Fairness and oversight

Show the reasoning, and let people answer for it.

Fairness is not solved by hiding a score or adding a human click at the end. The criteria, evidence, gaps, corrections, and decision record all matter.

Updated July 24, 2026

Criteria review

Hiring teams should identify which requirements are genuinely job-related, distinguish required from preferred criteria, and challenge proxy requirements that may exclude people without improving the decision.

Human oversight

Reviewers should see supporting context and uncertainty before acting. Human review must be substantive: a person can disagree, request clarification, correct evidence, and document a different conclusion.

Candidate agency

Candidates should be able to understand relevant profile information, correct it, control appropriate visibility, and raise concerns about inaccurate or unfair use.

Auditability

A production workflow should preserve the criteria used, evidence shown, material changes, reviewer identity, decision and reason, within privacy and retention limits.

Measurement gaps

The public repository does not yet provide validated fairness-study results. Aptaryn should publish measurement methods and limitations before presenting quantitative fairness claims.

Review the Responsible AI policy