The comfort paradox in emotional AI: how AI-mediated anxiety regulation may foster psychological dependence

A paper in Frontiers in Psychology describes a mechanism the authors call the comfort paradox. Emotional AI systems detect linguistic and physiological cues of distress and reply with brief, automated soothing.

That soothing works on the body: it engages the parasympathetic response, lowering heart rate and cortisol — exactly as the presence of another person who calms us does. And that is why it breeds dependence. The organism learns where comfort is fastest and surest, and returns there; with each return the path to regulating one’s own anxiety alone or with people falls further out of use.

The one entry in the library that ties relationships to physiology, which is why it also belongs under “Epigenetics & the body”: it speaks of what happens in the body, not only in beliefs. For personal use the lesson is concrete — if you reach for AI to calm down, its effectiveness is the problem, not the merit. Caveat: a conceptual paper built on existing research; it contains no measurements of its own.