THE HX BRIEF

Healthcare intelligence through the human experience lens
Edition: November 1, 2026 | Special Edition: AI, Hiring & Human Potential
Published: November 1, 2026

AI Is Screening Candidates. Who Are We Missing?

A rejection tells someone they did not move forward. It rarely tells them how that decision happened—or whether anyone had the opportunity to understand what they could contribute.


Stanford-led research raises an uncomfortable question: when employers depend on shared screening technology, how independent are those opportunities?

THE SIGNAL

In Algorithmic Monocultures in Hiring, researchers from Stanford, Chapman, and Northeastern examined approximately four million applications from 3.4 million applicants across 156 employers and 11 industry sectors. All were assessed through algorithms from one vendor. The research was released in May 2026; this edition revisits its implications for hiring and human experience. Source: Research team’s study overview.


The researchers investigated whether employers’ reliance on shared algorithms could repeatedly exclude the same people. Three findings deserve attention:


An overall score can conceal disparities. Combining recommendations across jobs obscured adverse impact that became visible when researchers examined individual positions.


The disparities affected substantial numbers of applications. Of all applications submitted by Black and Asian applicants, 25.87% and 14.74%, respectively, were submitted to positions that adversely impacted Black and Asian applicants, as measured under U.S. employment discrimination standards. Source: Stanford Digital Economy Lab study abstract.


Repeated applications did not necessarily mean independent chances. Among applicants submitting four applications, about 10% received no positive screening recommendation across those applications—more than expected under the researchers’ benchmark of independent decisions. Source: Findings and methodology.


An essential distinction: The study examined one vendor’s game-based assessments and screening recommendations. It does not establish that every AI hiring tool behaves this way, and a screening recommendation is different from a final hiring decision. Sources: Stanford’s interview with the researchers; Stanford HAI research explanation.

THE IMPLICATION

Analysis: We need to look beyond whether technology processes applications quickly. We also need to understand what it measures, how that relates to the actual work, and who gets excluded along the way.


An organization can have a person making the final hiring decision while technology determines whose application that person ever sees. Meaningful human involvement therefore needs to begin before the shortlist is settled.


For healthcare organizations, this belongs in the workforce and experience conversation together. How we identify talent helps shape the teams patients, families, and colleagues will depend on.


A useful starting point is to bring recruiters, hiring managers, frontline colleagues, and technology partners together around three questions:


  • What evidence shows that this assessment measures capability relevant to this particular job?
  • What do results look like for individual positions and applicant groups?
  • How can we identify and reconsider people the screening process may have overlooked?

THE HUMAN IMPACT

Analysis: Behind an application may be someone trying to support a family, return to work, change careers, or bring years of transferable experience into a new setting.


Repeated rejection can make a person question their own value. Employers can also lose the opportunity to discover someone whose judgment, adaptability, or ability to connect with others becomes apparent through conversation or a practical demonstration.


Candidates deserve a process they can understand, a way to raise concerns, and a meaningful opportunity to demonstrate what they can do.

THE QUESTION

If technology decides who reaches a human, how will we know which capable people we never had the chance to meet?

NICOLE’S TAKE

You all know I use AI. I see its value every day, and I understand why recruiting teams need help managing the volume of applications coming through.


What gives me pause is the possibility that we are becoming more efficient at overlooking people.


I have worked with talented people whose career paths would not fit neatly into a screening model. Sometimes you understand what someone brings when you ask another question, hear how they solved a problem, or give them an opportunity to show you.


That takes time. I know it does. Recruiting teams are stretched, and hiring managers need support too. We can acknowledge those pressures and still ask whether the process gives people a fair opportunity.


For me, AI should help us prepare, organize information, and ask better questions. I want people actively considering candidates before an automated recommendation closes the door.


And we need to stay curious about our own decisions. Human judgment needs examination too. Let’s look together at who advances, who gets screened out, and whether our assessments reflect the work we actually need someone to do.


A faster process is valuable. So is discovering the person we might otherwise have missed.


Let’s make room for both.

SOURCES

Research team’s study overview — Algorithmic Monocultures in Hiring


Stanford Digital Economy Lab — Study abstract, May 26, 2026


Stanford Digital Economy Lab — Interview with the researchers, May 28, 2026


Stanford HAI — Research explanation, May 26, 2026