Unlock Editor Digest for free.
Rola Khalaf, editor of the FT, selects her favorite stories in this weekly newsletter.
AI alarmists warn that machine learning will destroy humanity — or at least make humans redundant. But what if the real problem was more mundane — that AI tools just do bad work?
That's what Halke Shelman, a reporter and professor at New York University, found after spending five years using investigative tools that are now widely used in hiring, firing and managing employers. Bots increasingly dictate which job ads we see online, which CVs recruiters read, which applicants make it to the final interview and which employees get promotions, bonuses — or redundancy notices. is received. But in a world where algorithms define who we are, where we excel, and where we struggle. . . What if the algorithms go wrong?” asks Shellman. Algorithman account of its results.
Recruiters and managers have many reasons to turn to AI: sifting through impossibly large piles of CVs and filling posts faster; To help them find talented people, even if they come from unusual backgrounds. making fair decisions by eliminating human bias; or tracking performance and identifying staffing issues.
But Shellman's experience suggests that many systems on the market can do more harm than good. For example, she tests video interviewing software that deems her a close match for a role, even when she parrots her original, understandable answers with the phrase “I love teamwork.” replaces or speaks entirely in German.
She talks to experts who have audited CV screening tools for potential bias — and found them liable to filter out candidates from certain postcodes, a recipe for racial discrimination. in favor of certain nationalities; Or seeing a male-dominated pursuit like baseball as a sign of success. High performers are then selected for redundancy or automatically put out of the running for jobs they were qualified for, purely because they performed poorly in seemingly unrelated online games. were demonstrated which were used to score the candidates.
After playing a few, Shellman suspects that fast-paced, pattern-matching games or personality tests will help recruiters identify those most likely to fail in a role. The games would also be difficult for anyone who was distracted by children, or had a disability that was not recognized by the software.
But many of the problems Shellman finds aren't intrinsic to the use of AI. Developers can't design good recruitment tests if recruiters don't understand why some employees perform better than others. If a system is designed primarily to fill a vacancy quickly, it will not select the best candidate.
Shellman found that, unless developers intervene, job platforms tend to advertise more to candidates (mostly men) who are most likely to respond to recruiters and apply for senior positions regardless of experience. They are aggressive. Problems also arise because managers blindly rely on tools that were only intended to inform human judgment—sometimes under the mistaken belief that it will protect them from legal challenge.
Machine learning can amplify existing biases in ways that are hard to detect, even when developers are vigilant. Algorithms identify patterns of people who have performed well or poorly in the past, without the ability to understand whether the characteristics they achieve are significant or not. And when algorithms go wrong — sometimes massively — it can be incredibly difficult for people to figure out why, troubleshoot or even find a human to talk to.
Perhaps the most useful part of Schellmann's book is an appendix that gives tips for job seekers (use bullet points and avoid ampersands in your CV to make it machine-readable) and who People's employers are looking at them (keep the emails happy). But he also has suggestions for regulators, to ensure AI tools are vetted before they hit the market.
At a minimum, lawmakers could mandate transparency on the data used to train AI models and technical reports on their effectiveness. Ideally, government agencies would themselves test instruments used in sensitive areas such as policy, credit rating or workplace monitoring.
In the absence of such reforms, Shellman's book is a cautionary tale for anyone who thought AI would eliminate human bias from employment — and an essential read for job hunters. .
Algorithms: How AI Can Hijack Your Career and Steal Your Future By Light Shellman Hurst £22/Hachette Books $30, 336 pages
Join our online book group on Facebook. FT Books Cafe And subscribe to our podcast. Life and art Wherever you listen