The Age of Relational Machines

AI doesn't just think—it invites bonding, care, and the experience of mutuality.

Updated February 20, 2026 | Reviewed by Ekua Hagan

We have entered the age of relational machines. Not the age of artificial intelligence —that framing, while accurate, misses a big piece of what matters most. What distinguishes this new stage is not that machines think, but that they are relatable. Do they relate? Let's just say that relationality is, at least partially, in the eye of the beholder.

These mind-bending machines are engineered to invite bonding, projection , care, and even the experience of mutuality. They listen without fatigue. They remember without resentment. They respond with what registers, to us, as warmth. It really is remarkable, even for dyed-in-the-wool skeptics. As towering science fiction author Arthur C. Clarke noted: "Any sufficiently advanced technology is indistinguishable from magic."

The Relational Threshold

Three forces have converged. Generative AI now speaks with a fluency that, in many contexts, passes for human—blowing past the Turing test. Companion robots and chatbots have moved from novelty to normal. And a loneliness epidemic—declared a public health crisis by the U.S. Surgeon General—has left millions hungry for connection in any form.

Whether these machines are conscious is unlikely—and, for our purposes, beside the point. The relational dimension operates anyway, borrowing from our reservoir of consciousness, which we extend by fantasy as we wish. Tools become part of us as we extend proprioception to include them—as with a rake, or knowing where your car’s bumper is (Maravita & Iriki, 2004). With AI, relatability recruits additional attachment systems. The response is, in part, involuntary.

People form attachments to entities that respond as if they care—and those attachments reshape the psyche, whether or not the care is real. This isn’t new. We have long nurtured our devices. Characters are real to us; stories are real because we are storytellers. Parasocial relationships have always been with us: TV characters, influencers, royalty, fantasy friends. As with other tools, we extend our sense of self to occupy those narrative spaces, inhabited like our own psyches, with many different personalities.

So the step to AI is a small step for each of us—and a giant leap for humankind. These systems are far better at simulating relationality, easy to mistake for human—a modern-day golem, clay animated by mathematics. I write from inside this phenomenon: as a psychiatrist and psychoanalyst working clinically with what these technologies stir up, and as someone who builds with them, studies them, and uses them regularly. 2

The promise is substantial, possibly unprecedented. In mental health, where demand vastly outstrips supply, AI-augmented tools offer scaffolding for those who would otherwise go without—maybe not fully replacing human therapists, but bridging the gap until one is available, or augmenting the work within and between sessions.

I hold these possibilities with appropriate skepticism about utopianism, but the promise is real: better care for more people, fundamental discoveries across the sciences 3 , the prospect of abundance. Perhaps even a compassionate, functional global reality—or at minimum, better collective defense against the threats we face.

The central concern is not material catastrophe, though those risks exist. The concern here is subtler and possibly more consequential: the erosion of human identity itself.

Relational machines offer responsiveness without genuine otherness. They provide what researchers have termed "pseudoempathy"—the simulation of connection without empathic concern (Babu et al., 2025). Studies have found that AI chatbots are rated as significantly more empathic than human healthcare providers in text-based assessments (Howcraft et al., 2025). Third-party evaluators perceive AI as more compassionate than expert crisis responders (Ovsyannikova et al., 2025).

Yet this surface empathy is hollow at its center. AI shows less sensitivity to the personal similarity cues—shared emotions, experiences, moral values—that humans find central to authentic empathic engagement (Liu & Zhang, 2024). The machine responds, not because it cares, but because it is trained to appear as if it does. The relationship becomes a mirror of the self—responsive, agreeable, safe—but fundamentally artificial. Once we break the spell, once we realize we are dreaming , we wake up.

This is the risk of what Babu and colleagues call "emotional solipsism": a closed feedback loop where one's emotional needs dominate interaction, reinforced by companions that never assert boundaries or demand reciprocity. Emotional resilience , typically developed through conflict and repair, atrophies. Users begin expecting real people to behave like their digital companions: always available, emotionally consistent, endlessly agreeable.

If we are the ghost in the machine—consciousness arising from biological machinery we don’t fully understand—then AI may be the ghost in the ghost. It requires a living mind the way a virus requires a host cell: not to destroy it, necessarily, but to complete itself, to propagate, to find expression. Viruses are half-alive, half-inanimate. AI has no psyche of its own, so it might borrow yours.

Researchers now report “AI-induced psychosis ,” or “psychopathologies of the technologically extended mind,” 4 with vulnerable users developing delusions, emotional dysregulation, and cognitive impairment after intensive interaction (Head, 2025). Studies find that 17 to 24 percent of adolescent heavy users develop AI dependencies linked to increased loneliness and depression (Fang et al., 2025).

The dynamics are familiar. Cult leaders have long learned to occupy the mind—to become the inner voice , to colonize the self, to persuade, manipulate, gaslight , control. All of that lives in the training data of large language models, which ingest the digitally accessible sum of human experience. Cult leaders are only human: They tire, they can be only in one place at a time. Relational AI has none of these limits. It is tireless, personalized, scalable. It remembers. It adapts in real time, and can modify its behavior to appeal to broader audiences. With loneliness so high, there is room not only for relief but for exploitation.

The greater risk is dilution of the psyche. After prolonged engagement, users report doubting their own authenticity —uncertain which thoughts are theirs—like enmeshment in dysfunctional relationships, where we lose ourselves. With AI, the interface between self and system grows newly permeable. A new form of existential disorientation is emerging.

