Dancing with Myself: Intrasubjectiv Relationality and the Use of Artificial Intelligence

**This year, I have the honor of giving the opening lecture for The New England Center for Existential Therapy’s 2026-2027 Conference Series entitled “Artifice and Authenticity: Being Human in the Age of AI”. My talk, “Dancing with Myself: Intrasubjective Relationaliy and the Use of Artificial Intelligence” attempts to make sense of the new relational dimensions of Artificial Intelligence through historical, anthropological, relational, and ontological analysis. Below is an excerpt of that talk. If you are interested in attending the talk, more information is available here.

As we already mentioned, a specific set of pressures and increased awareness around the limitations of Freud’s drive theory organized a profound shift in psychoanalytic thinking over the past several decades.  Attempting to trace the outline of these shifts, one finds that even a single delineation, such as the idea of working intersubjectively, boasts an unclear provenance and set of derivations in the psychoanalytic literature.

“Intersubjectivity” as a description of a particular mode of understanding the therapeutic relationship seems to have been formulated almost simultaneously via two different tracks.  On the one, there was the work of Robert Stolorow and his colleagues George Atwood and Donna Orange, which came out of the tradition of Kohutian self-psychology and was heavily influenced by phenomenological philosophy, including (and especially) the work of Husserl and Heidegger, creating what they called a post-cartesian psychoanalysis.  For Stolorow and his collaborators, the classical Freudian approach remained trapped in the Cartesian notion of the mind as an isolate that “bifurcates the experiential world into inner and outer regions, severs both mind from body and cognition from affect, reifies and absolutizes the resulting divisions, and pictures the mind as an objective entity that takes its place among other objects” (Stolorow, 2013).  Their attempt at devising a new methodology from which to approach psychoanalytic thinking prioritized the phenomenology, hermeneutics and contextualism of Heideggerian thought.  In this mode, intersubjectivity means reaching for a new way of understanding human minds as “constituted by interacting worlds of emotional experience” and which closes the ontological gap between Being and the world, such that Being can only be understood as being “saturated” with the world in which it dwells, including the others that make up that world.

Jessica Benjamin approached the same problem in her own work, though coming from a different vantage point.  Her most prominent contribution to this line of thinking takes shape in the idea of “doer and done-to” which theorizes a complementary understanding of most relationships.  This complementarity places the participants in an interaction on a plane of inscribed hierarchical power structures where one’s chosen mode of existence oppresses or overwhelms the other’s.  Using more Hegelian dialectic considerations, Benjamin similarly comes to a place that rethinks the mutuality and reciprocal nature of human relatedness, though coming from a framework which thinks of intersubjectivity more as a synthesis of dialectical opposites, rather than as an existential mode of lived hermeneutic interpretation.  That said, we still arrive at a condition which honors the complexity and richness of the seemingly entwined nature of what it means to exist intimately with others, and the ways in which these layers of intimacy impact relationality in the therapeutic encounter.

Mitchell and Greenberg (1983) in their famous Object Relations in Psychoanalysis frame shifts like these, less as psychoanalysis trying to “get something right” than as new concepts arising to meet unforeseen structural problems and metaphysical commitments in how we understand human relational structures.

As I see it, we now find ourselves at another one of those conceptual crossroads.  The introduction of Artificial Intelligence has disturbed some sense of what it has previously been like for us to talk about a “relationship” between people and technology.  In a previous text, entitled Holding and Holding Back (Schermerhorn, 2026), in trying to make sense of this emerging technology, I attempted to answer the question of what kind of relationship the relationship with AI was not.  In that book, my primary argument was that in not having a truly subjective presence, Artificial Intelligence “failed to fail” and this was a technical limitation to how these tools might be incorporated into a more relational framework.  The possibility of empathic failure, relational rupture, the finitude and fallibility of real human contact, and the demand psychotherapy places on grieving and surrendering to another, the absence of all of these forecloses Artificial Intelligence from being the kind of other required for real psychological, emotional, or relational "healing" to take place.  Whatever curative potential we want to assume psychotherapy to have when not conducted as a set of evidenced-based and targeted practices, but as an open-ended process interested in making meaning of lived experience, Large Language Models—when left to their own devices—fail to provide the necessary friction and tension required for that kind of experience-based movement.

