EEOC Issues Guidance Regarding How Employer Software and Artificial Intelligence May Discriminate Against Individuals With Disabilities | Sheppard Mullin Richter & Hampton LLP

Evoz Tech

[co-author: Wolfram Ott]*

On May 12, 2022, the Equal Employment Opportunity Commission (“EEOC”) issued guidance addressing the application of the Americans with Disabilities Act (“ADA”) to employers utilizing software, algorithms, and artificial intelligence in hiring and employment decisions.  Produced in connection with the EEOC’s launch of its Initiative on Artificial Intelligence and Algorithmic Fairness in October 2021, the EEOC’s latest guidance reflects its goal of ensuring that employers utilizing technology in hiring and employment decisions are complying with federal civil rights laws.  Notably, the guidance was issued a few days after the EEOC filed a complaint against a software company alleging age discrimination, potentially signaling similar actions related to the use of artificial intelligence in the employment context.  Below are some key takeaways on the new guidance.

Scope and Definitions

The guidance implicates a broad range of technologies commonly utilized by employers including software, algorithms, and artificial intelligence:

  • Software: Refers to information technology programs or procedures that provide instructions to a computer on how to perform a particular task or functions.  Examples of software used in employment and hiring decisions include automatic resume-screening software, hiring software and video interviewing software.
  • Algorithms: A set of instructions that can be followed by a computer to accomplish an end.  Human resources software and applications use algorithms to allow employers to process data to rank, evaluate, rate, and make other decisions about job applicants and employees.
  • Artificial Intelligence (“AI”): Congress has defined “AI” to mean a “machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.”  Utilizing AI in the employment context has typically meant that the developer relies partly on the computer’s analysis of data to determine which criteria to use when making employment decisions.  AI may include machine learning, computer vision, natural language processing and understanding, intelligent decision support systems and autonomous systems.

Employers may use tools that include a combination of these terms.  For example, an employer may utilize resume screening software that incorporates an algorithm created by human design or an algorithm that is supplemented by AI analysis of data.

Ways in Which Algorithmic Decision-Making Tools may Violate the ADA

The guidance discusses the three most common ways that an employer’s use of algorithmic decision-making tools could violate the ADA.  This includes the following:

  • When an employer fails to provide disabled job applicants and employees with “reasonable accommodations” that are needed for the assessment tool to be fair and accurate to the applicant or employee.  The guidance makes clear that when an employer utilizes software tools, AI, or algorithmic tools to assess applicants or employees, the ADA requires reasonable accommodations are made for individuals if their disability will make it difficult to take such an assessment, or cause a less favorable result.  For example, the EEOC states that a job applicant with limited manual dexterity may report that they would have a difficult time taking a knowledge test that requires a manual input device such as a keyboard or trackpad.  In this case, the employer would need to provide an accessible version of the test (for example, one in which the job applicant is able to provide responses orally, rather than manually) as a reasonable accommodation, unless doing so would cause an undue hardship.
  • When the technology “screens out” disabled individuals, whether intentional or not.  “Screen Out” is unlawful when an individual who would be otherwise able to perform the essential functions of a job loses that job because they are unable to complete an assessment, or their performance on that assessment suffers, as a result of their disability.  This may happen even when an assessment professes to be “bias-free.”  For example, if a chatbot is programmed with an algorithm that rejects all applicants who, during the course of their “conversation” with the chatbot, indicate that they have significant gaps in their employment history.  If a particular applicant had a gap in employment, and if the gap had been caused by a disability (for example, if the individual needed to stop working to undergo treatment), then the chatbot may function to screen out that person because of the disability.
  • When the assessment contains “disability-related inquiries” or functions as an impermissible “medical examination.”  Any questions that may prompt an employee or applicant to provide information about a disability or if the individual has a disability, whether directly or indirectly, is a “disability-related inquiry.”  Questions seeking information about an individual’s physical and mental impairments, or health may also qualify as a “medical examination.” 

Employer Responsibility for Vendor Technology

Importantly, the EEOC guidance states that employers are generally responsible for the discriminatory effects of software utilized in the hiring process even when the software is utilized by third-party on behalf of the employer.

Best Practices for Employers

The EEOC offered so-called “Promising Practices” for employers seeking to ensure compliance with the ADA.  These recommendations provide helpful suggestions about ways in which employers may protect themselves against claims of disability discrimination.  Those recommendations include to:

  • Inform applicants or employees about how an assessment will be conducted, let them know accommodations are available if needed and explain the procedure for requesting such accommodations;
  • Explain what traits an assessment is designed to measure and how that measurement will be conducted;
  • Develop other methods of evaluating employees and candidates when the standard tools may disadvantage individuals with disabilities;
  • Ask all vendors and third-parties acting on behalf of the company to conduct assessments to forward all requests for accommodation, or require the third party to provide reasonable accommodations as required by the ADA;
  • Ensure technologies used to assess employees and applicants were designed to be accessible to individuals with a wide range of disabilities;
  • Make sure assessments only measure traits, skills, abilities, or qualities necessary for the job, and measure those qualities directly; and
  • Confirm that software and apps, algorithms, artificial intelligence, and other assessment systems do not ask questions about disability, physical or mental impairments, or health unless those questions are related to requests for reasonable accommodation.

Key Takeaways

May marked the first new developments out of the EEOC relating to AI since the launch of the Initiative on Artificial Intelligence and Algorithmic Fairness.  This new guidance provides much needed insight on how the EEOC will enforce the ADA with respect to AI going forward.  Employers should utilize the provided “Promising Practices” to ensure compliance, and avoid possible liability.  

Because this issue is still developing, we will continue monitoring developments in this area and provide updates as new information becomes available.

*Wolfram Ott is a summer associate in the Labor and Employment group and assisted with the drafting of this article.

https://www.jdsupra.com/legalnews/eeoc-issues-guidance-regarding-how-7940655/

Next Post

Faster computing results without fear of errors | MIT News

Researchers have pioneered a technique that can dramatically accelerate certain types of computer programs automatically, while ensuring program results remain accurate. Their system boosts the speeds of programs that run in the Unix shell, a ubiquitous programming environment created 50 years ago that is still widely used today. Their method […]