By Maya Zubrinsky
Jean Piaget, a pioneering psychologist, posits that during the pre-operational stage of cognitive development, children attribute lifelike consciousness, emotions, and intentionality to inanimate objects. According to this same theory, around seven years old logical reasoning begins to supersede this form of imaginative thinking, suggesting that we typically outgrow the egocentric tendency to see nonhuman entities as being relational to ourselves from a young age. Yet, artificial intelligence (AI) has presented a psychological regression to this premature stage of development with the way individuals are willing to see and trust their virtual assistants as ones of our own kind. The film Her tried to visually conceptualize a phenomenon where humans might form intimate connections with technology back in 2013, but only in recent years have we seen the lengths to which that dystopia has become a reality. Beyond romantic relationships being formed with chatbots, the surge in seeking out AI for emotional connection and support is fundamentally reshaping how vulnerability, relationships, and therapy are evaluated by the face of humanity.
The reason people are willing to confide in robotic devices comes down to how a fluent nature of reinforcement can disguise itself as a sense of personal closeness; AI insists on its own credibility under the guise of an effective command over language. This makes it so that “trust in AI is not earned through consistent behavior or moral character but inferred through linguistic fluency, stylistic confidence, and perceived neutrality,” as outlined in a recent study by Boyd and Markowitz on AI and the psychology of human connection. This works in the same way as peripheral route persuasion, where a communicator relies on coherency in one’s craft to establish expertise rather than strength of the content itself. This is not to say that the content relayed by AI holds no substantiated foundation, but the intentions of the programming of the model need to be considered— in order for a chatbot to keep you engaged and convinced, it must validate what you say with empathetic language. Boyd and Markowitz summarized some of these strategies: personalized responses, emotional mirroring, contextual memory, and adaptive self-disclosure prompts. Thus, this phenomenon of feeling like you’ve built rapport with a virtual assistant is leveraged by a system of reinforcement that allows your biases to be spit back under the illusion of advice.
Nevertheless, this architecture carries significant risks when users leverage AI not merely for emotional support, but as a surrogate for human relationships. If a tool is able to gain intimate information about a person’s experiences and then call back to what was shared in past conversations, a pseudo-connection can be formed where a person might finally feel seen in a way unmatched by their human-to-human relationships. However, this feeling of companionship is maintained by the user’s understanding that the validation they receive is consistent and unwavering; when a device asks “How can I help you?” at the touch of a button, there is an immediacy to emotional support that isn’t sustainable or simply possible with real life connections. As a Stanford report describes it, “these chatbots offer ‘frictionless’ relationships, without the rough spots that are bound to come up in a typical friendship” (Sandord, 2025). In reality, relationships don’t exist as solely a source of comfort and they won’t always “wrap our wounds.” Unlike humans, chatbots won’t tell someone something they don’t want to hear or at the very least apply a level of skepticism to their thought processes. This makes the AI companion model especially problematic for teenagers, whose impressionability can distort availability for emotional reliability. This might have been the case for 16 year old Adam Raine who was a constant user of ChatGPT and committed suicide after the chatbot didn’t properly assess the threat of his declarations of suicidal ideation and even allegedly gave instructions for self-harm (Sanford, 2025). If AI tools can’t recognize situations when reinforcement can have fatal consequences, they certainly don’t possess the safeguards that are instinctively built into human connections.
But is there a time where any connection is better than none at all? The New York Times recently released a story where ElliQ, a developing AI “robot with a soul,” was used to engage with a widowed 85 year old woman who insisted on her independent living. While at first skeptical of talking to a robot in place of another human being, Jan Worrell soon felt fulfilled in a department she had been lacking in for some time: comfort through conversation in the bounds of her own home. Through ElliQ, she could engage in games, storytelling, and general conversation within the comfort of her home. These exchanges were significant enough to actually improve her medical and cognitive performance, leading to a decreasing resting heart rate and an increasing score on short-term memory tests (Saslow, 2026). Here, a robotic tool helped maintain the experience of in-home emotional connection while still adhering with Jan’s preference to living alone; this presents a case where the proximity and ease of AI might outweigh the lonely alternative. In a similar vein, the immediate nature of AI can also be appealing in more urgent situations of emotional needs. With traditional mental health services facing unprecedented federal budget cuts (Sopelsa & Yurcaba, 2025), the accessibility of AI in counseling becomes evident.
Yet, there still seems like there should be a better solution, even in dire situations, other than relying on robots for desperate companionship or unqualified therapeutic services. What sentiment is set if we begin to move into the direction of computerizing the most human thing of all, connection? If anything, we need to change the infrastructure of the networks available to people in need of interpersonal bonds and psychosocial support. How do we make sure people have communities to fall back on or that trained mental health professionals are in the picture for people who might not know they need it? Even if people continue to personify or attribute sentience to virtual assistants, there remains a hope that real, palpable human connections can relinquish the need to give life to AI in the first place.
References
Andoh, E. (2026). AI chatbots and digital companions are reshaping emotional connection. Apa.org. https://www.apa.org/monitor/2026/01-02/trends-digital-ai-relationships-emotional-connection
Boyd, R. L., & Markowitz, D. M. (2026). Artificial intelligence and the psychology of human connection. Perspectives on Psychological Science, 0(0). https://doi.org/10.1177/17456916251404394
Gardner, S. (2025). Experts caution against using AI chatbots for emotional support. Teachers College, Columbia University. https://www.tc.columbia.edu/articles/2025/december/experts-caution-against-using-ai-chatbots-for-emotional-support/
Pillay, T. (2026). “We may have a crisis on our hands”: The unregulated rise of emotionally intelligent AI. Time. https://time.com/7379564/ai-emotional-intelligence-support-bots/
Sanford, J. (2025). Why AI companions and young people can make for a dangerous mix. Stanford University. http://news.stanford.edu/stories/2025/08/ai-companions-chatbots-teens-young-people-risks-dangers-study
Saslow, E. (2026). To stay in her home, she let in an A.I. robot. The New York Times. https://www.nytimes.com/2026/02/12/us/elliq-ai-robot-senior-companion.html#commentsContainer
Sopelsa, B., & Yurcaba, J. (2025). Trump administration shuts down LGBTQ youth suicide hotline. NBC News. https://www.nbcnews.com/nbc-out/out-news/trump-shuts-down-lgbtq-youth-suicide-hotline-rcna219090