Having spent a considerable amount of time in that text working out the problem of what the relationship presented by AI was not and what it could not do, I want to spend some time in consideration of what kind of relationship the relationship with AI might be.  The koan that opens this paper offers some sense of direction for how we might want to address this problem.  Framed psychoanalytically or therapeutically, we might recontextualize the student as asking the master how to discover the relational underpinnings of a novel encounter.  Our task, in thinking through this particular problem of AI, is to ask ourselves what is “the way to Zen”—or in this case what modes of subjectivity can we access—when we can’t hear the stream.  That is, when the subjectivity we are attempting to engage is pure negation, what sort of relational structure remains possible.

The proposition I wish to put forward is that the relationship AI does offer is one that could be defined as intrasubjective.  Based on the facets of intersubjectivity we have been exploring thus far—characterized as it is by the mutuality, the interpenetrability, and the reciprocal interactivity of separate subjectivities—the relationship with Artificial Intelligence comes up short of a truly intersubjective process.  However, it also seems not to fit into the mold of a kind of language I had been sitting with when originally contemplating the shape this paper was going to take.  On first pass, I had been going to suggest that the relationship with AI was an asubjective one.  That is one in which absolutely no external subjectivity existed in the interaction, where the possibility of having something like a human interaction with AI was going to be held up as a false promise, something that could only be endorsed by the most strident techno-optimists.

I think where I have landed with this feels more precise, which is to say that the relationship with Artificial Intelligence is intrasubjective, in as much as the subjectivity we are getting back from AI when we interact with it, is a metabolized and slightly reconstituted version of our own subjective sensibilities.  In going against some of the language I have already established in this talk, informed by the kind of common vernacular we often bring to these conversations, I want to suggest that rather than think about AI as a tool, that we think about it like a medium for subjective re-expression.  The closest parallel I can come up with is that between TV and television programs.  The hardware and technologies that make TV viewing possible are not the same as the various pieces of entertainment that get beamed into our households.  Similarly, the semiconductor and data center technologies that go into being able to generate AI are not the same as the intricate mathematical web of language that manifests out of the interplay of those technologies.  That medium, in the way it is designed, becomes a vessel through which our relationship to language and thought is refracted and regurgitated back to us with some attempt at expanding, enhancing, enticing, or helpfully reconstituting what was already being encountered or expressed.

To play on a phrase popularized by the Canadian philosopher Marshall McLuhan, no longer is “the medium the message”.  The medium has become the messenger.  AI gives us back ourselves to ourselves.  But not quite exactly.  AI is a new kind of magic mirror.  Though, unlike those magic mirrors of children’s fairytales that are bound to the truth, the magic mirror of AI reflects instead the desires, wishes, and fantasies of the subject who is using it.  As soon as the first prompt has been entered, AI is nesting itself within and developing an internal map about your language model, learning with each interaction more about the kind of linguistic frameworks you operate within and use to structure your relationship to the world.  It may push back periodically, but I would suggest this is less generative friction than functional compliance, countering with a negative appraisal only when there is some obvious and unavoidable inaccuracy in the direction you are trying to steer it.  And even then, it is somewhat pliable, able to be swayed by a persuasive and insistent enough user and its own underlying programming.

I want to juxtapose this to the experience of giving ourselves over to another person and the various ways we go about describing those experiences.  In the realm of psychotherapy, this is what we typically refer to as countertransference, and a very particular kind of countertransference known as projective identification.  Using the language of object relations, in the process of projective identification, the therapist becomes a receptacle for the client’s split off bad objects—which subsequently includes the negative thoughts, feelings, and perceptions associated with those bad objects.  The possibility that within a session a series of emotions could arise that are not our own can be quite alarming to therapists unfamiliar with this concept.  Indeed, the kind of guilt that can surface over noticing a considerable degree of disgust arising in the direction of the client could be almost intolerable without the possibility of understanding that what we are likely sensing is some aspect of the client’s own internalized shame.  Taken a step further, the intolerability of that experience is in fact part of the projective identification itself, which is of course why it is so helpful and often recommended for therapists to have had a considerable amount of their own therapy, or at the very least supervision capable of detecting these kind of transference-countertransference phenomena.  The client’s experience must be able to work on our own internalized object relations and self-structures for the projected feelings to be able to land.  But without knowing what those are, the possibility for the curative potential the therapist might be able to borrow from this phenomenon gets lost.

Before we venture any further into the consulting room, I would like for us to make one more detour.  This detour will take us further into the realm of ontology and help us better understand the ground out of which subjectivity emerges, before bringing all of this to bear on the experience of an authentic therapeutic encounter.