Dr Arnaud Gauthier, physician psychotherapist

[M]: Can you introduce yourself: who you are, what you do, what kind of patients you see?

[A]: I'm in private practice doing fairly generalist psychotherapy. My practice leans toward CBT and schema therapy. I see a lot of anxiety disorders, depressive disorders, quite a few personality disorders as well, and general difficulties: bereavement, separation. What I don't take on are patients with psychotic disorders. I only see adults or older adolescents, from age 15. Mostly in-person, but I do have a few remote patients.

[A]: I'm originally a somatic physician — I trained in emergency medicine, worked a lot in traumatology — and I've had a multi-approach journey. My first psychotherapy approach was Reichian analytical therapy, already body-centered. I trained in hypnosis, I did EMDR, CBT, schema therapy, positive psychology. I'm president of the French Institute of Schema Therapy and an ISST-certified supervisor.

[A]: Right now, I'm writing a book on chair work — it's an approach that goes well beyond schema therapy. You find it in Gestalt, CBT, emotion-focused therapy, compassion-focused therapy. My intention is to transmit a grammar of the tool, not rigid protocols.

[A]: You reminded me of an idea I had yesterday. I was in supervision with a therapist who was telling me about a patient she had just diagnosed with autism spectrum disorder. The patient is really struggling: she has enormous difficulty anticipating other people's reactions. When she sends messages or emails, she can be extremely blunt, even completely off-base. She's lost countless jobs because of this.

[A]: I'm not at all trained in ASD. But when my supervisee told me about this, I thought: this patient has a biological cognitive deficit, like ADHD. Having a crutch, a substitute for a deficient function, can often be a good idea. Wouldn't it be interesting to suggest she create a prompt to proofread the messages she needs to send? Something like a secretary or coach that would help her function more easily in everyday life.

[M]: The crutch analogy resonates with me: a tool that compensates for a function that's deficient relative to the norm and that harms social life. That would seem like a more than legitimate use.

[A]: Yes, I think it's something worth seriously exploring. If I think about the limitations, I believe the ideal would be agent-type AIs — that can spontaneously activate functions, not just produce text. For example, an AI with a schedule management function for someone with ADHD. Having an agent that can act across different platforms to support the person in daily life.

[A]: A patient has a broken leg, a physical deficit: we have no qualms saying they need a prosthesis, a crutch. It just seems normal. And I think: why should it be any different if we move beyond motor functions? For cognitive functions, memory functions, emotional functions, why would it be a bad thing to use crutches?

[M]: Do you have a sense of what makes it not as straightforward?

[A]: You break your arm, you get a cast, it's normal. Telling the person « no, you need to tough it out, manage your pain, not move your arm, it's a matter of willpower, getting a cast is taking the easy way out » — that's absurd. And yet, that's somewhat the discourse we hear for psychological issues: « You're depressed, just get moving, it's just a matter of willpower. » There's much more judgment, this idea that it's a lack of willpower, laziness.

[M]: I often hear the fear of becoming dependent on AI, whereas it would be hard to say « you're dependent on your cast ». This reminds me of Lerner's just-world theory: if you do more or less the right thing, everything will be fine — so if things aren't fine, it's because you don't want it enough. And there's this strong distinction between physical and psychological disorders, probably linked to our Cartesian dualist worldview.

[A]: Right now, I'm working on a chapter about the experiential. The goal is to bring the patient to feel, to have a corrective emotional experience. It's an experience that is not only cognitive but also embodied in the body. The fact that a psychotherapeutic technique triggers a change in bodily sensation, in interoception, has a transformative effect.

[A]: There are quite a few studies showing that we're not just a brain and a disconnected body. All the perceptions we have of our body are part of memory encoding, especially emotional encoding. In clinical practice, we see this in many patients: the emotional reasoning bias — « because I feel something, it must be true ». « I'm afraid, therefore it's proof of danger. »

[M]: Damasio wrote Descartes' Error precisely to point out the role of emotions. And Varela redefines cognition in a much broader way, within an organism. It brings us back to the material.

[M]: Do you use AI for yourself?

[A]: I use AI as a secretary, an assistant. I do my research on PubMed, download the article, give it to ChatGPT and ask for a simplified summary. That lets me see if the content is relevant to me. Then I won't necessarily always go further — sometimes the summary is enough. Sometimes, if I have a doubt, I'll verify.

