Candidate Scoring

Candidate scoring refers to AI systems that assign numerical values to candidates, enabling ranking and threshold-based selection decisions. Scores can be based on resume analysis, skills assessments, personality tests, video interviews, or combinations of multiple inputs. Candidate scoring introduces specific bias risks because the scoring function itself may weight features unequally across demographic groups. Even when the same scoring model is applied to all candidates, it can produce systematically different score distributions for different groups if the underlying features or their weights correlate with protected characteristics. Bias auditing of scoring systems requires not only comparing selection rates (who passes the threshold) but also analyzing the full score distributions across groups. Two groups might have similar means but different variances, or similar medians but different tails — and these distributional differences can interact with thresholds to create adverse impact. OnHirely analyzes score distributions as part of its audit process, identifying both threshold-based and distributional sources of bias.