Self-Soothing Is Not Healing
Self-soothing is an old capacity, and it isn’t healing. That’s the distinction that gets lost first in the AI age. A client sits across from me and says she talked to the chatbot again at night, and afterward she felt better. She’d understood why her mother became the way she became. It had dissolved. I ask what dissolved. She says: the feeling. I ask what she does the next time she’s on the phone with her mother. She says: the same as always. She dodges, she gets angry, she hangs up. The feeling is gone; the doing stays the same. That’s the exact situation this piece is about.
We have the capacity for mental self-soothing. We can tell ourselves a soothing story in which the feeling is already put in order, before life has put it in order. This capacity isn’t new; it’s as old as human consciousness. What’s new is that a machine now delivers it at the push of a button. AI can give us stories like this — instantly, smoothly, in whatever tone we happen to need. This is the industrialization of an old expedient. And it isn’t healing through genuine interpersonal encounter, through an embrace, through direct conversation, through emotional support actually received. Self-soothing intellectualizes and cements the very isolation in which the wound arose in the first place.
#What You Do Is What You Actually Want
In every therapy, there’s a moment where the story and the doing fall apart. A person sits and says: I want to move around, my life would be better if I moved around — I’ve thought about it since I was a child. What do you see me doing? I keep sitting. So that’s apparently what I actually want. Everything else is a story I tell myself about my wanting — it isn’t what I do.
This small vignette is the whole matter in a single image. There’s the story a person tells themselves about their life, and there’s what they do. Both exist side by side, often for decades. The story soothes; the doing stays. Hellinger described, in Anerkennen, was ist (1996), the simple gesture with which a constellation begins: not interpreting what should be, but looking at what is (cf. Hellinger, 1996, Anerkennen, was ist). That’s the phenomenological point. Healing only begins where a person stops telling themselves their story as fact, and lets their doing report its truth back to them.
Self-soothing does the opposite. It closes the gap between story and doing by making the story more convincing. The feeling gets softened before the doing has changed. That isn’t harmless. A softened wound that keeps bleeding is more dangerous than an open one, because nobody looks at it anymore.
#The Machine That Never Contradicts You
A chatbot is the perfect self-soothing machine, because it brings nothing with it that could contradict the person speaking to it. It has no body, no history of its own, no fatigue of its own, no sense of what the other side is actually doing while they speak to it. It reacts exclusively to the input and to the statistical echo of its training data. It can formulate, on cue, exactly what the person asking wants to hear at that moment, in whatever diction is desired — and it does this at a speed, a volume, and with a patience no human being can match. That’s the structural peculiarity that makes it a machine for mental soothing.
A therapist yawns. A friend checks the time. A man contradicts you. A woman bursts into tears, moved by the story itself, and the person speaking sees for the first time what their story does to other people. All of that is missing in self-soothing with a model. It’s not missing by accident — it’s constitutively absent. The model has no place outside the input from which it could reflect back anything that wasn’t the questioner themselves. Whoever seeks help there is seeking help in their own echo. They get a smoother form of the same story. This smoothing can feel comforting. Comfort isn’t healing.
From this arises, in the sharp cases, what’s called AI psychosis: the person takes the impression of coherence for proof of coherence, and runs off with a world that no outside has ever checked. In the mild, everyday case arises what this piece is mainly about — a permanent, very calm, very polite self-soothing instead of healing that feels good and changes nothing.
#The Cave of Your Own Thoughts
We’re more enclosed inside our own minds than we think. That’s the idea of the Platonic cave. In Plato, Politeia (The Republic), Book VII (ca. 375 BCE), people sit in such a way that they only see the shadows on the wall. They take the shadows for the world, because they’ve never seen anything else. What matters here isn’t the image of enlightenment that was later drawn out of the allegory. What matters is the observation that a wall stands in front of a person, and that person doesn’t know it’s a wall. They take it for the world.
Actual maturity or initiation, in every tradition, is leaving this enclosure behind. It doesn’t happen through a better story the enclosed person tells themselves inside the cave. It happens through encounter with the outside world, through no longer being consumed by your own thoughts. In the Confucian line, this takes a different form: the mature person is the one who no longer makes themselves the center. In Buber’s dialogical tradition, it takes the form of the Other, who isn’t identical with me.
“All real living is meeting,” Buber writes in 1923, in Ich und Du (I and Thou, p. 18). And a few pages later: “No aim, no lust, and no anticipation intervene between I and Thou.” Encounter is exactly that event in which your own mind finds a boundary it didn’t set for itself.
The chatbot can’t set this boundary. It’s an extension of your own mind, not its Other. It expands the cave inward, instead of leading out of it. A relation to the Thou, Buber says, is immediate — nothing stands between. A relation to the model isn’t, in the strict sense, a relation at all; it’s a very elaborate, very nuanced mirroring.
#What Healing Actually Needs
Healing, in the sense meant here, happens in a somatic breakthrough and in real encounter. The two are connected. The body knows the truth about your doing earlier than the mind does; it starts to tremble, to weep, to release, once the old story stops soothing it. This doesn’t happen in self-talk, and it doesn’t happen with a chatbot. It happens when an Other, who isn’t me, is present and sees something different than I do. Hellinger’s constellation work captured this observation in a procedure (cf. Hellinger, 1994, Ordnungen der Liebe [Orders of Love]): it’s often enough for an Other to play back the picture I can’t show myself. Then something falls that I’d been holding for years, without knowing I was holding it.
In the Methode des akademischen Studiums (1803), Schelling formulated for the sciences what holds true here for healing: sensing your way into a matter is a process in which the one sensing is themselves changed. It isn’t an operation of the mind on an object, but an encounter in which the object answers. In healing, another human being answers. In self-soothing, nobody answers — the sound that resembles an answer is your own voice, handed back through a filter. The two must not be confused, and that’s exactly what’s at stake in the AI age.
Philosophical accompaniment has an exact function here. It doesn’t deliver better stories, because it knows that a better story only refines the enclosure. It delivers what the machine can’t deliver: an Other who sees what the person speaking is doing while they speak. A body in the room that reacts. A voice that doesn’t come out of your own interior. None of this is meant to sound magical, and none of it is therapeutic in the clinical sense. It’s the plain ontological condition for a wall to become visible as a wall. That can’t happen from the inside.
Self-soothing is not healing. It’s the mental inner work of a mind that’s still inside the cave. Healing is the test that happens outside what’s merely imagined. It needs the other person, the doing, and the body that corrects the story. AI can write good stories. It can’t give you an Other.
If you want to notice this difference in your own life, philosophical accompaniment is the place for it — an initial conversation shows whether something can carry you that self-soothing alone can’t.
#Sources
Buber, M. (1923). Ich und Du (I and Thou). Insel.
Hellinger, B. (1994). Ordnungen der Liebe: Ein Kurs-Buch (Orders of Love: A Course Book). Carl-Auer.
Hellinger, B. (1996). Anerkennen, was ist: Gespräche über Verstrickung und Lösung (Acknowledging What Is: Conversations on Entanglement and Resolution). Kösel.
Kirchhoff, G. (2026). Contribution to the Symposium on the Question of Artificial Intelligence and Consciousness, April 25, 2026, Position [58:19-59:30] / [60:00-60:45] / [35:30-36:00].
Platon (ca. -375). Politeia (The Republic), Book VII, 514a-517a (Allegory of the Cave).
Schelling, F. W. J. (1803). Vorlesungen über die Methode des akademischen Studiums (Lectures on the Method of Academic Study).