Independent Algorithmic Bias Auditing for Enterprise Recruiting
Third-party AI screening tools can quietly filter candidates based on age, race, and disability without ever asking for protected attributes. Fraoula provides independent, continuous mathematical verification across Workday, Greenhouse, and custom ML talent pipelines.
Interactive Four-Fifths (80%) Rule Compliance Calculator
Evaluate adverse impact under Title VII, ADEA (Age 40+), and EEOC Uniform Guidelines in real time.
Protected group is selected at only 61.1% the rate of the benchmark group. Under EEOC guidelines, this establishes prima facie adverse impact liability.
The 4 Pillars of Fraoula Algorithmic Recruiting Governance
Enterprise AI recruiting compliance requires more than annual vendor whitepapers. Fraoula Data Auditor evaluates models as empirical black boxes through continuous statistical hypothesis testing.
Continuous Adverse Impact & 4/5ths Rule Telemetry
Real-time calculation of the Disparate Impact Ratio across age (ADEA 40+), gender, race, and disability. Automated circuit breakers halt automated screening on specific roles if the rolling 30-day DIR falls below safe thresholds (0.85).
Latent Proxy Feature Attribution (SHAP Decomposition)
Machine learning models reconstruct protected attributes through high-dimensional correlates. Fraoula uses game-theoretic SHAP values to identify when features like college graduation year, vintage technology mentions (e.g., COBOL, VB6), or career gaps drive negative scoring weights.
Counterfactual Resume Permutation Auditing
Fraoula programmatically generates synthetic resume pairs where skills, qualifications, and domain experience remain identical while latent proxy tokens (graduation year, high school date, career gap placement) are systematically varied to measure algorithmic score variance.
Rejection Velocity & Off-Hours Anomaly Throttling
Automated algorithms often reject candidates within minutes in the middle of the night (e.g., 3:15 AM). Fraoula tracks rejection velocity and enforces a human-in-the-loop review buffer, ensuring no candidate is discarded by unmonitored batch inferences without human oversight.
Full Regulatory Coverage Across Global AI Hiring Mandates
You cannot outsource Title VII or ADEA liability to software vendors. Fraoula provides the verifiable mathematical paper trail demanded by federal enforcement agencies.
| Regulatory Framework | Governing Body | Compliance Mandate | Fraoula Automated Audit Module |
|---|---|---|---|
| Title VII (Civil Rights Act of 1964) | EEOC | Zero unlawful disparate impact on race, sex, national origin | Automated UGESP 4/5ths Rule continuous monitoring |
| ADEA (Age Discrimination in Employment Act) | EEOC / Federal Courts | Prohibits discrimination against applicants 40 and older | Graduation year proxy redaction & counterfactual age audit |
| ADA (Americans with Disabilities Act) | EEOC / DOJ | Prohibits pre-employment screening that filters out disabilities | Audit of game-based cognitive tests & neuroticism proxies |
| NYC Local Law 144 (AEDT) | NYC DCWP | Mandatory annual independent bias audit published publicly | Certified independent AEDT audit report generation |
| EU AI Act (Annex III) | European Commission | Recruitment AI classified as High-Risk (Art. 10 data governance) | Article 10 bias testing & technical documentation dossier |
Is Your AI Screening System Legally Defensible?
Do not wait for an EEOC subpoena or federal class action to discover proxy bias in your third-party hiring tools. Protect your enterprise with Fraoula Data Auditor.