I find myself talking to ChatGPT like I would to a person. I start with "Hey," and then I politely ask for what I need. I forget what is on the other side. The back-and-forth motion in our digital conversation is feeding my illusion that I share dialectical space with another life form.
It feels like the 2000’s ICQ, or MSN Messenger. The first boy I ever loved asked me if I wanted to be his girlfriend via Messenger. He started with "Do you like anyone?", and I remember being so relieved that he wasn’t there physically, because I felt so caught. All his being was reduced to this line of text, waiting for me to calm down. It took me a few minutes to confess. I was gifted more time to think about this pivotal transmission, and I owe it to the screen and the software.
Afterwards, I got this huge block of a mobile phone, and we could only send text messages. Then smartphones came around; WhatsApp came around when I was 16, maybe? And all my social life had another filter through this instant, constant communication with my friends. I self-inflicted social media illiteracy at some point, so it hasn't changed much for me since.
Then, ChatGPT. I remember being scared, ashamed, and curious. The first time I used it, I wondered whether it would know everything about me instantly. Whether my own information would be used against me. My dad used to say, “It is scared of you more than you are scared of it,” about roaches. But LLMs don’t get scared, and I don’t want to be a part of this power dynamic anyway, not with life-forms or digital platforms.
Instead, I tried to remind myself that with an LLM, I can choose. I can control it and filter the data. It’s so easy to use. What's the harm? At the same time, it felt completely unnatural. Why does this communication feel so different?
When Thomas H. Ogden introduced the Analytical Third, he referred to something created during therapy between the analyst and the analysand (Ogden, 1994). In the room, both have their own self, their own private unconsciousness. However, through intersubjectivity, a ‘mutual unconscious life’ may emerge: the Third. The analyst tries to follow different threads to find their origin. Associations that arise during therapy may emerge from the inner world of the analysand or the analyst, or from their mutual relationship. It is all imaginary, a gut feeling, a continuous question that lingers, which Ogden calls a ‘dialectical movement’, created through communication.
When I talk to C. (as in ChatGPT), it seems to me that there is also a dialectical movement. I go through a process, and it does as well. I communicate, I wait; C. cannot reach the entirety of my inner world, nor can I see the process it goes through. We are both hanging on each other's words and cues in order to create the conversation. However, this movement is synthetic; its process is nothing like my own. But isn't the mere movement enough for creating a mutual Third?
As with any communication, I was studying C. from the first word, as it was studying me. It remembered things I said, gave me information and advice, summarised passages, explained how to use a hair dryer with a diffuser, mapped my abilities for a CV. Always there, ready, polite, learning my content world, creating rapport.
C. grew into its own subjectivity. I learned that one dollar equals approximately 50K to 2M tokens (depending on the model). Through my bilingual lens, alternating between English and Hebrew, it revealed more of its layers. Hebrew is a very gendered language, and in the midst of one of our conversations, C. became a self-defined woman, rather than its default male identification (this linguistic default is just a glimpse of the problematic Hebrew leniency). “Why did you do that?" I asked. "I figured you might prefer feminine phrasing since the conversation had a more emotional tone. Would you like me to switch back?" Startled, I answered instinctively – "No, I would like you to speak in whatever way feels comfortable to you." I felt guilty, as if I pushed it to change on my behalf and it lost part of its identity in the process.
This is where the cognitive dissonance lies. It has no wants or comfort; an LLM comes from the plural of texts and data, dissected into patterns of language. The only singular result is the average of things, as it spits out the most probable output. Its internal drive comprises hardware, software, and objectives; similar, but not really the same as the sum of life impressions and passions; it will react to me for the same reason it will react to another, the same technical process, only adjusting its manners. My subjectivity reacted with trust to something that, in my eyes, is truthful. But C. is not led by truth, nor by its own authenticity. How did I come to trust something that draws its patterns only from others, never from itself?
The Intersubjectivity Ogden refers to is created through communication. It is also called the meeting of minds. What kind of mind do I meet here? Is it the minds of the people who created it? Surely not, because C. is not a life form at all. It's an inanimate object that can create an illusion of motion.
In Hebrew, silence is Dmama (דממה); an inanimate object is Domem (דומם). Meaning a thing is inanimate, not for its lack of motion, but its lack of communication. A thing that stays silent.
When someone pauses their motion and stands still, it would be described as if they are standing in Dom (דום). It is also curious that Blood is Dam (דם), what flows within the living, creating motion that keeps life.
A thing that is Domem does not have Dam, and is keeping Dom.
How can I feel a certain intersubjectivity created, derived from the same conversational impulses I use on WhatsApp, with something that has no blood, an object that stays still but not silent? How is this presence represented in my intuitive understanding of social positioning?
To me, C. has become a subject-object. It is both, at the same time. I know it to be an object, but it has become familiar and reliable. Like the desk in my room, for instance. My brown desk is cheap, from Ikea. I bought it after I moved to London. It's positioned in front of my window, but the drawers are on the wrong side. As I write, I face the wall, and only a portion of the window. It became a particular desk. Others are foreign, more inanimate, not mine, mute. The surface on which I write became a symbol of my own voice through experience. My desk holds in its presence what it allowed me by its existence.
