Companies Like Microsoft and Amazon Are Developing AI-Assisted Hiring Policies
-
Although further incorporating AI into hiring practices may solve certain problems, experts emphasize that technology should not be expected to entirely overhaul how companies recruit new talent.
According to Criteria's 2023 Hiring Benchmark Report, only 12% of hiring professionals currently report using artificial intelligence in their recruitment or talent management processes. However, AI solutions targeting everything from streamlining talent acquisition to making informed selection decisions are being "very aggressively marketed," said Josh Millet, founder and CEO of Criteria.
AI is playing a role in various aspects of hiring, such as drafting job descriptions, screening candidates, creating and scoring assessments, filtering new applicants, communicating with candidates, and onboarding new hires. Criteria's report highlights that tools like OpenAI's ChatGPT, Google's Bard, recruitment chatbots, and proprietary solutions are facilitating these processes.
As organizations await the potential expansion of laws like New York City's AI Bias Law to other regions, the Center for Industry Self-Regulation (CISR) – a 501(c)(3) nonprofit foundation under BBB National Programs – has released principles and protocols for trustworthy AI in hiring.
These principles, developed in collaboration with companies like Amazon, Unilever, Koch Industries, and Microsoft, address transparency, fairness, non-discrimination, technical robustness, safety, governance, and accountability in the use of AI for hiring. The agreement also establishes certification standards for third-party AI vendors to promote accountability beyond employers.
How AI is Applied in the Job Market
Currently, AI is used in recruitment processes for tasks such as creating job descriptions, sourcing talent, designing and scoring assessments, screening applicants, communicating with candidates, and training new hires. Eric Reicin, CEO of BBB National Programs, stated that when AI tools are properly designed, deployed, and monitored, they have the potential to mitigate discrimination and bias on a broader scale.
Reicin emphasized that the primary goal of defining and publishing trustworthy AI hiring principles and protocols is to ensure effective and reliable systems that promote transparency and accountability, while striving to build secure, resilient, and scalable solutions.
The objective is to realize the promise of AI in hiring while managing inevitable risks. According to Reicin, AI enables organizations to fairly expand candidate pools: one organization in CISR's incubator reportedly received 20 million applications last year. "No human could fairly process all those applicants," he noted.
However, Reicin pointed out that issues of explicit and implicit bias persist in hiring, regardless of whether technology assists vendors, despite ongoing efforts to eradicate it.
The latest candidate experience report on hiring status shows that 39% of candidates were ignored in the past year. According to a 2021 study by the National Bureau of Economic Research, individuals with distinctively Black names had a 2.1% lower probability of employer contact compared to those with White-sounding names. Recently, iTutorGroup was forced to pay a $365,000 settlement for training its AI to screen out applicants over 60 years old, violating the U.S. Equal Employment Opportunity Commission's age discrimination laws.
Millet stated: "We can use evidence-based methods to help remove some biases from the selection or talent evaluation process. That's the promise."
However, Millet added that AI's promise doesn't always deliver. "In trying to eliminate bias, we sometimes create the opposite effect and amplify prejudice," he said.
Nearly a decade ago, Amazon attempted to automate hiring but abandoned the project when the tool "could not rate candidates for software developer and other technical roles in a gender-neutral way," according to a 2018 Reuters report.
Experts say AI has tremendous potential, but whether it develops along a positive trajectory depends on how it's used. Trustworthy AI hiring principles and protocols represent just one effort to mitigate risks and maximize its promise.
Maintaining human involvement in the hiring process
Millet stated that certain uses of AI in the recruitment process are relatively uncontroversial, such as using intelligent interview scheduling and chatbots to help candidates navigate the hiring process more smoothly. However, high-risk decisions or data-intensive use cases present different challenges.
Reicin recommended conducting regular vendor analyses, which involve asking questions such as: What debiasing measures is the vendor implementing? Who certifies the vendor, and what does this certification entail? How do they collect, protect, and use data?
Employers must recognize the importance of human oversight when implementing AI, Reicin emphasized. He illustrated the potential impact of video interviews, where AI technology can collect data on candidates' voices, tones, and eye movements. What if candidates are unaware this is happening?
Millet noted that employers and vendors should ensure data is used for reasonable and intended purposes. He said that communication and obtaining consent for appropriate use are two best practices employers should not overlook when using AI in recruitment.
Millet also highlighted that companies should familiarize themselves with 'glass box algorithms,' which are transparent in how they reach conclusions and can address issues related to 'black box algorithms.' The latter are opaque and may undermine trust or fail to establish it altogether.