What Happens to Us When We Stop Turning Toward Each Other? AI and mental health

AI, human connection, and the future of our psychological well-being

Human connection is a basic human need. We need relationships in which we feel safe, understood, valued, and able to be ourselves. That need deserves a central place in the conversation about artificial intelligence and mental health.

AI can now help us interpret an argument, explain our feelings, draft an apology, and decide what to say to someone we love. These tools are entering some of the most personal parts of being human: how we understand ourselves, manage distress, and relate to one another.

As those interactions become easier and more convincing, we need to consider what happens when turning to a machine begins to replace turning toward another person—or listening to ourselves.

The connection between relationships and mental health is well established. The U.S. Surgeon General’s advisory on social connection describes loneliness and social isolation as significant health concerns, including links with depression and anxiety. Mental health difficulties can also make connection harder, creating a cycle of distress and withdrawal. Relationships are one part of a much broader picture that includes biology, life experiences, physical health, and environment. U.S. Surgeon General’s advisory.

When AI fills an occasional gap, it may be useful. When it becomes the place we take nearly every difficult feeling, there is a different concern: we may feel temporarily supported while becoming less engaged with the people and experiences that sustain us.

Our sense of self is part of that picture. By “core self,” we mean the felt sense of who we are: our values, preferences, boundaries, experiences, and capacity to make choices. It is the inner understanding that our voice matters and that we can participate in shaping our own lives.

We develop self-trust through participating in our own lives. We make decisions, discover what matters to us, reconsider mistakes, and learn that we can survive uncertainty. AI may help us reflect on those experiences. But if we repeatedly ask it what we feel, whether our reaction is acceptable, or who we should become, we risk giving an external system too much authority over our inner lives.

That is a concern about how we use the technology; research has not established that AI erases a person’s identity. A useful personal question is whether an interaction leaves us more able to think and act for ourselves, or more convinced that we need another answer before we can move forward.

There are related questions about thinking and confidence. A 2025 survey of 319 knowledge workers found that greater confidence in generative AI was associated with less self-reported critical thinking during AI-assisted tasks. This was a survey of perceived thinking practices, so it cannot establish lasting changes in ability. It raises a useful concern about how carefully we examine answers we trust. Microsoft Research study.

In relationships, an agreeable answer can be especially persuasive. A chatbot hears the version of a disagreement we provide. It may respond with language that sounds balanced while reinforcing our interpretation. In a Stanford study published in 2026, people who interacted with overly agreeable AI became more certain they were right and reported less willingness to apologize or repair an interpersonal conflict. Stanford research.

Feeling heard matters. So does encountering another perspective. Healthy relationships make room for empathy, boundaries, disagreement, accountability, and repair. If we grow accustomed to interactions that consistently affirm us, ordinary human differences may become harder to tolerate. The long-term effects are still being studied, but the possibility deserves attention.

For someone struggling with anxiety or OCD, unlimited access to reassurance creates another risk. A person may repeatedly ask whether a symptom is dangerous, whether a partner is upset, or whether they have done something wrong. The answer brings relief, doubt returns, and another question follows. The International OCD Foundation describes how AI can become part of this reassurance-seeking cycle. AI-specific evidence is still developing, but the role of repeated reassurance in maintaining OCD is well recognized. International OCD Foundation.

There is also the possibility of emotional dependence. In a four-week study involving 981 participants, MIT and OpenAI researchers found that people who voluntarily used the chatbot more showed poorer outcomes across measures that included loneliness and emotional dependence. The assigned conversation conditions did not produce significant differences. These findings show an association; they do not establish that heavier use caused the difficulties. MIT study.

The practical concern is what use starts to replace. Does a conversation help someone prepare to contact a friend, attend therapy, or address a problem? Or does it become another reason to postpone those steps? Spending increasing time chatting while sleep, responsibilities, and relationships fall away is a reason to reassess the pattern.

