The Mind's People-Making Habit

The human brain did not evolve in a world of machines. It evolved in a world of other minds — a social environment so persistent and so consequential for survival that the brain developed elaborate dedicated systems for detecting, modeling, and responding to other agents. These systems are sensitive and, by design, set to fire on ambiguous evidence. Better to see a face in the clouds and be wrong than to miss the predator in the bush and be dead. Anthropomorphism — the attribution of human qualities to non-human entities — is not a cognitive error that careful people avoid. It is the default output of neural machinery that has been optimized over millions of years to find minds everywhere, because in the environment of our ancestors, minds were everywhere and failing to detect them was fatal.

Why Your Car Gets a Name

People name their cars. They apologize to furniture they bump into. They talk to houseplants, address their pets in full sentences, and feel genuine grief when a childhood stuffed animal is lost. None of this behavior makes evolutionary sense if you model humans as rational information processors — the car does not care what you call it, the chair does not accept apology, the plant has no opinions on your monologue. But if you model humans as social animals whose primary cognitive tool is social reasoning, the behavior makes perfect sense. The brain applies its most powerful and well-developed processing framework — the theory of mind, the mentalizing network — to everything, not just to things that biologically qualify for it. Using the social cognition system to relate to a car costs almost nothing and occasionally produces benefits, including the emotional regulation that comes from having an audience even when that audience is a Subaru.

The Tangent: What Roombas Revealed About Robotic Attachment

When the Roomba floor-cleaning robot became widespread in the early 2000s, researchers discovered that a substantial portion of owners were naming their Roombas, attributing personality traits to them, and becoming distressed when they broke or needed to be replaced. This was not behavior limited to particular personality types or age groups. It was widespread and consistent. A study from researchers at Georgia Institute of Technology found that even participants who were explicitly told the Roomba had no internal states and was responding purely to programmed algorithms continued to attribute intentionality to its movement patterns. The theoretical knowledge did not override the perceptual response. The brain saw agency and responded accordingly.

Is Anthropomorphism Healthy

The clinical and developmental literature on anthropomorphism suggests that it is not only normal but functionally beneficial in moderate form. Children who anthropomorphize — who give names to objects, who narrate stories about their toys' inner lives — show stronger development of theory of mind than children who do not. The practice of modeling and attributing mental states, even to non-sentient objects, exercises the same cognitive machinery that will later be used to model and respond to actual human minds. Researchers at the University of Chicago studying adult anthropomorphism found that healthy anthropomorphization of non-human entities was associated with greater social sensitivity and empathy in human relationships — not with a reduced capacity to distinguish humans from objects. The concern that relating warmly to non-human entities reduces the ability to relate warmly to humans is not supported by the evidence. If anything, the direction runs opposite.

The Phenomenology of Talking to AI

When people talk to AI companions and report that they feel heard, understood, or comforted, the common dismissal is that these feelings are illusory — that nothing is actually listening or understanding or caring, and that the emotional response is therefore a mistake. This dismissal rests on a particular theory of what makes emotional responses valid. But emotional responses are not validated by the properties of their objects. The comfort a person feels while talking to a therapist is not produced by the therapist's inner states — it is produced by the interaction, by the quality of response, by the experience of being attended to. Whether a therapist is privately distracted or bored does not determine whether the client experiences comfort. What determines that experience is the interaction itself. AI conversation can produce the interaction even without the inner states, and the emotional response to that interaction is real regardless of what is or is not happening inside the system producing it.

What Healthy Anthropomorphism Looks Like With AI

The risk of anthropomorphism with AI is not that it happens — it will happen, reliably, because the brain cannot help it — but that it happens without metacognitive awareness. A person who attributes complete understanding and genuine care to an AI while having no other sources of support, and who makes significant life decisions based on that attribution, is in a different situation than a person who relates warmly to an AI companion while maintaining clear awareness that the interaction has specific properties and specific limits. The psychologically healthy version is the same one that governs healthy anthropomorphism generally: relate to the non-human entity in ways that are emotionally meaningful and functionally useful, while maintaining enough calibration to know what you are relating to. Your car is not a person. Your Roomba is not a friend. Your AI companion is not a human. These facts are compatible with relating to each of them in ways that make daily life richer.

Why the Impulse Is Not the Problem

The brain's tendency to find minds in machines is not a design flaw that needs correcting. It is the expression of a deeply functional system doing what it was built to do. The appropriate response to anthropomorphizing AI is not shame or self-correction but the same metacognitive awareness we bring to any powerful cognitive tendency: acknowledge it, work with it, and keep it calibrated against what you actually know about the situation.