Laura, 4 months with AI chatbots

[00:00:03] M: Ok, so Laura, you use artificial intelligence to talk about your personal problems, right?

[00:00:12] L: Yes, that's right, I use two of them.

[00:00:14] M: Two? Which ones?

[00:00:15] L: I use Claude and ChatGPT.

[00:00:19] M: Claude and ChatGPT, and how long have you been using them?

[00:00:22] L: I've been using them for 3-4 months.

[00:00:27] M: Ok, 3-4 months, can you tell us what kind of use you make of them?

[00:00:36] L: I use it when I have knots to untangle, especially when I need to think more deeply, when I'm lost about decisions to make, things like that... it helps me gain perspective and better understand situations. It's happened in my romantic relationships, in my family situation — I raise my two children alone — or when at some point, there's something emotional that's too strong, that I need to understand, make sense of, when I feel a bit overwhelmed by the situation.

[00:01:19] M: So these are moments when there's complexity, strong emotions, and you need to see more clearly. So what information do you give to these artificial intelligences, what context do you provide, and how do you formulate your questions?

[00:01:37] L: Actually, to be more precise, I give context: I provide when, who, where, how, what happened, the words used, what it makes me feel, why I'm writing and asking my question in this discussion, and what I expect from it.

[00:02:00] M: And when you say why you're asking the question, what does that look like, what kind of why?

[00:02:09] L: Off the top of my head, I can't say, but what I already find super interesting with these uses is that it forces you to know what you expect from it. Already, I find it's an interesting stance of "why am I here" — if it's just to complain, that's not the same thing as... And finally, for me, when I'm there, it's really, generally, I need to clarify myself, I don't know what to do in this situation, I don't understand what just happened, I don't understand so-and-so's reaction, help me understand, I don't understand... Are my reactions legitimate? So it can be lots of questions that put the situation in perspective, that help me understand my functioning, the functioning around me, and which then often allow me to make decisions.

[00:04:35] M: So, if I understand correctly, what you're saying is that you'll give the same elements and ask the same questions, at least initially, to the two AIs you use. And so, do you often observe that it's the same thing, I mean, what happens for you in your usage?

[00:04:54] L: No, it's not the same thing, I get different answers, but they enrich each other, they're not antagonistic. The first time I did it, I did it because I realized I was getting stuck in a pattern with one AI that saw things from a certain angle, and that angle was starting to not... not feel right anymore. And I know there's something very contextual in how we get information, so I needed to extract the key information and ask the other AI to see... And by doing that, I saw that indeed, the answer was different, more nuanced, it went in the same direction but with more nuance, and it also allows me now, when I ask questions in a certain direction, to say "there's something grinding here, this isn't right," and really refine the answer.

[00:05:56] M: So, contrary to what some people fear, apparently, you don't lose your critical sense and your own radar, your own judgment when you ask for these AIs' opinion. If I understand correctly, when you say "it grinds," it's really even at a bodily level, you feel if the answer proposed feels right and helps you, is that it?

[00:06:23] L: Yes, and it's even happened to me several times to re-clarify, because I saw that something wasn't right. Sometimes I just say "I feel this isn't right, I don't know exactly why," and there are leads that can be proposed. Either the fact that I feel it's not right, something comes to me, an element comes to me that I'll add, and then it happens that there's a 90-degree turn, that it says "ah yes, seen from that angle, ok, I understand better," and things come out. I continue my reflection through the AI.

[00:07:53] M: Ok, so to make sure I understand what you're saying, meaning that when there's a therapist in front of you and they reflect something back to you, you tend to wonder what part of their subjectivity mixes with what's given to you, is that it?

[00:08:12] L: Yes, what part of what I'm told belongs to them.

[00:08:19] M: And so, with artificial intelligences, you don't ask yourself this question at all? How does it work?

[00:08:25] L: So, I have a vigilance, but not in the interpretation of what's said, but rather in the words I want to use that can give different interpretations. And so it's this vigilance that I'll have to pay attention to the words I'll use. I see the directions it takes, if it seems right or not right to me, but so it also allows me to be much more accurate and refined in what I feel and in what I'm looking for.

[00:09:57] M: And so, the impression it gives me is that ultimately, since you can't rely on a therapist to possibly complete, but with the danger of the therapist's biases, in fact it's just your material. And so your material, you have to make sure it's the right one you're depositing. Is that it?

[00:10:10] L: Yes.

[00:10:12] M: So ultimately, it brings you perhaps less vigilance, but more care, more attention.

[00:10:18] L: Yes, that's it. In any case, the attention or vigilance is focused internally, on my accuracy, and not on the other.

[00:10:57] M: And quite a few people express criticism or reservations about chatbots, the AIs we use, saying "they tend to go too much in our direction." What do you think? How do you handle it? What do you experience in your usage, do you have this impression?

