There is probably no better way to reduce hiring bias than to kill the job interview

Interviews are popular, both with employers and candidates. Employers love the ability to attend to everything they should (at least in theory) ignore: gender, race, class, attractiveness, etc. This is why people are suspicious, or even annoyed, when they can’t see candidates on Zoom. As for candidates, they mostly enjoy the interview because they overrate their own likability, and their ability to charm or influence interviewers. And yet, most people end up failing their job interviews, and for reasons that have little to do with their actual potential, or real fit with a given job or role.

It is for these reasons that interviews are unlikely to disappear anytime soon. Meanwhile, we continue to reject the possibility that if there is indeed a connection between what people actually do on an interview, and their subsequent job performance, artificial intelligence should be able to detect this better than irrational, prejudiced, political, or self-interested humans (which basically just means humans). Decades of scientific research show that the only valid or predictive job interviews are highly structured, standardized, and able to link behavioral signals to concrete indicators of future work performance. Yet when we rely on recruiters or hiring managers to pick up on these signals, our chances of success are severely limited.

Sure, most people think they “know talent when they see it”, especially when they are hiring managers or recruiters. However, few people are as good at spotting talent as they say, and this is even more true when it comes to inferring potential (talent in the making, or nascent talent). Moreover, it is not easy to reverse engineer the logic or instincts of those who do happen to be good at talent-spotting - a bit like Phillip Tetlock’s Superforecasters. They rely on years of clinical experience, high EQ, and an unusual ability to control their biases or at least be aware of them. And even they can have an off day, be hangover, or let their personal preferences interfere with their objective judgments of talent.

It is also true that you cannot just eliminate people’s biases. Humans are biased by nature and the same shortcuts that make us prejudiced or narrow-minded are essential for sorting out the information overload, and sensory ambivalence, ambiguity, and confusion we encounter in everyday life. Our brains are not for thinking, they are for making fast (hit-and-run-like) predictions, and when we try to slow people down, by making them more deliberately aware of their potential biases, they will malfunction. Imagine an interviewer who undergoes extensive unconscious bias training to discover that they are negatively predisposed towards women, hispanics, or short people: next time they interview someone who fits into any of these categories, they will not be able to think of anything else but their biases, and either overcompensate or get distracted...

The best case scenario is to make our intuition data-driven, by devoting years or decades to examining people, interpreting signals, and learning whether someone displays certain attributes, and whether that makes them a better or worse fit for a certain job. But if this can be done, and such connections truly exist, then we can safely assume that machines will be better able to identify them than humans, and that this would be a far more efficient, scalable, and reliable way to evaluate candidates for jobs. In turn, this would free up humans to perform the humane part of a job interview: answering questions, showing empathy, establishing rapport, and showcasing the cultural DNA of the company. Although ideally, assessments and AI would have also been used to provide candidates with a clear picture of their interviewers’ personality, values, and skills - as well as their organization’s culture.

Of course, the only way to test the unlikely possibility that eliminating interviews may actually reduce bias and increase predictive accuracy around talent and job fit, would be to objectively evaluate job performance in the first place. But when the people responsible for rating employees’ job performance, and evaluating their contribution to the firm, are the very people who interview candidates in the first place, we can expect the same level of objectivity of someone who is in charge of judging a cake they have baked, or evaluating their own skills on any task. More often than not, managers will persuade themselves that they have hired the right candidate even when they have not - they don’t even need to lie to others, they just need to fool themselves - because the same biases that led them to hire the person led them to evaluate their performance positively. And if they are aware of their mistakes, there is always an opportunity to save face, and lie to others.