Bias Audit Summary
NYC Local Law 144 Compliance
Audit Pending
This page will be updated following completion of the first independent bias audit. Cognaium conducts quarterly internal bias reviews and annual independent audits per our AI Governance Policy.
Audit Details
Independent Auditor
[Auditor Name — To Be Appointed]
Audit Date
[Pending — First audit scheduled]
Audit Period
[Will cover preceding 12 months of assessment data]
Tool Audited
Cognaium AI Assessment Scoring System
Methodology
The bias audit evaluates the AI assessment system for disparate impact using the following approach:
- Selection Rate Comparison: Calculate selection (pass) rates for each demographic group at each assessment stage
- Four-Fifths Rule: Apply the 4/5ths (80%) threshold — the selection rate of any group should not be less than 80% of the highest-selected group
- Demographic Parity Difference: Measure the absolute difference in selection rates between groups
- Equalized Odds Analysis: Evaluate true positive and false positive rates across groups
- Intersectional Analysis: Examine outcomes at the intersection of multiple protected characteristics where sample sizes permit
Selection Rate Results
By Race/Ethnicity (EEO-1 Categories)
| Category | Selection Rate | Impact Ratio | 4/5ths Pass |
|---|---|---|---|
| Data will be populated after first audit cycle | |||
By Gender
| Category | Selection Rate | Impact Ratio | 4/5ths Pass |
|---|---|---|---|
| Data will be populated after first audit cycle | |||
Questions About This Audit
For questions about our bias audit methodology, results, or to request the full audit report, contact [email protected].