NYC Bias Audit

Service · Audit & assessment · HR & recruiting by Holistic AI

Third-party bias auditing for NYC Local Law 144 compliance in AI-driven hiring.

NYC Bias Audit is a third-party assessment service for organizations to comply with New York City Local Law 144 regarding automated employment decision tools. It evaluates hiring algorithms for gender and racial bias using selection rates and impact ratios to ensure fair and transparent recruitment.

Category: RegTech & Compliance

Compliance Target
NYC Local Law 144 (NYC LL144)
Framework Used
JobFair Framework
Primary Metrics
Selection Rate (SR) and Impact Ratio (IR)
Bias Threshold
0.8 Impact Ratio
Task Coverage
Deterministic and Generative tasks
Assessment Types
Intrinsic and Extrinsic bias evaluation
  • Third-party verification for NYC Local Law 144 (NYC LL144) compliance
  • Calculates Selection Rates (SR) and Impact Ratios (IR) for hiring tools
  • Utilizes the JobFair framework to analyze resume scoring bias
  • Evaluates both intrinsic and extrinsic biases in recruitment algorithms

Overview

NYC Bias Audit is a specialized third-party assessment service provided by Holistic AI for organizations using automated employment decision tools. The service is designed to help businesses comply with New York City Local Law 144 (NYC LL144) by evaluating hiring algorithms for biases related to race, gender, and other protected characteristics. It utilizes the JobFair framework to analyze selection rates and impact ratios, providing the necessary verification for talent management systems. The audit addresses both intrinsic biases within model associations and extrinsic biases that manifest in practical applications like resume ranking. By calculating specific metrics such as the 0.8 impact ratio threshold, Holistic AI ensures that HR technology is transparent and legally compliant. This service is used by talent solutions companies, staffing agencies, and human capital management providers to validate their AI-driven recruitment tools.

Bias Assessment Metrics

  • NYC Local Law 144 Readiness: The audit service evaluates automated employment decision tools to ensure they meet the requirements of New York City Local Law 144.
  • Selection Rate Calculation: Assessments calculate Selection Rates (SR) to determine the proportion of a demographic group selected by the algorithm.
  • Impact Ratio Analysis: The service determines Impact Ratios (IR) by comparing selection rates between different demographic groups.
  • Bias Threshold Detection: Audits identify potential bias when an Impact Ratio falls below the 0.8 threshold.

Algorithmic Fairness Testing

  • JobFair Framework Analysis: The audit distinguishes between taste-based bias, which is consistent, and statistical bias, which varies with information levels.
  • Resume Density Evaluation: Assessments evaluate how the amount of detail in a resume affects the extent of bias exhibited by large language models.
  • Intrinsic Bias Detection: The service identifies internal associations in models, such as linking specific roles with gendered pronouns.
  • Extrinsic Bias Evaluation: Audits detect biases that occur during practical use, such as rating candidates differently based on demographic traits for specific roles.

Compliance and Reporting

  • Third-Party Verification: The service provides the necessary third-party verification for companies using AI in recruitment and talent management.
  • HR System Accountability: Audits help organizations establish transparency and accountability in their human capital management systems.
  • Downstream Task Reporting: The assessment covers bias detection in both deterministic and generative downstream tasks.

What's included

  • Automated employment decision tools (AEDT)
  • Hiring algorithms and talent management systems
  • Selection rates and impact ratios for protected characteristics

Deliverables

  • Third-party bias audit report
  • Compliance verification for NYC Local Law 144

Use cases

  • HR technology providers conduct third-party audits to verify their talent solutions are ready for NYC Local Law 144 compliance.
  • Staffing agencies analyze selection rates and impact ratios to detect potential bias in their automated resume ranking systems.
  • Enterprises evaluate the impact of resume density on LLM scoring to distinguish between taste-based and statistical biases in hiring.
  • Recruitment teams use the audit to identify intrinsic biases in model associations, such as gendered stereotypes linked to specific job titles.

Specifications

Industries served
Talent Solutions & Staffing; Human Capital Management; HR Technology Providers
Compliance Target
NYC Local Law 144 (NYC LL144)
Framework Used
JobFair Framework
Primary Metrics
Selection Rate (SR) and Impact Ratio (IR)
Bias Threshold
0.8 Impact Ratio
Task Coverage
Deterministic and Generative tasks
Assessment Types
Intrinsic and Extrinsic bias evaluation

FAQ

What does resume density mean in the context of hiring bias?

Resume density refers to the level of detail provided in a resume. It can impact bias in LLMs, where statistical bias may vary depending on the amount of information provided, while taste-based bias remains consistent.

What is the difference between intrinsic and extrinsic bias?

Intrinsic bias involves internal model associations, like linking "nurse" with female pronouns. Extrinsic bias occurs in practical use, such as a hiring tool rating male candidates higher for roles traditionally associated with men.

How does human feedback contribute to bias in LLMs?

Human feedback can introduce bias if raters prefer responses aligning with specific cultural or ideological views. The reward model in Reinforcement Learning from Human Feedback (RLHF) then learns to prioritize those perspectives.

Added by AntSpace from public sources (holisticai.com).

Related

More services by Holistic AI