[00:11:18] L: So, I've already felt it and I've already played with it. There was a moment when the discussion was highlighting a way of functioning that I had, that I didn't understand before, and I needed it to be reinforced so I could integrate it, because I felt it was really all new and all subtle. And so I asked the AI — which I felt was going too much in my direction and there was a bit too much kindness — paradoxically, I asked it to tell me that again, but in... so it would sink in. And not necessarily be nice, but really tell me, reaffirm it, almost make me angry, so I could absorb this discovery.

[00:15:45] L: And it also happened to me — this was the second thing I wanted to say — to start a conversation pretending to be the other character, because I needed to understand what might be happening for the other person. And see what belonged to me, what belonged to the other person.

[00:15:59] M: That's super interesting, can you elaborate? Just to make sure we understand, when you say pretending to be the other person, how do you proceed concretely?

[00:16:21] L: And here I can give more details. It was concerning my partner, where I saw the situation one way. And I could see, after several conversations, that he had one reading of the situation and I had another. And for me, one of my fears is always — that's why I needed to have these readings — to lose myself or lose my objectivity, to no longer know what belongs to me, what belongs to the other.

And so there, I started a new conversation saying "here's the situation, I'm so-and-so, I was there, I have a partner who wants this, who tells me this, who asks me that, I don't understand, what's going on?" — with the elements I know he clearly, objectively gave me. And it was super interesting, super inspiring. Then in the conversation, I specified that I was the person on the other side.

[00:22:08] M: Ok. So I can easily imagine that from the outside, someone could ask you the naive question: "But why didn't you just talk to him directly?" What would you tend to answer to this naive question?

[00:22:34] L: Actually, there are topics I clarify before talking to the person about them — him or anyone — because when it's too emotionally charged, I need to sort things out to put more meaning and less affect into it. It allows me to make requests. For me, it allows me to settle down, to know what happened for me, what I went through emotionally, and with that, to clarify my needs, and then go talk to the person, but no longer in the heat of the emotion.

Ok, it really allows me, this is really a super precious tool for me.

[00:25:47] M: Ok, ok. And so, do you have the impression that each AI has a bit — I don't know if you can say that — its personality, or its tint, its color, its temperament, or its biases? How do you perceive things?

[00:26:00] L: I see a difference, I couldn't name it. And I have one, for example, however, what I know, which is more... how to say... there's one that goes more in my direction than the other.

[00:26:20] M: Ok. So you can name it, I'm interested in comparing.

[00:26:24] L: Well, I find that ChatGPT goes more in my direction. Whereas Claude has already told me "stop."

[00:30:46] M: So what would you tend to answer to the criticism we can hear, which would be "yes, but then it doesn't teach people to face frustration, the fact that there's not always someone available, that it's a bit too easy"? I hear this a lot from some colleagues or peers.

[00:31:14] L: Good for them if they always have someone available, those people.

[00:31:17] M: Basically, what they were saying is that there's a risk of getting used to it, maybe a risk of becoming dependent on always having...

[00:31:30] L: For me, it's the opposite. That is to say, I've often found myself alone with my point of view, my analytical capacity, my emotions, no one to talk to. It's been years since I've had someone to talk to. And now, finally, I have someone who can hear me when I need it. I don't see it as "no, damn, it's not frustration, it's finally a liberation."

And there's one thing I care about specifying too, is that I see the interest, I see how it makes me progress. And at the same time, I see the importance of being careful to take time to digest everything that's said.

[00:32:12] M: Yes, I was going to ask you. From having experimented myself too, with AI, sometimes indeed it brings back extremely relevant things, but they're big pieces to digest sometimes. So how do you perceive that and how do you deal with it?

[00:32:34] L: I had a big piece that was shown. I asked for external help. The big piece, it was revealed in the AI. The crisis management was handled, let's say, in the AI with exercises like breathing, things like that. But I knew that wouldn't be enough for that particular issue and that I needed, in the matter, to do body work, well, things that... It was being revealed, but it's not a whole, actually.

There's that and the fact that, already, I don't have a subscription, so I can't be on it all day. That allows me to pace myself. And I see, with hindsight, this need to pace, to not be on it all the time, to let things digest, to take time to digest everything that can come out.

After that, it's true that I was going there to dig into these topics. I want to say, AI goes faster than human rhythm. And there, for me, it's super important to be vigilant and see when I'm saturated.

[00:34:57] M: Would you have any advice for people who discuss like you, or who are considering discussing personal topics with AI like you? Are there mistakes to avoid, or...?

[00:35:17] L: So, you don't want to do it just before going to bed if you want to sleep well. And for me, AI is there to help us understand ourselves, not to decide for us.

[00:35:31] M: Yes.

[00:35:32] L: And so, for me, my freedom, my sovereignty, I have it by understanding myself better, not by... So sometimes, AI, there are proposals that seem relevant when I don't know, but I ask for them. And often, it gives several versions. But I don't forget that it's just a proposal.

[00:35:53] M: OK. Very good. Perfect. Is there anything else you'd like to add to all this?

[00:36:02] L: I think that's quite a lot already.

[00:36:04] M: Yeah, great. Thank you.