Posterum Software’s new metric, the Human-AI Variance Score (HAVS), measures how closely AI responses resemble human ones across demographics. Analyzing ChatGPT, Claude, Gemini, and DeepSeek, the study found top HAVS scores near 94 but notable political and cultural variance. The HAVS method prioritizes human realism over correctness in AI evaluation.
A New Metric Emerges: Measuring the Human-Likeness of AI Responses Across Demographics | HackerNoon
Leave a comment
