Disclaimer: This article is for informational purposes only and does not constitute legal advice. Customers should consult legal counsel to ensure compliance with employment, AI, privacy, and other applicable laws, and provide candidates with a notice at data collection, outlining their practices (e.g., as required under NYC Law 144).
Why Spark Hire conducts regular bias audits
At Spark Hire, fostering diversity, equity, and inclusion is central to our mission. We understand the critical role AI plays in shaping hiring decisions and strive to ensure our tools promote fairness. While bias audits are essential for upholding these values, they also help us stay aligned with evolving laws.
By maintaining transparency and adhering to best practices, we aim to build trust in our AI tools and support equitable hiring practices across all organizations that use it.
AI Resume Review overview
The AI Resume Review tool is designed to rank candidates' qualifications based on specific job requirements. It assigns an overall score to each candidate, assisting recruiters in the evaluation process. Importantly, the tool does not autonomously reject or hire applicants; all final employment decisions are made by human reviewers.
💡 Tip: Learn more about the AI Resume Review for managers and recruiters.
Scope of use
The tool is employed during the recruitment process to evaluate resumes and rank candidates for various positions.
Data used
The AI Resume Review tool processes the information provided in candidates' resumes and the corresponding job description and requirements. To maintain fairness in the evaluation process, AI Resume Review was designed to minimize the use of Personally Identifiable Information (PII) as much as possible.
Compliance measures
In adherence to New York City's Local Law 144 of 2021, the following measures are implemented:
Bias audit
The tool undergoes an independent bias audit annually to assess its impact on individuals based on race, ethnicity, and sex. The audit evaluates the selection or scoring rates among different demographic groups to ensure fairness.
Summary of results
🗓️ Date of the most recent bias audit: October 9, 2024
Source and explanation of the data used
Spark Hire conducted a bias audit with a third-party auditor, where the processing of candidates' data via the AI Resume Review tool (candidate ranking model) was reviewed, including the score from 1 to 5 reflecting their fit for the job.
The dataset included 14,053 candidates across 12 real jobs (low-skilled, skilled, and management roles). After filtering for self-reported gender and race/ethnicity, 9,469 candidates remained. Due to technical issues, such as unreadable resume file formats, 584 candidates were excluded. Three jobs with more than 1,000 candidates were capped at 1,000 for sampling purposes, resulting in a final dataset of 5,154 candidates. Categories with less than 2% representation were excluded due to small sample sizes.
Number of individuals assessed in unknown categories
4,584 candidates had unknown demographic data.
Selection or scoring rates and impact ratios
By race/ethnicity:
Ethnicity | Number of applicants | Scoring rate | Impact ratio |
Asian | 1,464 | 0.547 | 1.000 |
Black/African American | 896 | 0.452 | 0.826 |
Hispanic/Latino | 412 | 0.500 | 0.914 |
Two or more races | 262 | 0.500 | 0.914 |
White | 2,087 | 0.474 | 0.866 |
By gender:
Gender | Number of applicants | Scoring rate | Impact ratio |
Female | 2,711 | 0.489 | 0.977 |
Male | 2,410 | 0.500 | 1.000 |
By intersectional categories (gender * race/ethnicity):
Gender | Ethnicity | Number of applicants | Scoring rate | Impact ratio |
Female | Asian | 610 | 0.487 | 0.825 |
Female | Black/African American | 594 | 0.460 | 0.779 |
Female | Hispanic/Latino | 245 | 0.514 | 0.871 |
Female | Two or more races | 162 | 0.494 | 0.837 |
Female | White | 1,100 | 0.500 | 0.847 |
Male | Asian | 854 | 0.590 | 1.000 |
Male | Black/African American | 302 | 0.437 | 0.741 |
Male | Hispanic/Latino | 167 | 0.479 | 0.812 |
Male | Two or more races | 100 | 0.510 | 0.864 |
Male | White | 987 | 0.445 | 0.754 |
Public disclosure: The results of the most recent bias audit are made publicly available on the Spark Hire's website. This transparency allows candidates and the public to review the tool's compliance with fairness standards.
Have more questions? Contact us at support-ats@sparkhire.com