One of the strangest things about AI is that we keep insisting it is not human, and then we judge it like a person anyway.
This came through strongly in the lesson summary on how people perceive AI. The psychology is very familiar. People judge others on warmth and competence. Are you capable? And also, are your intentions good? Apparently we bring the same mental machinery to AI, which is both understandable and slightly unfair to the machine, since it has no feelings and still somehow gives off “competent but cold intern” energy.
The warmth part matters more than people think. Cuddy and colleagues’ warmth-competence model suggests that people often judge intentions before ability. In AI terms, this helps explain why a technically strong system can still be rejected if it feels uncaring, opaque, or misaligned with the user.
Mind perception theory adds another layer. People perceive AI through agency and experience. If the AI seems highly agentic, people may think it is more capable. But that also creates a risk: when it fails, the failure may feel less like a malfunction and more like betrayal. A broken calculator is annoying. A confident AI assistant that misleads you after sounding helpful feels personal, even when it is not.
This is why adoption is not only a model-quality problem. It is a perception problem. Transparency, reliability, tangibility, immediacy, and even how human-like the system appears can change how people trust it.
That does not mean we should make every AI sound like a warm therapist trapped in a browser tab. Please no. It means organisations need to design AI interactions with psychology in mind.
My reflection is this: people do not adopt AI because it is powerful in the abstract. They adopt it when it feels useful, understandable, and safe enough to rely on. Competence opens the door. Trust decides whether people walk through.