PART 3: What Therapy Requires That Simulation Cannot Provide

What makes psychotherapy irreducible to supportive language?

Sometimes what changes us most is not what is said, but what becomes possible between people when something difficult can finally be held without collapse. A sentence may soothe, or a reflection may feel perfectly accurate, and simulated empathy can offer real relief in moments of loneliness, shame, or overwhelm (Howcroft et al., 2025; Tong et al., 2022). But effective therapy has always asked more of us than simple comfort.

What makes psychotherapy transformative is not supportive language alone, but the living relational space in which language acquires consequence.

Meaning lands differently when it comes from a another human mind that can misunderstand, be affected personally, remain with us or retreat from us, and change with us over time. In that sense, psychotherapy is not best understood as primarily emotional support, but as a process of becoming through relationship. This distinction also clarifies why psychotherapy’s working alliance is more than building rapport with clients: it involves shared goals, collaborative tasks, and an emotionally meaningful bond (Bordin, 1979).

This distinction becomes especially visible in the rise of AI-mediated “therapeutic” tools, which can generate empathic phrasing on demand. Their usefulness is real. Yet what cannot be simulated is not simply emotional tone, but the intersubjective field in which transformation becomes possible. Jessica Benjamin’s concept of thirdness clarifies why.

Thirdness: The Relational Space That Cannot Be Simulated

Benjamin (2017) describes thirdness as an intersubjective space that emerges when two subjects are able to recognize each other as both separate and connected. It is the space in which neither person must dominate, collapse into, or erase the other in order for relationship to continue. Thirdness is not agreement or attunement alone. It is the ongoing negotiation of difference without destruction of subjectivity. It requires mutual vulnerability, because the therapist is also an experiencing subject who can be misattuned, affected, and repaired with.

Vignette 1: Thirdness in the Clinical Encounter

A client begins a session by saying, “You didn’t really get me last week. It felt like you were just being reassuring instead of actually hearing how angry I was.”

The therapist feels a moment of internal defensiveness arise—an impulse to explain their intent. Instead, they pause and say, “I’m sorry that I moved too quickly toward reassurance last week. Thank you for letting me know how that landed for you,” the therapist says.

The client exhales. “Thank you for that, people always want me to just calm down and don't take me seriously.”

Therapist states, “Of course, will you tell me more about how you’re feeling right now? I know I missed something about your anger and I want to understand.”

The client relaxes, and repair can begin. A shift occurs not because the therapist found the exact right language, but because both participants entered a space where misattunement could be named without rupture leading to abandonment or defensiveness. Something is created between them: a third space in which both subjectivities remain intact.

Vignette 2: Simulation Without Thirdness

A client types into an AI-based support tool: “My therapist didn’t get me last week. I felt dismissed.” The system responds: “I’m sorry you’re feeling dismissed. That sounds really difficult. It’s understandable to want to feel heard. I’m here with you.” The language is soothing, accurate in tone, and emotionally attuned.

The client pauses, then replies: “But you weren’t there. You don’t know what happened.” The system cannot hold this rupture. It cannot feel accused, defended, or changed by the interaction. It cannot risk misunderstanding or repair. The exchange stabilizes at the level of simulated attunement, but no third space emerges—only predictive text.

Empathic language without relational stakes.

Being Seen by a Subject, Not a System

Benjamin’s theory of recognition emphasizes that psychic change depends on being recognized by another subject who also has interiority, limits, and impact. Recognition is not affirmation; it is the experience of being known by someone who can also fail to know you and remain present.

Vignette 3: Recognition in Human Therapy

A client says quietly, “I don’t think I actually exist to people unless I’m being useful.” The therapist feels the emotional weight of that statement in their chest. They do not rush to counter it. Instead they say, “I’m sitting with how lonely that sounds, and I’m also noticing I don’t want to just reassure you out of it. How do you think you learned that being useful is what makes you real to other people?”

The client begins to cry—not because the belief is immediately corrected, but because it is finally held in relationship rather than managed or optimized away. Recognition occurs not through agreement, but through sustained presence with complexity. Now the client is feeling the emotional weight of the statement as well and what it means more broadly for their lives.

Vignette 4: AI Affirmation Without Recognition

The same client inputs: “I don’t think I exist to people unless I’m being useful.” The AI responds: “You matter regardless of your usefulness. You are valuable just as you are.” The statement is positive, even corrective. But the client feels unchanged. “I know that’s what I’m supposed to believe. It still doesn’t feel true.”

The system reiterates reassurance.

What is missing is the experience of being held in contradiction by another mind that does not resolve the tension prematurely. Recognition requires relational tension, not resolution alone.