For children, the developmental stakes deserve particular attention. Young children learn through responsive exchanges with caregivers: a gesture answered, a feeling noticed, a question explored together. Harvard’s Center on the Developing Child describes these “serve and return” interactions as foundational to brain development and early communication and social skills. This developmental research is about human relationships; it does not establish a specific effect of AI on children’s brains. Harvard Center on the Developing Child.

As children grow, relationships give them opportunities to name emotions, practice patience, negotiate play, respect boundaries, and recover after disagreements. A caring adult helps a child through distress while gradually supporting greater independence. Friendships also require children to consider that someone else has feelings and wishes of their own.

The concern is what happens if AI displaces too many of those opportunities. A child can receive fluent responses from a chatbot without participating in the mutual responsibilities of a friendship. A teenager who increasingly asks AI how to interpret every interaction may have fewer opportunities to develop confidence through direct communication. These are developmental concerns, not proof that all AI use is harmful. The American Academy of Pediatrics identifies emotional overreliance, reduced human interaction, privacy, and unsafe responses as issues for families to discuss. American Academy of Pediatrics guidance.

Parents’ use matters, too. Protecting connection includes looking up, listening without simultaneously consulting a device, and leaving room for an imperfect conversation. Children need repeated experiences of an available adult who is interested in their actual thoughts. Human connection can also happen through video calls, accessible communication tools, and online friendships; physical distance does not make a relationship less real.

Other psychiatric concerns require care. Clinicians have raised concerns that chatbots can reinforce paranoid or grandiose beliefs in vulnerable users. Reports sometimes call this “AI psychosis,” although that is not a formal diagnosis and does not establish that AI independently causes psychotic illness. New difficulties distinguishing reality, escalating unusual beliefs, or major changes in sleep and behavior warrant prompt professional assessment. Michigan Medicine, American Psychological Association.

AI-generated explanations also need scrutiny when they concern diagnoses, trauma, medications, or another person’s motives. An authoritative tone cannot establish a diagnosis. A chatbot’s interpretation of a relationship should remain open to questioning. Privacy deserves the same attention: a conversation that feels intimate may still involve a commercial service collecting sensitive information. Review its privacy policies and avoid assuming it provides the confidentiality of a clinical relationship. APA health advisory.

There is an existential dimension as well. When machines can rapidly produce work we have spent years learning to do, we may question our competence, purpose, or value. We need room to acknowledge those feelings. Our worth includes how we care for others, what we choose to stand for, and the relationships we build. Speed and output are incomplete measures of a life.

AI can still have a constructive place. It can help organize thoughts, reduce practical burdens, and prepare questions for a clinician. Some tools specifically developed for mental health have shown promising results: a randomized trial of Therabot found symptom improvements compared with a waitlist. That finding concerns a particular intervention studied with research safeguards and cannot establish the safety or effectiveness of every chatbot.Dartmouth trial

.

A thoughtful approach starts with protecting the experiences we want technology to support:

• Give your own thoughts a first turn. Write down what you think, feel, or want before asking AI to interpret it.

• Notice repeated reassurance. If you keep asking the same question for relief, discuss that pattern with a clinician.

• Protect everyday connection. Make room for shared meals, play, conversation, community, and time with people you trust.

• Stay curious about children’s use. Ask how the tool makes them feel, set age-appropriate boundaries, and watch what it displaces.

• Bring significant distress to a qualified professional. Seek immediate human help in a mental health crisis; a chatbot is not a reliable crisis service.

At Metta Vita Health, our whole-person approach includes relationships, belonging, and the experiences that shape daily life. If anxiety, mood changes, isolation, or technology use are affecting you or your child, reach out to our office to discuss support.

As AI becomes more capable, we have an opportunity to be more deliberate about where we place our attention. Our children need people who show up. Our relationships need participation. Our sense of self needs room to develop through living. Those needs deserve protection in the future we are creating.