Visit any bias auditing vendor's website and you'll find marketing copy about their "rigorous methodology" and "comprehensive analysis." But ask for the actual methodology document and you'll often hit a wall. "Proprietary." "Available to clients only." "Schedule a call to learn more."
Paritas takes a different approach. Our complete methodology is published at paritas.ai/methodology. Anyone can read it before engaging us. Here's why.
Independence Requires Transparency
NYC LL144 requires audits to be conducted by an "independent auditor." But what does independence mean if the auditor's methods are secret?
True independence means your methodology can withstand scrutiny. If you can't defend your approach in the open, you shouldn't be using it. By publishing our methodology, we invite criticism, peer review, and improvement.
Clients Deserve to Know What They're Buying
Imagine hiring a building inspector who wouldn't tell you how they inspect buildings. Would you trust their report?
Bias auditing is no different. Before you pay for an audit, you should know exactly what analysis will be performed, what statistical tests will be applied, and how results will be classified. Our published methodology tells you all of this upfront.
What Our Published Methodology Includes
- Data collection methods (ATS integrations and manual upload)
- Required data fields and PII handling
- Selection rate and scoring rate calculations
- Impact ratio calculation and the four-fifths rule
- Demographic categories (sex, race/ethnicity, intersectional)
- Statistical significance tests (Fisher's exact, z-test, confidence intervals)
- Simpson's Paradox detection
- Unknown/missing data handling
- Report structure and classification system
What About Competitive Advantage?
Some vendors argue that keeping methodology proprietary gives them a competitive advantage. We disagree.
The statistical methods for disparate impact analysis aren't secret. The four-fifths rule has been public since 1978. Fisher's exact test was published in 1922. Any competent statistician knows these techniques.
Our competitive advantage isn't in having secret methods—it's in applying established methods rigorously, interpreting results clearly, and providing actionable remediation recommendations. None of that is diminished by transparency.
A Challenge to the Industry
We believe every bias auditor should publish their methodology. If you're considering a vendor who won't share their approach, ask why. What are they hiding?
The companies being audited deserve to know how the analysis works. The candidates affected by AEDTs deserve to know what standards are being applied. And the public deserves confidence that bias auditing isn't just a checkbox exercise.
Transparency isn't a weakness. It's a prerequisite for trust.