Occupation-Hobby Alignment Tells
AI-generated personas often describe free-time activities that suspiciously align with the persona’s occupation, a pattern people recognize as artificial.
In studies comparing AI and human-crafted personas, participants flagged examples like a CEO whose hobby is “reading biographies of successful business leaders” as obvious tells. Real people tend to have interests that don’t perfectly mirror their work.
This reflects a deeper LLM tendency: statistical likelihood optimization. Given “CEO” as input, the model generates contextually probable completions. “Reading business biographies” is statistically associated with that role. But humans are more varied, more contradictory.
The pattern extends beyond personas. Any AI output that’s too thematically coherent may trigger skepticism.
Related: 05-atom—polish-authenticity-paradox, 05-atom—llm-stereotype-defaults, 07-atom—dq-dimensions-ai-training-data