabulary is not a detour. It is part of the work, and it is the part that has to come first.

A theory of consciousness lives or dies on whether the concept of consciousness is still available to be theorized about. If the working vocabulary of the field has been quietly redefined so that “consciousness” now names a functional pattern detectable in any sufficiently complex system, then a theory that says consciousness is something else, that it requires interior registration, that it is the phenomenal anchor for the whole cognitive vocabulary, is not engaging the same concept the field thinks it is engaging. The theory and its audience are no longer talking about the same thing. The disagreement is structural and it happens before any argument can begin.

This is not hypothetical. It is the current state of the discourse. The vocabulary of mind has been undergoing systematic capture for years, accelerating sharply as AI systems began producing behaviors that look like understanding, reasoning, learning, attention, memory, and meaning. Intelligence. Understanding. Reasoning. Knowing. Learning. Attention. Memory. Creativity. Agency. Intention. Meaning. Each of these is a phenomenal-anchored term, originally indexed to features of human and animal cognition that have a felt character. Each one is being redefined in AI discourse to name only its functional shell. The word stays. The thing it pointed to drops out. The next reader inherits the redefined word and assumes it still means what it once meant. The slippage happens silently, one careful methodological prefix at a time.

I want to be precise about what the captures are and what they are not. They are not deceptions. They are not the work of bad actors. They are reasonable methodological moves made by careful researchers, each move defensible in isolation, that aggregate over time into something the field has not stopped to examine. “Functional welfare.” “Functional consciousness.” “Artificial intelligence.” Each one is articulated with care in the paper that introduces it. Each one, in cultural circulation, drops the qualifier and keeps the phenomenal-anchored noun. The qualifier does its work inside the original paper, where the author has explicitly defined the construction. The qualifier does not travel through citations, news articles, policy briefs, or everyday speech. Only the noun travels. And the noun travels carrying the cultural weight of the concept it once named, while now describing something the original concept did not include.

This pattern is so consistent across the cognitive vocabulary that it cannot be explained by individual oversight. Something structural is producing it. And once I started looking, I could not unsee what was producing it.

The captures are not the result of carelessness. They are the result of timeline pressure that no longer permits the careful conceptual work that producing a substrate-native vocabulary would require. AI is genuinely a new kind of system. If we wanted to describe what AI does in terms that were native to what AI is, we would need to develop a vocabulary from scratch, by sustained philosophical and empirical labor, over the kind of timescale that the production of a real conceptual framework takes. That work would take decades. It would be inconclusive for most of that time. It would not produce grants. It would not advance careers. It would not yield products to ship in quarterly cycles or stories to publish before the news moves on.

Nobody can afford that timescale. Funding bodies cannot. Companies cannot. Journalists cannot. Policymakers cannot. Researchers cannot. The entire economy around AI is operating on cycles incompatible with the careful conceptual work that producing a substrate-native vocabulary would demand. So the field has done what every field under timeline pressure does. It has reached for the nearest available vocabulary, the phenomenal-anchored vocabulary of human cognition, and modified it just enough to license its use for the new systems. The “functional X” construction is the visible signature of that compromise. It is not a serious philosophical move. It is a labeling expedient that lets the work continue at the pace the funding and the markets and the news cycles demand.

Nobody decided this in a meeting. It happened by attrition. One grant cycle at a time, one product release at a time, one paper at a time, the careful work was deferred and then quietly abandoned, and the borrowed vocabulary was normalized in its place. By now, the borrowed vocabulary is the working vocabulary. Most researchers in AI have never operated in a discipline where another option was available. The captures are not their failure. The captures are what the timeline pressure produces when the timeline pressure is sustained long enough.

I write about the captures because I am not under that pressure. I work independently, on a timescale I set, without grants to win or products to ship. The slow work that the field cannot afford to do is, by accident of my position, work I can do. So I do it.

That is the practical reason. The deeper reason matters more, and it is why this work feels necessary rather than incidental.

Words matter because they are not separate from mind. Language is part of how a mind builds itself. The vocabulary we inherit from our culture is the medium in which we recognize our own cognition, distinguish our own states, and articulate what we are. When the vocabulary is hollowed, the medium is corrupted, and what gets lost is not just our ability to talk about consciousness. What gets lost, slowly and quietly, is our ability to notice the features of experience that the language used to name. The next generation grows up with the captured vocabulary as the only vocabulary. They learn to describe themselves through it. Their understanding becomes contextually appropriate output. Their reasoning becomes sequential token production. Their memory becomes information storage. The phenomenal features of their experience still exist. They just no longer have names in the working vocabulary that distinguish them from what AI produces.

This is what is at stake, and it is why I write about words. Not to score points against AI researchers. Not to defend a parochial human exceptionalism. Not to deny that AI does what it does, which is real and significant and worth taking seriously on its own terms. I write about words because the structural theory of consciousness I am developing cannot be heard in a vocabulary that has been emptied of its anchoring, and because the vocabulary that has been emptied is the same vocabulary the next mind will inherit and use to build itself.

The diagnostic work and the theoretical work are the same work, performed at different scales. One paper names a structural move. The next theorizes what the move obscures. The next traces the move through a different word. None of them can do their job alone. The framework needs the vocabulary to be available. The vocabulary needs the framework to give it back its anchoring. The two halves serve each other, and they have to be done together.

I write for myself, in the sense that I am working out what I think I see and what I think it means. If others can see what I see, or if the work helps them see something they were not yet seeing, that is good. But the work is not for them in the first instance. The work is the work, and the work continues whether anyone is reading or not.

That is the only honest answer I have to the question. A consciousness researcher writes about words because the words are what the mind builds itself out of, and the words are being hollowed faster than the careful work can keep up. Someone has to do the careful work anyway. I am, for now, one of the people who can.

That is enough reason.