Can AI really debias recruitment?  

pixabay geralt

Are AI video hiring tools the cutting edge of recruitment and create more diversity? A deeper look at HireVue, myinterview and Retorio and a personality machine to try it out.

If you’re applying for a job today, you might be asked to attend an interview with an AI. There are now many companies that use AI to analyse videos of candidates before putting you through (or not) to the next round of the recruitment process. These are called AI video hiring tools.

These tools claim to assess a candidate’s personality to tell if they are the right fit for the job. They say that they can know your personality from looking at your face. The idea is that, like a lie-detector test, AI can see ‘through’ your face to the real you. The real you, they say, can be given a Big 5 personality trait score for: extroversion, agreeableness, openness, conscientiousness, and neuroticism. They claim that by looking at specific regions of your face AI can assess your Big 5 scores. From there, they can invite you through to the next round of the recruitment process or reject you.

These tools have become incredibly popular. By mid-April 2020, when we were a few months into the COVID-19 pandemic, a poll of 334 HR leaders found that 86% of organisations were incorporating new virtual technology into their hiring practices. It’s now October 2022, but these technologies are markedly on the rise: a 2020 study of HR professionals from 500 mid-sized organisations in a range of industries in five different countries found that 24% of businesses have already implemented AI for recruitment purposes and 56% of hiring managers plan to adopt it in the next year.

COVID-19 isn’t the only factor. Burdened with the immense challenge of volume recruitment, hiring managers are exploring ways to free up more of their time to engage with selected applicants in person. Furthermore, AI HR tools now claim to not only streamline recruitment but help corporations fulfil corporate diversity, equality and inclusion (or DEI) goals. For example, HR AI tool Retorio says that “85% of talent acquisition leaders feel the pressure to increase diversity” while HireVue emphasises how their tools helped the Co-Operative Bank “bring their values of inclusivity to life”. A lot of companies are deeply concerned about the lack of diversity in their current workforces and are struggling to meaningfully grapple with this problem. Hence, it’s unsurprising that many of these companies’ HR teams are excited by the prospect of technologies that claim to be able to combat human bias in the hiring process.

Creating a more diverse workforce?

In our study, we analysed the marketing materials offered by AI hiring firms – Hirevue, myInterview, and Retorio – and identified four key claims made by these AI hiring companies. These were: (1) their tools remove bias from the hiring process, providing a more objective and neutral assessment of candidates; (2) AI-powered hiring will promote the recruitment of a more diverse workforce; (3) AI recruitment tools can identify a candidate’s internal characteristics from their face; and (4) this process of categorization and sorting allows AI recruitment tools to identify a company’s ‘ideal candidate’. However, using feminist and critical race theory, we challenge these four claims, arguing instead that AI tools can’t simply remove race and gender from the hiring process (and in doing so, neutralise gender- and race-based discrimination that arises during recruitment).

Our research shows that these tools can’t be trained to only identify job-related characteristics and strip out gender and race from the hiring process because the kinds of attributes we think are essential for being a good employee are inherently bound up with gender and race (which also intersect with class, disability, age, and other ways that we are socially classified). Our gender and racial identities are not only central to who we are, they’re also part of wider systems of power that make some people seem more hireable than others. For example, these tools claim to be able to identify people who are a good ‘culture fit’ with a company; Retorio claims that its tools “can determine person-culture fit and estimate whether a person is more or less likely to thrive within an organisational culture”.​​ However, the idea of ‘culture fit’ has long been used as a euphemism to justify exclusive work cultures and make some employees feel like they ‘belong’ while others do not. Indeed, we argue that AI-power…

Dr Kerry Mackereth, MPhil and doctorate degree in Multi-Disciplinary Gender Studies at the University of Cambridge Centre for Gender Studies. She is a Research Associate at the University of Cambridge Leverhulme Centre for the Future of Intelligence. In February 2023 she will be a Visiting Fellow at the Sarah Parker Remond Centre for the Study of Race and Racialisation at the Institute of Advanced Studies, UCL.

Eleanor Drage, International Dual PhD from the University of Bologna and University of Granada. She is a researcher at the University of Cambridge Leverhulme Centre for the Future of Intelligence, specialising in feminist ideas to make artificial intelligence safer for everyone.

Diskutieren Sie mit