Recruiter Trains AI On Hired Candidates To Eliminate Bias

A recruiter at Symmetry Staffing Solutions in Austin said Tuesday that she had built a machine learning system to reduce hiring discrimination by training it exclusively on candidates the company had already hired. The algorithm, called Precedent, analyzed seventy-three thousand résumés of people Symmetry had employed over the past decade and learned to identify the traits that correlated with employment at the firm. The company plans to use Precedent to screen new applicants and flag candidates whose profiles most closely resemble those of previous hires.

Symmetry's leadership framed the tool as a breakthrough in objective recruitment. We wanted the machine to learn from our best people, and our best people are the ones we hired, said Karen Ng, Symmetry's head of talent acquisition, in a statement to HR Wire. The system removes human bias from the equation by letting data speak for itself.

We wanted the machine to learn from our best people, and our best people are the ones we hired.

The logic of the approach generated enthusiasm among Symmetry's board, which approved a budget expansion last month. Because the algorithm relies on measurable factors rather than subjective judgment, it cannot discriminate on protected characteristics, argued Marcus Fewell, Symmetry's chief technology officer, in an internal memo obtained by this publication. He noted that the system scored applicants on education, work history, and language use, all of which were objectively quantifiable.

What Precedent could not do was isolate which traits it had actually learned. A statistical analysis of the algorithm's decisions, conducted by an outside consultant in late August, found that candidates matching the profile of Symmetry's current workforce were forty-one percent more likely to be flagged as strong matches. The consultant's report, marked confidential, did not specify whether this outcome constituted a problem or a feature of the system.

Symmetry has since integrated Precedent into its first-pass screening process for entry-level positions. David Cheng, a senior recruiter who beta-tested the system, said he was pleased by the consistency it brought to his work. The algorithm never gets tired or distracted, he said in an interview. It applies the same standard to every person who applies. When asked whether the standard itself might be flawed, he noted that Precedent had been trained on Symmetry's most successful employees and therefore could not be wrong about what success looked like.

At press time, Symmetry announced that it was expanding Precedent to management-level hiring and had begun licensing the underlying framework to fourteen other staffing firms.