Intelligence

ai

Applied and Filtered: An End-to-End Algorithmic Fairness Audit of A Public Employment Agency

Source
arXiv — Computers and Society
Published
Last verified
14 Aug 2026
Confidence
High
Evidence
Original document retained
Reading time
1 min
Country
International
Relevant to
Technology & Data, Research & Evidence, Operations & Delivery, People & Capability, Risk & Compliance

Executive summary

What happened, and why should leadership care?

A research paper from arXiv details an independent, end-to-end algorithmic fairness audit of a semi-automated hiring system used by Barcelona Activa, a public employment agency. The audit, covering approximately 497,000 candidate-vacancy pipeline entries over five years, evaluated the third-party TalentClue platform. While aggregate outcomes across binary genders appeared statistically indistinguishable, this parity was found to mask substantial underlying disparities within the system.

Why this matters

Why is this strategically important?

This research highlights that relying solely on aggregate fairness metrics can be misleading, as underlying biases and disparities may persist within complex algorithmic systems. For organizations deploying AI in critical functions, it underscores the necessity of comprehensive, granular audits to ensure true equity and mitigate hidden risks.

Key insights

What should be noted from the evidence?

  • An independent, end-to-end algorithmic fairness audit was conducted on a semi-automated hiring system used by a public employment agency (Barcelona Activa).
  • The audited system involved the third-party TalentClue platform for candidate search and shortlisting.
  • The analysis covered approximately 497,000 candidate-vacancy pipeline entries from September 2017 to September 2022.
  • Seven pipeline stages were examined, encompassing automated processing, human discretion, candidate data, and employer decisions.
  • Aggregate outcomes across binary genders were statistically indistinguishable, implying overall fairness at a high level.

Evidence and confidence

How far can this assessment be trusted?

High confidence. Named institution, original document retained and analysis corroborated.

Analysis is prepared editorially by Aziz Shuaib Ausi. The original publication remains the authoritative record, and executive judgement remains entirely human.

Source

Where does this originate?

Reported by arXiv — Computers and Society · International. This briefing summarises the publication for executive use; the document itself is not reproduced here.

Read the original publication