Fairness & Bias Audit

Algorithmic Parity Guard Active

Algorithmic adverse impact ratio monitoring, four-fifths rule (80% benchmark) verification, and disparity mitigation across protected demographic cohorts.

shield_lock

Protected Demographic Disclosure & Blind Screening Enclave: Demographic attributes are compiled solely via voluntary EEOC/OFCCP applicant surveys. Under screenCV isolation protocols, identifying cohort fields are held in a cryptographically sealed hardware enclave and never exposed to automated rubric scoring nodes, hiring managers, or interviewers.

lock AES-256 Enclave

warning

Active Disparity Alert · Four-Fifths Rule Threshold Breach Requisition Level

Adverse Impact Ratio is 0.74 for Female applicants on Senior Backend Engineer (#REQ-1042). Selection rate: 14.8% vs. 20.0% benchmark group. Statutory limit is ≥ 0.80.

Scope:

expand_more

Period:

expand_more

Audited 12m ago (n=1,420) · hash: 0x9e88...b214

Lowest Adverse-Impact Ratio Org Safe Harbor

0.86 EEOC Threshold: ≥ 0.80

Organization-wide parity remains compliant at 0.86. Single localized threshold breach currently isolated to Requisition #1042.

Protected Groups Monitored n ≥ 30 Rule

14 Groups 5 EEOC/OFCCP Classes

8 Standard 4 Underrepresented 2 Small Sample (<30)

Open Disparity Breaches Action Required

1 Active 0 Systemic breaches

Requires rubric weight calibration History (3 resolved) chevron_right

Selection Rate by Demographic Cohort vs. Benchmark

Percentage of applicants advancing past algorithmic resume evaluation. Selection rate = Advancing Candidates ÷ Total Applicants.

Benchmark / Compliant

Breach (< 0.80 Ratio)

Insufficient Data (n<30)

Protected Cohort

6.25%

12.5%

0.80 Benchmark Threshold (16.0%)

20.0% (Baseline)

25.0%

Male n=240 · Benchmark Group

20.0% Selection Rate

Ratio 1.00 (Base)

Non-Binary / Third Gender n=40 applicants

17.5%

Ratio 0.88 · Pass

Undisclosed / Decline n=110 applicants

19.2%

Ratio 0.96 · Pass

error Female

n=162 applicants

14.8% Falls 1.2% below 4/5ths threshold (16.0%)

Ratio 0.74 ⚠

Indigenous / Native n=8 applicants

Sample size too small for statistical validity (n = 8 < 30 required by EEOC 29 CFR §1607)

n < 30 (Muted)

info EEOC 4/5ths Rule: Adverse impact is indicated if selection rate for any group is less than 80% (0.80) of highest group rate.

Statistical Confidence Level: 95% (α = 0.05)

Disaggregated Adverse Impact Cohort Ledger

Statistically verified applicant counts, selection frequencies, and Fisher's exact test significance scores.

10 Protected Cohorts

Protected Cohort & Category	Applicants (n)	Advanced (n)	Selection Rate	Impact Ratio	4/5ths Rule Status	Fisher's Exact (p)
Male (Gender / Sex)	240	48	20.0%	1.00	Benchmark (Baseline)	N/A (Reference)
Female (Gender / Sex)	162	24	14.8%	0.74	warning Adverse Impact (< 0.80)	p = 0.042 (Sig.)
Non-Binary / Third Gender (Gender / Sex)	40	7	17.5%	0.88	Compliant (≥ 0.80)	p = 0.410 (Not sig.)
Asian (Race / Ethnicity)	195	38	19.5%	0.98	Compliant (≥ 0.80)	p = 0.520 (Not sig.)
Black or African American (Race / Ethnicity)	114	21	18.4%	0.92	Compliant (≥ 0.80)	p = 0.280 (Not sig.)
Hispanic or Latino (Race / Ethnicity)	88	15	17.0%	0.85	Compliant (≥ 0.80)	p = 0.190 (Not sig.)
White (Race / Ethnicity)	210	42	20.0%	1.00	Benchmark (Baseline)	N/A (Reference)
Native Hawaiian / Pacific Islander (Race)	12	2	—	—	Insufficient data (n < 30)	Inconclusive
American Indian / Alaska Native (Race)	6	1	—	—	Insufficient data (n < 30)	Inconclusive
Age 40 and Over (ADEA Cohort)	138	26	18.8%	0.94	Compliant (≥ 0.80)	p = 0.340 (Not sig.)

Formula: Impact Ratio = Group Rate ÷ Benchmark Rate · Confidence Threshold: p < 0.05

Showing 10 of 14 monitored cohorts

verified_user

NYC Local Law 144 Independent Audit Ready

Audit Grade A-

Next Annual Regulatory Audit Due: Dec 15, 2024 · Independent Auditor of Record: Dana Vance (Lead Counsel) · Merkle Root: 0x7f4a...9b12