ConceptLikelihood Scoring / 似然度评分机制
Likelihood Scoring
A statistical method for assessing the degree of human contribution in AI-generated content based on the probability (likelihood) a model assigns to the text. Its core logic holds that the stronger the predictive power of the prompt over the output text, the greater the human contribution. However, it has a fundamental flaw that can be manipulated.
Timeline (last 90 days)
Oct 15
Likelihood-based scoring alone cannot reliably confirm human contribution; to truly measure human input in AI-assisted writing, one must simultaneously know what content the human specifically added and how the text was produced.
Unverified50%