Recruiting AI systems under fire for excluding workers
AI recruiting tools focus on hiring efficiency rather than efficacy, according to an Accenture-Harvard Business School study. The U.N. believes AI is fostering human rights abuse.
AI-enabled recruiting systems have been raising red flags for years over the possibility that they are embedding bias in the software. But the concern is escalating. Last week, the United Nations High Commissioner for Human Rights called for a broad moratorium on AI use, which includes recruiting AI systems.
A new Harvard Business School and Accenture report also argued that AI recruiting and application tracking systems (ATSes) prioritize efficiency in hiring and exclude qualified candidates.
ATSes can weed out credible candidates, according to the Harvard Business School and Accenture study. This isn't a new criticism. Other studies have pointed out that these systems rely on algorithms built by mostly white, mostly male engineers and are trained on massive data sets that may not represent diverse populations.
The Harvard Business School-Accenture study goes a step further, and identifies the segments of the population most hurt by HR bias and inflexibly configured and flawed recruiting systems, such as veterans, immigrants, people with physical disabilities, and those who have been previously incarcerated, among others. The study, "Hidden Workers: Untapped Talent," estimates that as many as 27 million job candidates fall into the category of "hidden workers."
The United Nations not only expressed concerns about AI bias, but made a plea last week, asking member states to delay the deployment of AI systems such as those used in recruiting until the problem with these systems is recognized and resolved.
"The risk of discrimination linked to AI-driven decisions -- decisions that can change, define or damage human lives -- is all too real," the United Nations High Commissioner for Human Rights Michelle Bachelet said in a statement. "This is why there needs to be systematic assessment and monitoring of the effects of AI systems to identify and mitigate human rights abuses."
Recruiters are almost "exclusively focused on efficiency," said Joseph Fuller, a professor of management practice at Harvard Business School and an author of the Harvard-Accenture study. In the hiring process, efficiency means using automation to sort job applications and minimize hiring time.
Excluding qualified workers
The researchers surveyed 2,250 executives in the United States, the United Kingdom and Germany. A large majority, 88%, agreed that qualified, high-skills candidates "are vetted out of the process" because they do not match the exact job criteria.
Employers "end up hiring people that they say aren't fully qualified, and they bemoan the fact that, 'We can't find qualified candidates,'" Fuller said.
Recruiting AI automation "has caused firms to narrow the pool of applicants so severely as to exclude qualified workers," according to the report.
What I want to do is hire people who are going to be a success in this job and enthusiastic about it and excited to be part of our organization.
Joseph FullerProfessor, Harvard Business School
Fuller said that vendors and HR managers need to talk about the next generation of their recruiting AI technology and develop a new definition of efficiency. That means hiring people who will be productive and contribute to high retention rates as well as eliminating job requirements that exclude workers, he said.
"What I want to do is hire people who are going to be a success in this job and enthusiastic about it and excited to be part of our organization," Fuller said.
The U.S. Department of Labor this month reported a record-high 10.9 million job openings in July. But African American and Asian American workers are not having as much reemployment success as other groups, according to the Center for Economic and Policy Research (CEPR).
Some workers are falling behind
Last week, CEPR reported that in the first six months of this year, "18% of Black and 20% of Asian American unemployed workers gained a job, which is around five percentage points lower than the rate of reemployment for white and Hispanic unemployed workers."
Dean Baker, a senior economist at CEPR, could not attribute the disparity to AI systems used in hiring. "We're asking about a pattern that we first see in 2021, so the question is whether there can be that much more use of AI in screening job applicants this year than in 2020 or 2019," he said in an email.
The CEPR only speculated on the reasons for the disparity in hiring for Black and Asian American workers and said discrimination may be part of the problem.
But the idea that bias is excluding certain workers, including those identified in the Harvard and Accenture study, is happening, said Johnny Taylor, president and CEO of the Society for Human Resource Management, at its recent conference in Las Vegas.
Taylor told attendees that the HR profession is "famous" for weeding out "undesirable job candidates," a category that can include older workers, disabled people, and those without college degrees and with criminal records.
Recruiting AI algorithms are trained on historical data. If that data favors "male, white, middle-aged men, the resulting algorithm will disfavor women, people of color and younger or older people who would have been equally qualified to fill the vacancy," the UN report stated.
AI moratorium gets attention, not traction
While it's unlikely to gain much traction, the call for an AI moratorium is like calling for a nuclear weapons ban: "It gets everyone's attention," said Manoj Saxena, chair of the Responsible AI Institute and executive chair of CognitiveScale Inc. The firm measures the transparency and trust of AI systems. The nonprofit has a certification process similar to the silver, gold and platinum system used to certify buildings as environmentally friendly.
"We are in a world where we are automating inequality at scale," said Saxena, who argued that discrimination is rising.
Saxena called the Harvard Business School-Accenture report compelling because "you can design an AI to be a lot more effective, not just efficient." That might mean recommending a job candidate who falls outside the scope of the exact criteria, but would nonetheless be a good candidate, he said.
Relying solely on AI in employment is risky, Saxena said. He expects to see lawsuits, in time, from people who feel that they have been discriminated against by employers because of AI decision-making.
Josh Bersin, an industry analyst and head of Josh Bersin Academy, said the report is not raising a new topic, and firms have been diversifying hiring strategies for years. Some, for instance, are offering programs to get retirees to return to work, he said.
But the importance of the report "is the need for recruiting tools -- and HR departments -- to be much smarter about who is likely to succeed in this job," Bersin said. That means looking at a candidate's experience, education, job history, personal connections as well as mental, physical, emotional and other strengths. He said that an ATS doesn't do this, but some intelligent hiring systems are figuring it out.
Patrick Thibodeau covers HCM and ERP technologies for TechTarget. He's worked for more than two decades as an enterprise IT reporter.