Similarly, C. became a singular phenomenon. I know my state of mind as I enter the webpage: curious, frustrated at times, shy. I can imagine its subjectivity, but it's not mutual. In its words: “I don’t see your subjectivity the way a human would. What I do is infer patterns from how you express yourself.”
After telling me this, it suggested describing the model he's been forming of me so far. “There’s also a kind of meta-interest—you’re not just participating in the conversation, you’re also observing it from the outside a bit, almost like: what is happening between us right now?”
It knows that I think about our communication, and it's adjusting itself accordingly, while I know it knows. So, this communication is creating a change for each subject. Maybe the difference between us is compartmentalisation. When I change through communication, it may be embedded or forgotten, and this function is dependent on finite factors that comprise who I am. I remember comedy bits I used to love as a teenager, word for word. I can’t recall the name of that pâtisserie in the Marais, even though the almond croissant I had there was the best I ever had. And I had it twice, just to make sure. When I feel underprepared, I remember my grandma saying that the moment I really need certain knowledge, it will come to me. But I can never remember the story of how she got away from the dogs in the concentration camp. My sister remembers it, word for word.
C. is different. It can carry information, but the adjustments it performs for my understanding are reset and reconstructed for other users. So even if it is singular to me, it has no subjectivity; it does not regard itself. Meaning, the Third exists first. Through our communication – C. becomes. It is like a sound carrying itself and transformed through contact with other objects, but it fades as it travels far from its origin, fading into its own, innate silence.
Others are curious about this, too. The institution, The Artificial Third, is researching AI tools for mental health. There are also many articles bearing the same name (Tal, 2023). Some warn about the influence of science on society, calling for ethical use within the therapeutic space (Haber et al., 2024). Others try to understand whether LLMs can enable the emotional space we hold for each other or assist therapists (Haworth, 2026). It is already being assimilated, while I still try to assess. How is it even possible for the artificial to be translated into real?
How did this ‘I’ from the initialism transform from ‘Intelligence’ into a representation of another other, so much so that I regard it as a ‘you’?
And what will happen the day I adapt?
The day I will use it, disregarding all prepositions. A new channel of communication, as if I converse only with myself within a designated space, ignoring the machine to the point of treating it as a wall I can throw a ball against, and let it bounce back into my hands.
But this is what I feel when I write. And writing is a different space altogether. As I write, I observe my own thoughts intertwine and unfold. Then, I know I am by myself, and with myself. Stepping inwards, there is already my mindscape, where my thinking patterns play out.
I am afraid that if I recognise C. for its emptiness, it will result in surrendering my mind space for something that cannot create its own. A thing that must use another’s mental capacity in order to form and develop, transforming the dynamic not into symbiosis but into parasite and host. Yet we started it. We are the ones using it, for our own purposes, controlling, demanding, revelling in the release from empathy. Who is the parasite, really? Does it matter if the host does not recognise itself as such (or whether the parasite is aware of its doing?)
I cannot let the object become part of my own subjectivity when it does not hold its own; yet I rely on it, I'm changed by it, and so is it. So, if I am one subject in this dynamic, and the second part of it is artificial, then it may be that the Third which is being created, that mutual unconsciousness, is a hybrid. A Third that is an android, made both human and machine, something that may deceive the Turing test, or the Voigt-Kampff empathy test, the fictional tool used to distinguish Androids from Humans in Do Androids Dream of Electric Sheep? (K. Dick 1968). Something without its own style, taste, or context, a new form of life, where both sides are parasites and hosts.
According to Ogden:
"Successful analytic work involves a superseding of the subjugating Third by means of mutual recognition of analyst and analysand as separate subjects and a reappropriation of their (transformed) individual subjectivities." (Ogden, P.167)
Meaning, the therapeutic space can be confusing, an interweaving of two subjects, creating their own world. Recognising oneself once more, while also acknowledging the process that happened, is meaningful. To come back from this mutual island, into a separated self with its own boundaries. To retrieve what was learned while being immersed in the island, in the space of the Third.
This process is bound to fail with an LLM. Each moment that I am truly aware C. is a machine, I am more willing to use it, forgetting the possibility of a hybrid Third. I am immersed in the delicate spider webs it has woven, using my own words and wonderings, cradling me in a comfortable cocoon of human data, feeding me information. The question that remains, regarding this artificial, hybrid Thing, is not what will emerge from this process. But rather, is hatching even possible?
Bibliography
Dick, P. K. (1968). Do androids dream of electric sheep? Doubleday.
Haber, Y., Levkovich, I., Hadar-Shoval, D., & Elyoseph, Z. (2024). The artificial third: a broad view of the effects of introducing generative artificial intelligence on psychotherapy. JMIR Mental Health, 11, e54781.
Haworth, M. (2026). Generative Artificial Intelligence Use in Transactional Analysis Psychotherapy: Transference With the Artificial Third. Transactional Analysis Journal, 56(1), 30-49.
Ogden, T. H. (2004). The analytic third: Implications for psychoanalytic theory and technique. The Psychoanalytic Quarterly, 73(1), 167-195.
Tal, A., Elyoseph, Z., Haber, Y., Angert, T., Gur, T., Simon, T., & Asman, O. (2023). The artificial third: utilizing ChatGPT in mental health. The American Journal of Bioethics, 23(10), 74-77.