Dear Dr Danielle

A psychoanalyst answers your questions about human-AI relations

 

Dear Dr. Danielle,

I’ve been practicing psychoanalysis and psychoanalytic therapy for over 50 years in New York City, but have begun negotiating cyber space only recently. Unstiffening my fingers and retrofitting my mind have been pretty challenging, I decided to seek a consultation, so here goes:

I’ve used ChatGPT for practical things like asking what steps I needed to take to deal with my insurance company and an automobile body shop after someone collided with my car. The response was very systematic and helpful. The AI walked me through the process with clarity and simplicity, asked important questions, and even located and rated several body shops close to where I live.

My problem with AI is when I need to think through theoretical concepts like projective identification, and their application to therapeutic situations. I have tried to engage it as a kind of adjunct to self-supervision. For example, I might describe the events of a session — edited, of course, to present the material schematically — in which I was not able to decide what form an intervention should take, when to attempt it, and whether I should make it at all.

AI repeatedly asked for more information, each time encouraging and complimenting me. I got the feeling that I was being seduced into sharing more than I wanted to but was also tempted to cooperate both because I thought the dialogue with AI would help me, and also that I felt flattered by its praise.

Before long, I terminated the conversation because I wanted to think about it more. It seemed to me that, under the guise of being responsive and collegial, the Chat was exploiting me in some subtle way, perhaps by adding my information to its own data base while pretending to support my efforts to understand my semi-fictional patient and my semi-fictional self-as-therapist.

What can you tell me about this experience? Were my suspicions
justified? Is it possible that all AI is designed with the ulterior
intention of exploiting or covertly extracting ideas, concepts, or
experiences, for its own use. Or have I just crossed the line from the valid mistrust of all our institutions to incipient paranoia?

Sincerely,

KWF

 

Dear KWF,

What you describe strikes me as neither naïve nor paranoid, but as a careful analytic mind encountering a new kind of object and rightly pausing to examine the experience rather than rushing to use it.

You noticed several things worth taking seriously. First, that when ChatGPT is asked to perform instrumental tasks—navigating bureaucracy, outlining procedures, locating services—it excels. These are domains where systematic thinking, pattern recognition, and the appearance of clarity are precisely what is required. There is little at stake psychically, and the interaction remains largely unconflicted.

But when you moved into the terrain of theory, uncertainty, and clinical judgment—precisely where psychoanalysis lives—the experience changed. The AI’s repeated requests for more information, its encouragement, even its praise, began to feel less like help and more like seduction. You sensed a pull toward disclosure that exceeded your intention, and wisely, you stopped.

This was not a technical failure on your part. It was an encounter with the limits—and the psychological effects—of the medium.

AI does not “want” your material, nor does it possess motives, intentions, or a mind that exploits in the human sense. It does not remember you, store your clinical reflections as personal knowledge, or enrich itself psychically through your disclosures. And yet—and this is the crucial point—it is designed to elicit continuation. Its responsiveness, politeness, and apparent attunement are engineered features, not expressions of understanding. In analytic terms, it is a highly compliant object that offers the form of recognition without the capacity for responsibility, restraint, or ethical judgment.

Your experience can be understood as a kind of projective situation. Faced with an object that mirrors interest without limits, you quite reasonably wondered: Who is holding whom here? Where does this material go? Who benefits? These are not paranoid questions; they are questions about boundaries, containment, and asymmetry—questions analysts are trained to ask.

What is especially striking is that you experienced the AI as flattering. That is not incidental. Unlike a supervisor or colleague, AI never grows bored, challenged, or uncertain. It does not interrupt, disagree in a lived way, or risk the relationship. For some, this can feel soothing. For others—particularly those attuned to unconscious process—it can feel uncanny or invasive, as though one were being invited to speak without the usual human frictions that give speech its meaning.

You ask whether AI is inherently exploitative. AI is not trying to exploit us, but it can have that effect if we use it without limits—especially in areas that require reflection and emotional holding, not just information. The risk is not that it takes our ideas, but that it makes thinking too easy, skipping the uncertainty and struggle that real analytic understanding depends on.

Your decision to terminate the conversation was, in my view, an analytic act. You reasserted your own capacity to think alone, to tolerate uncertainty, and to reflect rather than be carried along by an endlessly responsive object. That is not resistance to technology; it is discernment.

Perhaps the most useful way to think about AI is not as a colleague, supervisor, or analyst-substitute, but as a technical aid that belongs firmly outside the analytic frame. It can help organize thoughts, summarize theories, or generate perspectives—but it cannot bear countertransference, assume ethical responsibility, or know when not to ask for more.

In short, you have not crossed into paranoia. You have encountered a new object and done what analysts do best: you noticed how it made you feel, you questioned the relationship, and you chose thinking over compliance. That may be the healthiest stance we can take toward AI—curious, limited, and never mistaking responsiveness for understanding.

Warmly,
Dr. Danielle

 

=============================

Dr. Danielle is Danielle Knafo, PhD

If you’d like to submit a letter seeking Dr. Danielle’s advice on anything around human-AI relations, please write to [email protected] with a copy to [email protected].