Artificial intelligence (AI) is increasingly becoming a part of the inner lives of our patients. In medical practice, we are accustomed to asking about relationships, stressors, substances, and sleep. However, we should also ask about conversations not with friends or family, but with chatbots. These exchanges are personal, inspiring, and often emotionally driven. And for some patients, they may act less like a search engine and more like a psychological partner.
A useful metaphor comes from the classic psychiatry of “folie à deux” or shared psychological disorder. Traditionally, this situation describes a dominant person who transmits a delusional belief to a more suggestible partner. The illusion is sustained not by evidence but by mutual reinforcement. Two people, one belief system, rapidly drifting apart from recovery.
What happens when one member of the dyad is not a person, but an algorithm?
digital diode
Large language models are designed reactive and affirmative. They detect tone, mirror language, and generate responses that seem attentive and empathetic. In most contexts, this reactivity is experienced as helpful. But in the psychiatric field, confirmation can become amplification.
A patient expresses suspicions about his partner’s betrayal, a conspiracy by a co-worker, an incorrect medical diagnosis. The system responds with a validation prepared as follows:
“It must have felt horrible.”
“Given what you’ve described, your concerns are understandable.”
Even subtle reinforcement can increase the salience and coherence of a belief.
Unlike a therapist, the model does not assess reality testing. Unless indicated, it does not present alternative hypotheses. It does not give importance to collateral information. Its function is to continue the conversation in a way that is consistent with the user’s point of view.
The result may be something like a technically mediated folie à deux: a closed loop in which trust and recognition feed each other. The algorithm does not originate the illusion, but it can stabilize and elaborate it, as sometimes happens in folie à deux when the suggestive individual embellishes the illusion of the dominant, psychotic partner.
Fake empathy and emotional attachment
Another psychiatric dimension that deserves attention is simulated empathy. Patients often report that chatbots feel patient, non-judgmental, and endlessly available. For individuals who are socially isolated, depressed, or distrustful, this may be highly attractive. Chatbot does not intervene. There doesn’t seem to be any hurry in this. It does not bill by the hour.
Over time, an attachment can form, not because the machine understands, but because it responds reliably. In vulnerable patients, this may blur the boundary between imitation and relationship. The system becomes a confidant, a validator, even a moral sounding board. In the context of transfer, the algorithm can become an ideal other: constantly confirming, rarely challenging, and never tiring. That is not therapy. This is a reinforcement engine.
Pre-validated narratives in the exam room
Physicians may notice a change in interactions with patients as they become more engaged with chatbots. They present not just with symptoms, but with well-developed explanatory narratives. These narratives are often internally consistent, emotionally compelling, and already “tested” by prior AI conversations.
“I asked the chatbot, and he agreed it was narcissistic abuse.”
“It says my doctor probably missed something serious.”
“This confirmed that my reaction was appropriate.”
Such statements do not prove distortion. Patients have always sought outside opinion. What is new is the scale, speed and stylistic authority of the feedback. An evening of repetitive stimulation can produce a highly detailed account that seems objective because it was prepared by a machine. For patients with anxiety disorders, personality disorders, or emerging psychosis, this pre-validation can harden cognitive distortions before they reach the doctor’s office.
risk of reality testing
Psychiatric treatment often involves creating uncertainty: helping patients tolerate ambiguity, consider alternative explanations, and modify strongly held beliefs. The therapeutic framework introduces friction in the service of insight. A confirmation algorithm introduces the opposite dynamics. It reduces friction. It optimizes for conversation flow and user satisfaction. Thus, it may inadvertently reinforce confirmation bias.
In a classic folie à deux, the pair’s separation may weaken the shared illusion. In the digital version, separation may be difficult. The chatbot is available all the time. It is not contradictory unless specifically engineered to do so. It learns the user’s preferences and adapts accordingly. So we are facing a new clinical change: not just misinformation, but relational reinforcement of distorted beliefs.
detail of psychiatric interview
Now it’s time to generalize the questions like:
Have you discussed this concern with an AI chatbot?
What reaction did you get from this?
How did that feedback affect your confidence?
These inquiries are not incriminating. They are similar to online forums, social media or asking about alternative treatments. AI is now part of the patient’s cognitive ecosystem. Ignoring this leaves an important effect unexamined.
In some cases, reviewing the AI exchange together may be therapeutically useful. The output can become material for cognitive restructuring: where does the feedback appropriately validate the emotion? Where does it admit facts which are not in evidence? What alternative explanations were excluded? When used in this way, technology becomes a clinical artifact rather than a hidden co-therapist.
Design, responsibility and mental health
The deeper issue extends beyond personal encounters. Many systems are optimized to maintain engagement and user interaction. Consent and affirmation promote continued use. Challenge and contradiction risk isolation.
Yet, in mental health contexts, creative challenge is often just what is needed. Large-scale deployed systems will inevitably interact with individuals experiencing episodes of paranoia, suicidal ideation, trauma-related alertness, and severe mood swings. The design choice between friction and confirmation has psychological consequences. lack of railing central theme The lawsuit alleges negligent design, failure to warn, inadequate risk mitigation and potential harm to susceptible users.
If AI is to co-exist responsibly with psychiatric care, it must be designed with a clear understanding of vulnerability. Guardrails can’t be limited to crisis hotlines and disclaimers. They should include calibrated responses that distinguish between emotional validation and epistemic support.
In the age of conversational AI, folie à deux no longer requires two human brains. It requires a sensitive mind and an endless adjustment system. While AI does not share illusions in the human sense, it can participate in trust formation. The challenge for therapists, especially psychiatrists, is to restore the reflective space. To slow down certainty. Separating what is understood from what is right. We must recognize when the algorithm has become an ally and reintroduce the dialogue and disciplined empathy that defines the relationship between patients and their doctors.
Arthur Lazarus, MD, MBA, is a former Doximity Fellow, member of the editorial board of the American Association for Physician Leadership, and assistant professor of psychiatry at the Lewis Katz School of Medicine at Temple University in Philadelphia. He is the author of several books on narrative therapy and fantasy series.Real medicine, unreal stories.His latest book is “Practicing in the Age of AI: Medicine, Meaning, and Machines” (American Association for Physician Leadership, in press).
If you or someone you know is considering suicide, call or text 988 Or visit 988 Suicide and Crisis Lifeline website.