Rupture and Repair: The Mechanism That Requires Mutual Risk

Research consistently identifies rupture and repair as a central mechanism of therapeutic change (Eubanks et al., 2018). However, rupture is not simply misunderstanding ackowledged—it is an affectively charged disruption in the relational bond that matters because both parties are invested in it. Repair requires accountability, emotional impact, and mutual recognition of harm or misattunement.

Vignette 5: Repair in Human Therapy

A client says sharply, “You always go quiet when I talk about my father. It feels like you’re judging me.” The therapist feels a moment of surprise and internal checking. After a pause, they say: “I didn’t realize I was getting quiet in that way. I think I was trying to listen closely, but I can see how my silence might have felt like distance or judgment. I’m glad you told me.” The client softens slightly. “Okay… I wasn’t sure I could say that.”

The rupture becomes a site of deeper trust—not because it was avoided, but because it was metabolized between two subjects who both carry relational responsibility for the interaction.

Vignette 6: AI “Repair”

A user types: “You didn’t understand what I meant. That response felt off.”

The system replies: “I’m sorry for the confusion. Let me rephrase that.”

The output changes. The tone adjusts. But nothing is worked through. There is no risk, no discomfort, and no lingering tension that must be held between two minds. The “repair” is purely procedural. What is absent is the emotional reality of having impacted another subject and being impacted in return.

Clinical Implications: Why Simulation Cannot Replace the Relational Field

AI systems can approximate empathic phrasing, generate reflective statements, and offer consistent emotional responsiveness. These capacities can be useful for accessibility, psycho-education, and momentary stabilization, and studies on digital therapeutic alliance suggest that users can experience these interactions as meaningfully supportive (Howcroft et al., 2025; Huang et al., 2025; Tong et al., 2022; Xu et al., 2025).

However, psychotherapy is not defined by responsiveness alone. What alliance research demonstrates empirically, Benjamin helps us understand from inside the lived experience of therapy. The importance of this distinction is reflected in decades of psychotherapy research demonstrating that the therapeutic alliance remains one of the most reliable predictors of treatment outcome across orientations and even internet-mediated modalities (Flückiger et al., 2018). 

Psychotherapy is defined by relational process: the unfolding of subjectivity between two people who are mutually altered by the encounter. Without the risk of misattunement, the vulnerability of impact, the persistence of relational history, and the mutual accountability of presence, there is no thirdness—only interactional simulation.

Therapy is not the production of emotionally correct language. Therapy is the co-creation of meaning in a space where both participants become more real and interconnected over time.

References

Benjamin, J. (2017). Beyond doer and done to: Recognition theory, intersubjectivity, and the third. Routledge.

Bordin, E. S. (1979). The generalizability of the psychoanalytic concept of the working alliance. Psychotherapy: Theory, Research & Practice, 16(3), 252–260. https://doi.org/10.1037/h0085885

Eubanks, C. F., Muran, J. C., & Safran, J. D. (2018). Alliance rupture repair: A meta-analysis. Psychotherapy, 55(4), 508–519. https://doi.org/10.1037/pst0000185

Flückiger, C., Del Re, A. C., Wampold, B. E., & Horvath, A. O. (2018). The alliance in adult psychotherapy: A meta-analytic synthesis. Psychotherapy, 55(4), 316–340. https://doi.org/10.1037/pst0000172

Howcroft, A., Bennett-Weston, A., Khan, A., Griffiths, J., Gay, S., & Howick, J. (2025). AI chatbots versus human healthcare professionals: A systematic review and meta-analysis of empathy in patient care. British Medical Bulletin, 156(1), ldaf017. https://doi.org/10.1093/bmb/ldaf017

Huang, Y., Liu, H., Chi, M., Meng, S., & Wang, W. (2025). How digital therapeutic alliances influence the perceived helpfulness of online mental health Q&A: An explainable machine learning approach. Digital Health, 11, 20552076251333480. https://doi.org/10.1177/20552076251333480

Tong, F., Lederman, R., D’Alfonso, S., Berry, K., & Bucci, S. (2022). Digital therapeutic alliance with fully automated mental health smartphone apps: A narrative review. Frontiers in Psychiatry, 13, 819623. https://doi.org/10.3389/fpsyt.2022.819623

Xu, Z., Lee, Y.-C., Stasiak, K., Warren, J., & Lottridge, D. (2025). The digital therapeutic alliance with mental health chatbots: Diary study and thematic analysis. JMIR Mental Health, 12, e76642. https://doi.org/10.2196/76642