The Numbers Nobody Was Required to Count
Even now, somewhere in the United Kingdom, an applicant tracking system is deciding who gets an interview before a human being has read a single line, dear reader — and for once, someone actually checked the math. People Like Us, a workplace-equity nonprofit, commissioned Censuswide to survey 500 UK hiring decision-makers and 2,000 jobseekers, and the results were, in the driest possible sense, illuminating: 82% of employers who bother to monitor their AI screening tools by ethnicity found candidates progressing at meaningfully different rates. Thirty-six percent called the gap "meaningful." Twenty-eight percent of the rest don't monitor outcomes by ethnicity at all, which is one way to guarantee a clean conscience.
One presumes that not looking is, for many HR departments, a feature rather than an oversight.
The Great CV Un-Naming
And yet — as if this were not enough — the jobseekers noticed first, and adapted the way people always adapt to a system that won't explain itself: they started editing themselves out of it. Seventy percent of ethnically diverse applicants report altering their CVs specifically to survive automated screening — scrubbing country of origin, trimming religious and cultural markers, in some cases changing names outright — compared with 57% of white candidates doing the same for unrelated reasons. Minority candidates were also rejected within the hour at meaningfully higher rates than white candidates, which is an impressively fast way to never find out why.
Nothing says "equal opportunity employer" quite like a workforce quietly editing its own identity to satisfy a piece of software nobody will describe out loud.
"Inherently Objective," and Other Load-Bearing Myths
Seventy-six percent of the hiring managers surveyed believe automated screening tools are inherently objective, because — sources within the Human Resources Community confirm — they make decisions "based on data," which is a sentence that has never once in the history of software meant what it sounds like it means. Dr. Claire Tyler of UCL, who ran a companion analysis of over 100,000 live graduate applications, was rather less sanguine about the data in question. Employment lawyer Tom Heys of Lewis Silkin noted that most employers deploying these tools couldn't produce an audit of their outcomes if asked — which, one imagines, is precisely why nobody has been asking.
The Algorithm remains, as ever, unavailable for comment, and also for an actual bias audit.
Westminster Notices, Slowly
People Like Us has now put its findings on nineteen JCDecaux screens across London and is pushing lawmakers to fold mandatory automated-screening disclosure and independent bias audits into the UK's Equality (Race and Disability) Bill. In the meantime, the group built a tool letting rejected applicants formally ask whether a machine made the call, and demand a human redo it — a right that, in a development that will surprise no one who has been paying attention, currently exists mostly on a poster.
The gateway, it appears, has been automated. The receipt for that automation, at least, now has a name on it.
Sources: Fair Play Talks — Reject the Rejections campaign launched as study warns AI is filtering ethnic minority candidates out before interview · Campaign Brief — Reject the Rejections campaign for People Like Us
