picked.
Log inPost a role
Trust · Fairness
Monitored per role. Per stage. Per protected group.
The four-fifths rule is the floor. We report against tighter thresholds. We pause syndication when a flag fires.
See the model cards
Methodology
Five protected pairs. Disclosed voluntarily. Post-decision only.

We monitor adverse impact across five pairs: sex (female / male), race (non-white / white), age (over-40 / under-40), education (non-graduate / graduate), and disability disclosure. Categories are voluntary, declared by the candidate post-decision (so the disclosure cannot influence the decision), and used only for aggregate adverse-impact monitoring.

  • ●Voluntary self-identification. Candidates can decline every question.
  • ●Post-decision only. Disclosure is captured after the candidate has been advanced or rejected.
  • ●Aggregate only. Per-role and per-stage aggregates. Never used as an input to scoring.
  • ●4/5ths rule is the regulatory floor (EEOC). We report against tighter thresholds where the role volume supports it.
Redress
Adverse impact is a stop-the-line event.
01
Detection. Per-role drift detection runs daily. The 4/5ths ratio is recalculated on every candidate decision; a breach pauses syndication for that role.
02
Notification. The hiring manager and Picked's internal review team are both notified within minutes. The candidate dashboard shows a banner explaining the pause.
03
Investigation. Item-level differential item functioning (DIF) analysis runs on the assessment items used. Any item showing DIF above threshold is retired.
04
Resolution. Once the calibration is corrected and the role re-passes the 4/5ths threshold, syndication resumes. Time to resolve is published quarterly.
Related
TRUST · COMPLIANCE
Compliance
Eight frameworks across four jurisdictions. Per-framework posture.
Read →
TRUST · MODEL CARDS
Model cards
One per model in production. Capability, training, evaluation, limits.
Read →
TRUST · RESPONSIBLE AI
Responsible AI
Six principles. Eight candidate rights. The lines we will not cross.
Read →
Last reviewed · 22 May 2026 · v1.0
Still on the legal review?
We'll send you the file.
Pre-filled vendor security questionnaire (SIG Lite), DPA, sub-processor list, SOC 2 letter, model card, latest bias audit. One zip. Reply within one business day.
Request the legal pack →Email legal@picked.ai ↗
Operated by Neuroworx Ltd · ICO #09910326623
picked.
The AI-native replacement for inbound hiring. For the hiring manager, not HR.
© 2026 Picked · Made in London · Three finalists in forty-eight hours.
Fairness · Picked.ai