In 2021, U.S. Equal Employment Opportunity Commission (EEOC) launched an agency-wide initiative to ensure that the use of software, including artificial intelligence (AI), machine learning, and other emerging technologies used in hiring and other employment decisions comply with the federal civil rights laws that the EEOC enforces. On January 31, the EEOC held a public hearing on the use of how artificial intelligence and algorithmic employment decision-making interact with federal civil rights laws such as Title VII, the Americans with Disabilities Act (ADA) and the Age Discrimination in Employment Act (ADEA).
The meeting was led by the EEOC Commission which includes Chair, Charlotte Burrows; Vice Chair, Jocelyn Samuels and Commissioners, Keith Sonderling and Andrea Lucas. The hearing included witness testimony on the various aspects of use of AI in employment, including the benefits and potential unintended outcomes that may result in violations of civil rights. The witnesses included a range of stakeholders who were experts in the field of AI and/or thought leaders on AI and civil rights such as the US Chamber of Commerce, AARP, various tech and law academics, attorneys (e.g., ACLU, plaintiffs’ bar and employment and labor counsel) and Industrial and Organizational Psychologists. However, notably absent from the witness list were any AI vendors and employers who currently use AI systems in their workplaces.
The overall tone of the hearing was that the EEOC understands AI tools can be useful but also acknowledges that they may often be flawed, and without proper guidance and oversight that use of these tools could create biases in employment decisions. The central objective of the hearing (based on witness testimony and the EEOC’s line of questioning for the witnesses) was determining how the EEOC can provide meaningful standards and/or technical assistance for employers who use AI in employment while ensuring its mission of protecting civil rights. Below are some of the reoccurring themes from the witness testimony which may be relevant in understanding what guidance may be issued by the agency, which includes but is not limited to:
- Transparency
- Validation of AI Prior to Deployment
- Auditing of AI Tools
- Education/Raising Awareness on “UGESP”
Transparency. To create trust in the use of AI systems, it must be understood what these tools do and how they do it. Due to the lack of inherent objectivity in AI tools, it is imperative that employers, vendors and government agencies understand and disclose exactly how these technologies work. In most cases, candidates and employees are unaware that these technologies are being used to make decisions about them. A reoccurring question was whether candidates and employees should know that they are being assessed by an AI tool, and what types of disclosures regarding the use of AI will be meaningful?
Testing of AI Prior to Deployment. Unique to AI’s functionality is its needs to be allowed to take in many pieces of data that are related to protected classifications in order to learn. Collection of this type of information is counter to the employment practice of not collecting information on protected classes to reduce bias and discrimination claims. With AI, there is a fine line in that AI systems need protected class information in order to de-bias these systems by permitting this information to be analyzed in order to produce less discriminatory effects. Determining what the balance is of protected information needed for machine learning without violating civil rights will be key. To ensure that these systems work as intended before being deployed, they must also be tested and adjusted accordingly.
Auditing of AI Tools. There was a consensus that requiring an audit of AI tools will be a necessary component of any standards or guidance. However, there are lingering questions about to whom the duty to audit should fall on? The EEOC likley lacks the resources and technical expertise to conduct audits and permitting employers to self-audit can be problematic for a number of reasons. Third-party audits seem to be the likely outcome, with the possibility that the EEOC may set vendor standards for third-party auditors to ensure their integrity and to provide some oversight of any potential ongoing relationship between the third-party auditor and the employer.
Education/Raising Awareness on UGESP. There was an acknowledgment that some of the concerns related to civil rights and the use of AI for employment tests and selection procedures may already be examined using the framework of the 1978 Uniform Guidelines on Employee Selection Procedures or “UGESP” under Title VII. However, there was some concern on the reliance on the 4/5th rule for determining whether an AI tool is creating a disparate impact. The 4/5th rule states that if the selection rate for a certain group is less than 80 percent of that of the group with the highest selection rate, there is adverse impact on that group. EEOC Chair, Charlotte Burrows, signaled that further clarification from the EEOC is necessary regarding total reliance on this rule for determining disparate impact as the rule is not intended as a legal definition, and it is one of many factors that may be used for determining disparate impact. But because UGESP provides some guidelines that may be useful to the context of AI, there was agreement that raising awareness about these guidelines and providing additional educational materials for employers would be useful.
Other Topics of Discussion
Other topics discussed included: how and from where AI data is sourced; high-level explanations of how AI tools learn; how AI is currently being, or can, be deployed in the workplace (i.e., recruiting and selection, performance management, assigning of tasks, determining compensation, termination/RIF decisions, etc.); the critical need for job analyses to be built into AI systems; and issues created with use of AI including underrepresentation, differential validity and proxy variables.
Conclusions
Based on the witness testimony and the EEOC’s line of questions for the witnesses, it is safe to predict that the EEOC is not actively seeking ways to curtail or prohibit the use of AI in employment decisions. In fact, the agency acknowledges that properly deployed AI holds some potential benefits for employers by creating efficiencies in processes such as recruiting and selection, assisting in reducing biases and cost effectiveness. The agency is simply working to issue standards and/or guidance on the use of AI in employment in order to mitigate civil rights violations that may be an unintended consequence of using these tools. Of note, the agency has already published a technical assistance document titled, The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees.
It is also reasonable to expect that any guidance that is released will clarify and/or expand upon how the use of AI correlates to the standards already in existence under UGESP.
The speed at which the EEOC will develop, and release standards or guidance is unknown, as it is still within the agency’s interest to ensure that they consider the perspectives of other key stakeholders such as AI vendors and employers who have deployed this technology in their workplaces. Development of meaningful guidance should also include consultation with other experts such as Industrial and Organizational Psychologists, AI Developers and possibly, other governmental agencies, who also have a joint interest in protecting civil rights, such as the Department of Justice (DOJ), the Office of Federal Contractor Compliance Programs (OFCCP), and the Federal Trade Commission (FTC). What can be said, is that is not a matter of if the EEOC will continue to move forward with its AI initiative – but at what speed!
HR Works will continue to monitor the agency’s progress on this topic and provide updates as needed.