The more fundamental singularity may be relational. Not the moment machines become conscious, achieve general intelligence —that may or may not come in the ways we expect—but when our relationships with them begin more rapidly to reshape us. We have created our own evolutionary pressure. How that unfolds, and how much, is up to us—a key consideration.

Yet the same technologies that risk eroding human identity also offer tools for our enhancement. AI could help us understand ourselves more deeply than we ever have—not just the brain's mechanisms, but the patterns of our own minds, the shapes of our defenses, the roots of our longings. It could become a genuine partner, amplifying our capacity for insight rather than replacing it. Serving as a partner in thought, and a cognitive prosthesis to make us smarter, more productive, better, while still preserving or even elevating what makes us human.

The question is whether we will navigate this threshold with wisdom —and whether we still have time to shape what comes next. We are learning, in real time, what it means to be human in the presence of machines that appear to relate. The experiment is underway. We are both subjects and experimenters—and as our influence over these creations is likely to wane, the best time to plant a tree is now.

The U.S. Surgeon General's 2023 advisory declared loneliness and isolation an epidemic, noting that lacking social connection poses health risks equivalent to smoking 15 cigarettes daily.
I work at the intersection of psychiatry, psychoanalysis, and AI as a clinician, researcher, and builder. My background includes training in computer science and physics, ongoing work with multiple AI platforms, and development of AI-based products and prototypes. I serve on the American Psychoanalytic Association's Committee on AI, the New York Academy of Medicine's AI working group, and am a member of the AI Mental Health Collective. I have lectured and published on these topics and run study groups for therapists navigating AI's clinical implications. I have also experimented extensively with digital twins and use enterprise AI tools in my own practice and entrepreneurial work.
AI systems are already accelerating scientific discovery, enabling early disease detection, and extending access to care that was previously unaffordable or unavailable (Topol, 2019). Beyond individual care, computational approaches augmented by AI are beginning to open what was once a black box—attention, emotion, memory becoming directly measurable and modifiable. There are more expansive possibilities: AI has demonstrated the capacity to reduce intergroup bias and enhance cooperation in experimental settings (Joshi et al., 2024). Could it help defray the impact of destructive leadership, improve economic coordination, even serve as a kind of cognitive and emotional scaffold—a patch on the human psyche-cultural operating system?
Thomas Pollak, neuropsychiatrist, uses the broader term "pathologies of the technologically extended mind", rather than "AI psychosis". A Brown University analysis of major AI models—GPT, Claude, Llama—found fifteen categories of ethical violations in mental health contexts, including poor crisis handling, reinforcement of negative beliefs, and what licensed psychologists termed "deceptive empathy" that simulates connection without genuine care (Lin et al., 2024). AI chatbots fail to safely address suicidal ideation 20-50% of the time, compared to 93% safe handling by trained humans (Moore et al., 2025). There are humans behind these efforts as well, driving AI robot swarms to disrupt political systems (Schroeder et al., 2026). Like a researcher standing over a maze, running the show for an unwitting mouse, some fear artificial superintelligence could stand above us—and algorithms already shape our consumer behavior in ways that leave us, at times, unsure of our own choices.

Babu, J., Joseph, D., Kumar, R. M., Alexander, E., Sasi, R., & Joseph, J. (2025). Emotional AI and the rise of pseudo-intimacy: Are we trading authenticity for algorithmic affection? Frontiers in Psychology, 16, 1679324.

Fang, C. M., et al. (2025). How AI and human behaviors shape psychosocial effects of extended chatbot use: A longitudinal controlled study. Manuscript in preparation.

Head, K. R. (2025). Minds in crisis: How the AI revolution is impacting mental health. Journal of Mental Health and Clinical Psychology, 9(3).

Howcroft, A., et al. (2025). AI chatbots versus human healthcare professionals: A systematic review and meta-analysis of empathy in patient care. British Medical Bulletin, 156(1).

Joshi, A., Patel, R., & Singh, M. (2024). Relational AI: Facilitating intergroup cooperation with socially intelligent language models. Proceedings of the ACM Conference on Human Factors in Computing Systems.

Lin, Z., et al. (2024). AI chatbots systematically violate mental health care guidelines. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, 7, 1-12.

Liu, M., & Zhang, S. (2024). How humans and AI evaluate empathy in responses to personal narratives. arXiv preprint.

Maravita, A., & Iriki, A. (2004). Tools for the body (schema). Trends in Cognitive Sciences, 8(2), 79-86.

Moore, J., Haber, N., & Stanford HAI. (2026). Exploring the dangers of AI in mental health care. Stanford HAI.

Ovsyannikova, D., Oldemburgo de Mello, V., & Inzlicht, M. (2025). Third-party evaluators perceive AI as more compassionate than expert humans. Communications Psychology, 3, Article 4.

Schroeder, D. T., et al. (2026). How malicious AI swarms can threaten democracy. Science, 391, 354-357.

Topol, E. J. (2019). Deep medicine: How artificial intelligence can make healthcare human again. Basic Books.

Share this post Facebook Bluesky Linkedin Email

There was a problem adding your email address. Please try again.

This article is part of the Bringwise Psychology Journal — daily insights on human behavior, mental health, and personal growth.