4 Artificial Intelligence-Based Hiring Practices That Can Be Discriminatory Towards People With Disabilities

When the COVID-19 pandemic hit, so did a shift in most companies’ perspectives on technology.

As Accenture CEO, Julie Sweet, recently told Fortune Magazine, the COVID-19 pandemic has caused many companies to reevaluate their views on technology. “We are no longer spending time talking about, is technology good or bad and what are the risks”, she notes. With over 500,000 employees and $40 billion in revenue, Sweet understands that during the pandemic, “tech became the lifeline for individuals, societies, businesses and government.” Because of this increasing reliance on technology, employers may be tempted to turn to artificial intelligence as part of their hiring processes. However, a growing body of research demonstrates a record of inaccurate results, as well as inherent disadvantages for women and people of color, and people with disabilities. Decision-centric data are based on profiles of current employees – a pool already sorely lacking those of us with disabilities.

Although artificial intelligence is often marketed as an objective tool, it can be yet another barrier to securing employment for people with disabilities. Here are four artificial intelligence-based hiring practices that can be discriminatory towards candidates with disabilities:

1. Applicant tracking systems (ATS). Applicant tracking systems are a computer software that uses technology to process job applications that often use CAPTCHAs (Completely Automated Public Turing test to tell Computers and Humans Apart). CAPTCHAs can be difficult, if not impossible, for people with vision and hearing impairments, learning disabilities, or dexterity limitations. I utilize four protheses and this practice is not so user-friendly for those of us wearing clamps. Requiring potential talent to prove their humanity with CAPTCHAs can lead to smaller talent pools.
2. Resume Screeners. Resume screeners, software that scans resumes for keywords relevant to the employer’s desired characteristics in an applicant, can be a significant hurdle for people with disabilities. These systems may not have been trained on data or writing styles of users with diverse cognitive and intellectual abilities.
3. Facial Recognition. Facial analysis software assessment is, in my opinion, highly risky. WHY? Here are only a few examples of applicants who might easily be screened out by AI: people with facial differences based on ethnicity, disfigurement or paralysis; people with blindness who may not be facing a camera or be able to make eye contact; people with speech delays; people with hearing loss who need captions to interpret the questions. Let us not forget that IBM, through its new CEO Arvind Krishna, is ceasing facial recognition tech development for these discriminatory and other values-based reasons.
4. Conversational Agents. Also known as chatbots, employers use conversational agents anywhere in the hiring process, from asking screening questions to scheduling an interview with a recruiter. These agents may not be able to correctly support non-writing communication methods, such as text-to-speech and VoiceOver, or interpret words and phrases it has not previously heard.

Assessments of the fairness of artificial intelligence in hiring has been largely overlooked in research. Artificial intelligence is a reflection of who builds it. If those who are building these types of software lack experience working with us, that bias is likely to seep into the way it performs and makes decisions. Although likely not intended, artificial intelligence can be biased and discriminatory. Recruitment artificial intelligence systems must be fair for all before employers decide to purchase them.

NBDC is an employer resource that addresses contemporary issues, such as AI, and how they affect the development and maintenance of a diverse workforce – one that includes a disability culture. Read more about NBDC’s resources and the benefits of membership.