Editor's Note
Is Your Pre-employment Testing Biased?
The ACLU has filed claims with the FTC claiming that several types of candidate screening tools discriminate against people of color and disabled people. And they're probably right.
A few years ago, a startup claimed they could determine and assess someone's skills by looking at their social feeds and providing a report. So I put them to the test and asked for my report based on my public Twitter feed. The results said I was a terrible writer.
I write for a living—legal writing, newsletter writing, blog writing. I have taught writing for 20 years and just published a book on pay equity. Despite my bouts of imposter syndrome, I'm pretty confident that my skills involve writing.
So why would the program say I'm a terrible writer? Well, for the past few years, each morning I've posted a landscape photograph with the word, "breathe." It's my small way of dealing with the political, legal, and social chaos of the last 8 years or so. In one of my lives I'm a photographer so I have a lot of photos. And putting them in the world makes me feel better.
But to a computer, my twitter feed is full of posts with one word. Even worse, it's the same word. And that is just one example of why computers can't and should not judge humans.
This is an excellent article on the ACLU claim that screening tools discriminate. The tools involved are far more sophisticated than the one in my example, but they still can't do what they are designed to do without bias. So if you use personality tests, video interviews, or cognitive assessments, you will want to look at the tools and how you use them. You probably need to do a diversity assessment with your friendly employment lawyer too.
- Heather Bussing
The ACLU recently fired a clear warning shot to employers by asking the FTC to investigate a personality assessment test, a video interview tool, and a cognitive ability assessment screening device – all powered by artificial intelligence – because of alleged discrimination. The May 30 complaint follows the ACLU filing a charge with the EEOC over the same AI tools, alleging they unfairly screen out applicants with disabilities and unfairly target those with certain racial backgrounds. What do employers and AI developers need to know about these aggressive maneuvers and how should employers adjust when it comes to hiring?
3 Hiring Tools Targeted by ACLU
The three tools under attack were developed by Aon Consulting, a major player in the assessment field.
- Personality Assessment Test – The ADEPT15 is a scientifically based assessment tool that has been administered to applicants and workers over 8 million times. It uses algorithmic processes to assesses broad workstyles and determine how a worker performs tasks, adapts to changes, performs in a team, and more. The developer says using this tool, which can be completed in about 25 minutes, allows employers to uncover hidden traits and talents to find the best fits for their organization.
- Video Interviewing Platform – The vidAssess-AI tool enables employers to conduct a video interview asynchronously, relying on a set of pre-developed questions presented to each candidate. The program identifies positive and negative indicators in each response, but does not record or score visual cues. The developer notes that the platform eliminates bias and introduces objectivity since each candidate has the same experience, without small talk, unstructured questions, or unconscious bias tainting the process.
The ACLU claims that these tools assess general personality traits such as positivity, emotional awareness, liveliness, ambition, and drive that are not job-related or necessary for a specific job. It claims they can unfairly screen out people based on disabilities, as those with depression, anxiety, and autism (to name a few) can be scored low without regard to their actual job skills.
The ACLU specifically says that the use of AI to fuel these tools does nothing but exacerbate these fundamental problems. In fact, it alleges that the video interviewing tool is likely to discriminate based on disability, race, and other protected characteristics because of the AI data relied upon by the programs.
- Cognitive Ability Test – The developer also uses gridChallenge, a gamified cognitive assessment tool powered by AI that assesses working memory – essentially testing how well a worker can process information in the presence of distractions or under stress.
The ACLU cites data showing that the average gridChallenge scores for assessment-takers of different races reveals troubling disparities. It claims that a study revealed Asian, Black, Hispanic or Latino, and other test-takers scored lower than white test-takers on average.
It also claims that gridChallenge can discriminate against those with cognitive impairments and various mental health disabilities. For example, it contends that autistic people score significantly lower on various measures of working memory compared to individuals in the general population.
Why Did the ACLU File With the FTC?
The ACLU’s complaint with the FTC centers around the developer’s claim that these hiring tools are “fair,” “bias free,” and “without adverse impact” – which the ACLU alleges amounts to a deceptive marketing tactic. It claims there is a representation, omission, or practice that is likely to mislead consumers – in this case, employers looking for an efficient and effective hiring experience – to their detriment.
It asks the agency to launch an investigation into the developer to determine whether it is engaging in unfair and deceptive acts and practices. It also seeks an order from the agency to force Aon to stop selling or administering these tools until these problems are fixed.
Employers Aren’t Necessarily Off the Hook
While the ACLU’s complaint with the FTC doesn’t name any employers who might have relied upon these tools, the ACLU is not necessarily letting them off the hook quite yet. While it claims that the developer may have misled employers into a sense of comfort through its allegedly deceptive statements about the tools, it also says that “employers are legally obligated to ensure that the assessments they use do not discriminate, including when using technologically complex assessments.” Plaintiffs’ lawyers will certainly watch the FTC investigation closely.
EEOC Process Also Underway
In fact, the ACLU has already initiated an employment-related attack against Aon and one of the employers using these tools. It claims that a biracial autistic job applicant who was required to take the ADEPT-15 and gridChallenge assessments as part of the employer’s hiring process faced racial and disability discrimination as a result. It alleges violations of the Americans with Disabilities Act (ADA) and Title VII of the Civil Rights Act of 1964 on behalf of not only that applicant but all others similarly situated.
What’s Next?
The EEOC charge was filed late last year and is still pending with the agency. That administrative process could take several more months before any action is revealed, but it is likely that the charge was simply a precursor to the ACLU filing federal litigation against Aon and the employer on these same claims.
The FTC complaint is very recent, having just been filed in the past few weeks. The agency will soon determine what steps to take as a result of the allegations presented in the complaint and the developer’s response, though the administrative process could take months to unfold.
What Should Employers Do?
If you currently engage an AI developer or seek to use an AI-fueled tool to improve your hiring process, you should consider the following steps in light of these recent actions:
- Establish an AI Governance System –Develop clear systems and procedures – including a human oversight component – before deploying AI in the workplace. This is consistent with the guidelines issued by the National Institute for Standards and Technology (NIST) on artificial intelligence (NIST AI RMF). You should also evaluate the processes once in action to make sure they effectively govern the AI systems.
- Vet Your Vendors – As the ACLU noted in its complaint, employers cannot simply play dumb and point fingers at AI vendors should problems arise. The EEOC said as much when it issued guidance about AI bias last year. Plaintiffs will increasingly seek to hold employers accountable for any resulting discrimination, so make sure you ask the right questions (with the help of your lawyers) when you are vetting products and at the time of implementation.
- Be Transparent – Employers should be transparent with workers and candidates about AI systems being used in the workplace. This is an increasingly common theme, and one that may soon be required in states like Colorado, California, and New York.
- Offer Accommodations if Feasible – Some states could soon require employers to allow workers and applicants to opt out of any AI process that makes determinations about their workplace future. You might want to offer the same. Some options could include specialized equipment, alternative tests or exam formats, or the opportunity to work in a quieter setting.
- Align Your Questioning – To the extent you develop a set of questions to be fed into the AI system and posed to applicants, make sure they are closely aligned with the job requirements for the position and don’t range too far afield.
