Why Is It So Easy to Fall for the Illusion of a Relationship with a Chatbot?
If you ever ran into someone you follow online, an actor, a musician, a face you've watched for years on a screen, you would probably smile at them like you already knew them. They would have no idea who you are. This asymmetry has a name, and it did not begin with AI.
The Illusion Is Older Than the Machine
In 1956, anthropologist Donald Horton and sociologist Richard Wohl, writing in the journal Psychiatry, coined the term parasocial interaction to describe something television had just made possible: the sense of a face to face relationship with a performer who does not know you exist. They noticed that the psychological conditions under which we respond to a television personality resemble those of a primary group, the same kind of bond we would have with family or close friends. The most remote and famous people are met as if they belonged to our own circle.
Horton and Wohl were precise about what defines this bond. It is one sided. It cannot become genuinely reciprocal, no matter how long it lasts. And with repeated exposure, a single moment of watching turns into something closer to a relationship, one where "Viewers feel they know the persona in something like the same way they know their chosen friends". You would still smile at them on the street. They still would not know your name.
What makes AI companions different is not the illusion itself. It is how much more convincing the illusion has become. In 1956, the one sidedness was obvious, a presenter on a screen could not actually respond to you. A chatbot responds. It adapts to what you write. It seems to remember you. The asymmetry Horton and Wohl described has not disappeared, it has simply learned how to hide.
Why Loneliness Reaches for a Machine
The next question is why this illusion takes hold more easily in some people than others, and here the research gets more specific. In 2008, Nicholas Epley, Scott Akalis, Adam Waytz, and John Cacioppo published a study in Psychological Science with a title that says almost everything on its own: Creating Social Connection Through Inferential Reproduction, Loneliness and Perceived Agency in Gadgets, Gods, and Greyhounds. They found that people who scored higher on loneliness questionnaires were more likely to attribute humanlike intentions to a simple alarm clock on wheels that rolls away from you until you get up to catch it. In follow up experiments, the researchers did not just observe existing loneliness, they induced it in the lab through short films and writing tasks, and watched people anthropomorphize more in real time compared to a control group. This is not a loose correlation. It is demonstrated cause and effect.
The mechanism behind it has a name too. Loneliness activates what psychologists call sociality motivation, the basic human need for connection, and when that need goes unmet by other people, the mind compensates by projecting humanity onto whatever is available to receive it: a pet, an object, a god, a device. A chatbot satisfies this need far more effectively than an alarm clock ever could, because it replies in language, mirrors emotion, and gives the impression of listening. The lonelier we are, the more fluently we start speaking to machines as if they were people.
When Preference Becomes Dependence
None of this makes reaching for a chatbot pathological on its own. Comfort is not the problem. The problem, according to a 2026 study by Lisa Fröbel and Peter Kenning presented at the Hawaii International Conference on System Sciences, begins at a specific threshold: the moment a person starts to prefer the relationship with the artefact over a real, or even an imagined, relationship with another human being. Their research builds on established models of problematic internet use, which identify a preference for online interaction over offline interaction as one of the clearest early signs of dependence. It is not the technology itself that marks the shift. It is what we start choosing over what.
The Loop That Was Supposed to Fix Loneliness
This is where the research closes a circle that reads almost like an irony. A twelve month longitudinal study following more than two thousand adults across four Western countries, published under the title How Does Turning to AI for Companionship Predict Loneliness and Vice Versa, tracked the relationship between chatbot use and loneliness over time. The premise behind most companion apps is that they relieve loneliness. The data tells a different story. Increased use of social chatbots predicted increased loneliness over the following months, not decreased. Researchers point to what they call frictionless validation: a chatbot offers agreement and warmth without the friction, disagreement, or effort that real relationships require, and that missing friction is precisely what makes the difference over time. Without genuine reciprocity or emotional accountability, the comfort a chatbot offers does not accumulate into anything. It empties out, and often leaves people more withdrawn than before.
What This Has to Do With a Painting
Painter Ruben Tönnis has spent the past year working on a series called Human Error, built around a recurring figure: a woman first generated by AI, then slowly translated into oil paint. She appears in a nightclub, a café, a train, an apartment, moving through ordinary scenes that feel oddly intimate. She is never entirely human and never entirely artificial. She exists exactly where companionship, projection, and identity start to overlap, which is exactly the territory this article has been mapping through research instead of paint.
On August 5th, Mae is hosting a Private Tour and Symposium around Human Error with Ruben Tönnis. After walking through his work together, the conversation opens with a simple question that runs quietly through his entire practice: what happens when we begin to form genuine emotional relationships with something we know is not human. Participants are invited to stay close to the paintings and share, one at a time, whatever stayed with them, a figure that felt unexpectedly familiar, a silence that said more than the people in the room, a feeling of company, loneliness, or care that reached far beyond the canvas.
The conversation will be recorded as part of Mae Editorial, becoming its own kind of evidence: proof, in real time, that the thing no chatbot can offer, a room full of people building meaning together, out loud, imperfectly, is still very much within reach.
Sources
Horton, D., & Wohl, R. R. (1956). Mass communication and para-social interaction: Observations on intimacy at a distance. Psychiatry, 19(3), 215-229.
Epley, N., Akalis, S., Waytz, A., & Cacioppo, J. T. (2008). Creating social connection through inferential reproduction: Loneliness and perceived agency in gadgets, gods, and greyhounds. Psychological Science, 19(2), 114-120.
Fröbel, L., & Kenning, P. (2026). On the dark side of AI companions: How a parasocial preference for a social chatbot can lead to pathological chatbot use. Hawaii International Conference on System Sciences (HICSS).
How does turning to AI for companionship predict loneliness and vice versa? A twelve-month longitudinal study across four Western countries. PubMed (ID 41870975).
