HireLens

Docs · Understand

Blind review & bias audits

Two mechanisms, one goal: decisions based on evidence, and evidence that your decisions are fair.

Blind review

One toggle in the review queue masks identity cues: name, contact details, schools, graduation years. The implementation matters as much as the feature:

  • Masking happens at the presentation layer with character alignment preserved — masked text keeps the same length and offsets as the original.
  • That means evidence spans (which are character offsets into the document) still land exactly. Blind mode doesn't break the product's core promise.

Demographics: opt-in, isolated, never in scoring

Bias auditing needs group data, and group data is sensitive. HireLens takes the only defensible position:

  • Candidates self-report, per dimension (e.g. gender), entirely opt-in.
  • Reports live in an isolated table that is never joined into scoring queries — the schema makes it structurally impossible for demographics to influence a score.
  • People who don't report are bucketed as “undisclosed” in audit reports. Not knowing who you rejected is itself an audit finding, and hiding it would defeat the purpose.

The adverse-impact audit

One click computes, per group and per dimension:

  • Selection rate — advanced-or-shortlisted ÷ considered.
  • Impact ratio — each group's rate ÷ the reference group's rate (or the overall rate when no majority group is defined).
  • The four-fifths rule — ratios below 0.80 are flagged. The math uses exact integer arithmetic, so 4/5 passes cleanly instead of failing on a floating-point 0.7999…
  • Demographic parity difference — the largest gap between group rates.

What the audit is — and isn't

Statistical screening metrics, not a legal conclusion. HireLens helps you meet audit obligations; it does not make you compliant.

A flagged ratio is a prompt to investigate, not proof of discrimination. A passing report is not a shield. Sample sizes matter: with five people in a group, one decision moves the rate by 20 points. Read the numbers with that in mind, and involve people who know the law in your jurisdiction (NYC LL144, the EU AI Act, and state laws each have their own duties).

Export

Every audit report can be exported as JSON — and the export itself is written to the hash-chained audit log, so running an audit is on the record too.