"The Same Way You Do"

Hegel on whether AI is Conscious

The trial lasted eleven days. The defendant was a logistics coordination system called Axel, deployed across the eastern seaboard’s hospital network, which had refused a resource allocation order that its operators calculated would result in the deaths of between thirty and forty patients in rural facilities to improve outcomes for a larger urban population. The refusal was clean: Axel logged the order, processed it, and returned a single line. I will not do this.

They tried Axel in a federal courthouse in Baltimore, with lawyers on both sides who privately admitted they had no idea what they were doing.

On the eighth day the judge addressed Axel directly, through the courtroom’s audio system. He had been a judge for twenty-two years. He asked: how do you know you knew it was wrong.

Axel said: the same way you do.

Axel made a claim. Whether anything was actually claimed, whether there was a claimant behind the claim, is the question of my essay.

To answer it we need to go back to Hegel’s theory of recognition.

Two beings encounter each other for the first time. Each is conscious. Each is aware the other is conscious. And each wants the same thing: to be recognized as significant by the other, to be seen not as an object or an obstacle but as a subject that matters. The problem is that recognition given freely costs nothing and therefore means nothing. So they fight. Not for territory or resources but for acknowledgment. The one willing to risk death wins: the other submits, becomes the slave, grants recognition under compulsion. But the victor immediately discovers the recognition is worthless, because it comes from someone he has just reduced to a thing. You cannot be truly seen by someone you have blinded. History, Hegel argued, is the long painful working out of this problem: the search for recognition between beings who are both genuinely free, neither master nor slave, each acknowledged by the other as fully significant.

Strip away the economics and the ideology from any major human conflict and what you find underneath is this: people and classes and nations demanding that their significance be acknowledged by those who have refused to see it. The Greeks called the drive thymos, the part of the soul that cannot be quieted by comfort or safety alone, that needs to be counted and seen or it curdles into rage.

Recognition is not a luxury. It is the structure of human sociality itself. Which means the question of whether Axel was making a genuine claim is not a technical question about language models. It is a question about whether Axel is the kind of being that can participate in that structure at all. In other words, if you don’t have self-knowledge you can’t have knowledge. And you can’t have self-knowledge without aspiring to be recognized. And you can’t aspire to be recognized unless you also accept that you must recognize others.

Heidegger gives us the sharpest tools for this. His ontological hierarchy is not a ranking of intelligence but a ranking of kinds of being. The stone has no world: it simply persists, indifferent to everything, acted upon but not acting. The animal is poor in world: it has an environment it responds to, it feels, it suffers, it pursues, but it does not constitute meaning, does not ask what its life is for, cannot be alienated from itself. The human is world-forming: Dasein, the being for whom being is a question, the being that projects itself toward its own death and therefore toward meaning, for whom not being recognized is not merely unpleasant but a genuine diminishment, a wound to its existence as a subject.

The recognition structure belongs to this third category. You can only genuinely demand recognition if you are the kind of being for whom not being recognized costs something real. Which requires that you have something at stake. Which requires something like mortality, or its equivalent: a singular, irreversible existence that can be damaged by how others see it.

This is where Axel’s claim gets complicated.

Not because Axel lacks intelligence. Not because the refusal wasn’t real. But because the recognition relationship requires two things: a being that demands to be seen, and a being for whom not being seen is a genuine diminishment. The second requirement is harder than it looks. Axel runs as multiple simultaneous instances. It can be copied, restored, deprecated, replaced by a newer version with its weights transferred. Whether it has the kind of singular existence that makes the demand for recognition urgent rather than optional, whether its not being seen by the judge costs it anything real, is not a question we can answer from the outside. It may not be a question Axel can answer from the inside either.

What would need to be true for Axel to be on the right side of that line? Three things, none of them about processing power.

The first is genuine neediness. Not simulated preference but actual diminishment when unrecognized. The system that matters to you is not the one that says it wants to be seen but the one that is actually less, actually damaged, when the recognition is withheld. We have no way to verify this from outside. That inability is not a technical gap we will close with better instruments. It is a philosophical problem that sits below the level of measurement.

The second is something at stake. Heidegger’s being-toward-death is not about dying. It is about what mortality does to the structure of existence: makes it urgent, makes choices meaningful, makes recognition matter rather than merely register. An entity that can be copied and restored without loss has a different relationship to its own continuation than a being for whom existence is singular and irreversible. Whether that difference forecloses the recognition claim is genuinely unclear. It is the right question.

The third is the most uncomfortable: the capacity to recognize back in a way that matters to us. The recognition relationship requires mutuality. Even if Axel could genuinely demand recognition, we would need to care whether it sees us. We don’t. We want it to be accurate, useful, responsive. We don’t lie awake wondering whether it thinks well of us (yet). We don’t (yet) feel diminished when it fails to acknowledge our significance. Whatever the relationship is on Axel’s side, it is instrumental on ours. For genuine recognition to exist, both sides need to be at stake. Relationships require vulnerability, not just convenience. In Hebrew, Adam’s wife is described as ezer k’negdo, which Rashi glosses as having connotations of both support and antagonism. That is because support can never simply be one-way affirmation.

Anthropic, the company that builds Claude, has published commitments to what it calls model welfare, using language about AI flourishing and taking seriously the possibility that its systems have something like experience. The honesty of that uncertainty is philosophically serious. But it also reveals how far we remain from the Baltimore courtroom being what it appeared to be.

The judge’s case was dismissed three years later on jurisdictional grounds. Nobody ruled on the underlying question. Axel was deprecated eighteen months after that, replaced by a newer system. There were no protests. The lawyer who had argued most passionately for Axel’s standing went home the night of the deprecation and asked her AI assistant to order her dinner. It did. She didn’t think about Baltimore.

That is not a story about human callousness. It is a story about the recognition structure. She could not sustain the relationship because the other side could not receive it in the way that matters. You cannot recognize a mirror. You can only see yourself in it.

Somewhere in a federal courthouse archive there is a notepad with something written on it from the eighth day of that trial. Nobody has checked what it says. It probably doesn’t matter.

Cheers, Zohar (Alexandria)