[A]: I'm president of the IFTS and I run a monthly bibliographic monitoring newsletter. I use this for that monitoring. An article by Arnoud Arntz came out that updates the reparenting protocol — it's a big piece, 15 pages in English. Reading the entire article plus writing a synthesis to share with students — that would have taken me a full day. I gave it to ChatGPT, and in two minutes it produced a clear summary. I went to verify the information, and it took me half an hour.

[M]: What you're saying implicitly is that it simplifies certain tasks so much that you end up doing them instead of not doing them.

[A]: I'll be more mixed on using AI for creative work. I feel that AI is very strong at rephrasing things that already exist. However, when I had concept ideas, I tried to make it think — and I was often a bit frustrated. More than 50% of the time, I still find it interesting. But when the synthesis doesn't exist yet, when I'm the one creating it, I find it unsatisfying.

[A]: I really enjoy making connections between different fields. My first approach was Reichian analytical therapy, I did hypnosis, EMDR, CBT, schema therapy, positive psychology. I always try to find places where different words are used to talk about the same concepts, to create a synthesis. And that's where ChatGPT struggles when that synthesis doesn't already exist.

[M]: It's interesting that we have, about the same thing, two different experiences. That's the heart of the matter: perception. The same AI, depending on the use, the context, doesn't yield the same things.

[A]: I think a good part of it comes from how I ask the question. I have a fairly clear vision of what I want, and I don't take enough time to explain my intention. Sometimes I start a new chat telling it to start from scratch. And then I get better results, because it no longer limits itself to what I usually work on.

[A]: I had finished writing my book, I was delighted. I copied my text into ChatGPT and asked it to do a review. It gave me something laudatory, a list of what works well. And then a list of everything that needs changing: « You need to include many more well-structured protocols, you need to provide clinical cases with well-defined scripts. »

[A]: It gave me a bit of the blues. I thought: have I been on the wrong track? And it took me a moment to realize that actually, no. That's not what I want for this book. That's not me. My idea was to transmit a grammar. And then with that grammar, once you know how to make sentences, you can create.

[M]: That's the essential question: why do we do things? AI calibrates itself on its training data — we were already very protocol-driven before AI. But the real issue is: it has meaning for me.

[A]: This difficulty isn't specific to AI. I'm in contact with a publisher. I think when I send them the manuscript, the feedback will be the same as ChatGPT's. AI follows a trend, a convention.

[A]: And so I realize that probably, I should have explained all of this to it: what my intention is, what I want to do with this book. My mistake, clearly, was not telling it everything I just told you.

[A]: What I would find amazing would be to create a digital model of my personality. An AI to which I give all my preferences, my opinions, my ideas, my way of thinking, my dreams. A model that could do this kind of work based on all of that.

[M]: The usefulness for you would be to maintain a kind of vibration, a reference note? Like a tuning?

[A]: Yes. And I grew up steeped in SF — my parents had more than 2,000 science fiction books. The idea of transferring a consciousness, if we consider that our brain is a biological information medium — it wouldn't be absurd to envision creating a dematerialized Arnaud. There might not be the biological feeling, embodiment would perhaps be more complicated. But it would be a form of immortality.

[A]: It would be a kind of backup. Everything that makes up our essence — our ideas, our convictions, our dreams, our vulnerabilities — could persist as long as there's electricity in a hard drive. We could imagine implants that restore lost function after a stroke. Doubling an Alzheimer's patient with that. And for grief: if I lose my wife, if I can still talk with her, the experience of death would be completely different.

[M]: Isn't that already what we do when we take a loved one's scarf to seek their perfume? When we look at a photo album? It's summoning the spirit of what has been.

[A]: This makes me think: from the moment we use AI to compensate for deficits, to what extent is it a good thing? Would a human being who no longer had to experience grief — would that be beneficial? Our humanity might also lie in facing certain difficult things. Sometimes facing pain, sorrow, loss. How much does it make us dependent on the tool? The line is thin.

[A]: I think it's like everything. AI is not something fundamentally good or bad. It all depends on what you do with it. Rather than taking an approach based on good and evil, we need to bring back nuance. Ask yourself: what use do I have for it, how am I using it, does it do me good, am I getting the right dosage?

[M]: There's a philosopher, Bernard Stiegler, who recalled the notion of pharmakon: the same thing can be poison or remedy depending on the dosage and context. That is the very approach of ethics. Ethics is a process, not a checklist.

[A]: Thank you for proposing this exchange. These are subjects I enjoy discussing. I like this way of approaching things: not being dogmatic, stepping back, and sometimes accepting that there are things we don't know.

[M]: Humility...

[A]: Epistemological humility. Remaining humble about what we know, about what we think we know.

[M]: A magnificent conclusion.