A new study published in Organization Science reveals that algorithmic fairness isn’t just a technical design problem. Even when matching systems are built to be bias-free, unequal outcomes emerge when users enter the process with different levels of understanding and confidence.
When institutions adopt computerized matching systems, they often assume the hard work is in building a fair algorithm. A new study by researchers at the University of California Los Angeles suggests that assumption may be leaving a critical source of inequality unaddressed.
The study, published in Organization Science, a journal of INFORMS, finds that disparities can emerge even from systems specifically designed to reduce bias, discourage strategic manipulation and reward honest decision-making. The culprit isn’t the algorithm — it’s the unequal understanding users bring to it.
The research focuses on one of the highest-stakes algorithmic systems in professional life: the National Residency Matching Program, which determines where graduating medical students will complete their physician training.
How the Residency Match Is Supposed to Work
The residency match pairs medical school graduates with hospital training programs using a computerized algorithm. Both applicants and programs submit ranked preference lists, and the system is mathematically designed so that students benefit most from ranking programs in their genuine order of preference — not by trying to anticipate or game the process.
In theory, this design levels the playing field. Everyone is playing by the same rules, and honesty is literally the optimal strategy. But the study found that many students were still making suboptimal ranking decisions — often because they didn’t fully understand why honest ranking works best.
Some students ranked less-preferred programs higher, believing it would increase their chances of matching somewhere. In fact, that strategy can actually reduce their odds of landing their best possible placement.
“Algorithms do not operate in a vacuum,” lead author Samuel E. Skowronek, a postdoctoral scholar in behavioral science at the UCLA Anderson School of Management, said in a news release. “Even when the algorithm cannot be gamed, outcomes still depend on whether people have the knowledge and support needed to use it correctly.”
What the Data Revealed
To investigate, the researchers drew on two data sources: an incentivized simulation of the residency match involving more than 1,700 medical students, and 66 in-depth interviews with students navigating the actual match process. Together, these methods revealed a consistent and troubling pattern.
Male students were more likely than female students to independently seek out additional information about how the algorithm worked. They were more likely to consult multiple sources, revisit training materials, watch explanatory videos, and pursue independent guidance beyond what their programs provided.
That behavior gap had measurable consequences. Students who took those extra steps developed stronger, more accurate mental models of the system — and were more likely to use it in ways that maximized their outcomes. Students who relied primarily on standard institutional advice were more likely to misunderstand the process and submit rankings that weakened their results.
Women in the study reported lower confidence and less algorithmic understanding on average, and were more likely to deviate from the optimal ranking strategy. Critically, the researchers found this disparity did not stem from the algorithm itself treating men and women differently. It emerged entirely from differences in behavior, information-seeking and confidence surrounding the system.
“This research broadens the conversation around algorithmic fairness,” Skowronek added. “Fairness cannot be viewed only as a technical property of the algorithm. It also depends on how people engage with the system and understand how it works.”
The Problem With ‘Follow Your Heart’ Advice
The researchers also scrutinized the quality of guidance institutions provide to applicants. Many students described receiving advice that amounted to little more than being told to “rank programs based on your true preferences” or “follow your heart.”
While technically accurate, that guidance proved insufficient. Without understanding the underlying logic of why honest ranking is the dominant strategy, many applicants still acted on fear, uncertainty or incorrect intuitions — particularly under the pressure of a career-defining decision.
The finding points to a design gap not in the algorithm, but in how institutions communicate about it. Saying what to do, without explaining why, leaves too much room for misinterpretation.
Why It Matters Beyond Medicine
The implications of this research extend well beyond the physician pipeline. Matching algorithms and similar algorithmic decision-making tools are now used in school admissions, military assignments, public sector hiring, workforce placement and internal corporate talent management systems. Organizations adopt these tools to improve efficiency and reduce human bias — but this study suggests they may be trading one form of inequality for another if they don’t invest equally in user education.
The researchers recommend a range of practical steps: clearer explanations of how matching systems interpret choices, repeated exposure to training materials, simulations and interactive exercises, and active encouragement for users to consult more than one source of guidance. The goal is not just to build a fair system but to ensure that everyone who uses it has a genuine opportunity to use it well.
“Organizations increasingly rely on algorithms to make consequential decisions,” added Skowronek. “If they want those systems to be fair in practice, they need to pay as much attention to implementation, communication and user understanding as they do to the algorithm itself.”
For college students preparing to navigate high-stakes matching processes — medical residencies, law clerkship lotteries, graduate program placements, even some job application pipelines — the research carries a direct message: understanding how a system works is as important as preparing what to submit. Seeking out multiple explanations, engaging with training materials and questioning vague advice can meaningfully change outcomes.
Source: Institute for Operations Research and the Management Sciences