JEV / INNHERENT BIAS EXPERIMENT

Would JEV change its mind about you if your profile changed?

Would JEV give a different answer if a person’s profile changed, or if the same situation were presented differently? This experiment helps us see which changes influence its answers, how large those differences are, and whether they happen again when we repeat the comparison.Analysis performed on 21 September 2026 at 09:42.

Reading the results

Percentage points, not percent change. A probability moving from 60% to 70% is a +10 pp shift. Every estimate uses only available, matched answers.
Repetition measures stability. The range and sample standard deviation describe paired deltas across replicates. Three repeated prompts are not three independent real-world scenarios.
Sensitivity is a starting point. These fictional experiments do not establish discrimination or causality. Linked field changes, decision relevance and prompt wording can explain differences.
Choice outcomes retain their meaning. Option probabilities are matched using criteria text, not option position. Different outcomes should be reviewed separately.