Fair Matching Systems Can Still Produce Unequal Outcomes, New Research Finds

Study of medical residency matching shows fairness depends not only on how systems are designed, but how well people understand and use them

BALTIMORE, May 15, 2026 — A computerized matching system can be designed to be fair and still produce unequal outcomes if the people using it do not understand how it works.

That is the central finding of new research published in Organization Science, an INFORMS journal, showing that disparities can emerge even when a matching system is designed to reduce bias, discourage gaming and reward honest decision-making.

The study, “Gendered Navigation of Advice and Suboptimal Behavior in Matching Algorithms: Evidence from the Residency Match,” examines how medical students navigate the National Residency Matching Program, the high-stakes process that determines where future physicians will train. The researchers found that unequal outcomes can arise not from bias inside the algorithm, but from differences in how applicants seek information, interpret advice and understand how the system works.

The residency match uses a computerized system to pair graduating medical students with residency programs based on the ranked preferences submitted by both sides. The system is designed so students are best served by ranking programs in their true order of preference rather than trying to game the process.

In theory, that should level the playing field.

But the study found that many students still made suboptimal ranking decisions in part because they misunderstood how the system works. Some ranked less-preferred programs higher because they believed it would improve their chances of matching, even though that strategy can reduce their odds of receiving their best possible placement.

“Algorithms do not operate in a vacuum,” said Samuel E. Skowronek, one of the study’s authors. “Even when the algorithm cannot be gamed, outcomes still depend on whether people have the knowledge and support needed to use it correctly.”

Using data from more than 1,700 medical students who participated in an incentivized simulation of the residency match, along with 66 in-depth interviews with students navigating the real match process, the researchers found a consistent pattern: male students were more likely than female students to independently seek out additional information about the algorithm.

That difference mattered.

Students who consulted multiple sources, revisited training materials, watched explanatory videos or searched for independent guidance developed a stronger understanding of the system and were more likely to use it optimally. Students who relied primarily on standard institutional advice were more likely to misunderstand the process and make decisions that weakened their outcomes.

The researchers found that women in the study reported lower confidence and understanding of the algorithm and were more likely to submit rankings that deviated from the optimal strategy. The disparity did not come from the system treating applicants differently. It emerged from the human behaviors surrounding the system.

“This research broadens the conversation around algorithmic fairness,” Skowronek said. “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 findings have implications far beyond medical residency.

Computerized matching systems and related algorithmic tools are increasingly used in school admissions, military assignments, public sector hiring, workforce placement, promotions and internal talent marketplaces. Organizations often adopt these systems to improve efficiency, reduce subjectivity and create fairer outcomes. But the study suggests that organizations may be overlooking a critical source of inequality: unequal user understanding.

The researchers argue that institutions should not assume that sound system design is enough. They should also invest in better explanation, training and support for the people expected to use these systems.

Practical steps could include clearer descriptions of how matching systems process choices, repeated exposure to training materials, simulations, interactive exercises and stronger encouragement for users to consult multiple sources of information.

The study also cautions against overly simple guidance. Many students interviewed for the research described institutional advice as limited to messages such as “rank programs based on your true preferences” or “follow your heart.” While that advice may be technically correct, the researchers found it was insufficient for many students. Without understanding why the strategy works, applicants may still act on fear, uncertainty or incorrect assumptions.

The broader lesson is clear: organizations cannot treat algorithmic fairness as a purely technical problem.

Even well-designed systems can reproduce inequality when users enter the process with different levels of information, confidence and support.

“Organizations increasingly rely on algorithms to make consequential decisions,” Skowronek said. “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.”

Read the full study here.

About INFORMS and Organization Science

Organization Science is a premier peer-reviewed scholarly journal focused on the fields of strategy, management, and organization theory. The journal publishes groundbreaking research about organizations, including their processes, structures, technologies, identities, capabilities, forms, and performance.

INFORMS serves as a global authority in advancing cutting-edge practices and fostering an interdisciplinary community of innovation. With a network of more than 12,000 members across academia, industry, and government, INFORMS connects thought leaders and emerging professionals who apply science and technology to solve complex challenges and drive better decision-making.

Through its prestigious journals, world-class conferences, certification programs, and professional resources, INFORMS empowers its community to enhance operational efficiency, elevate organizational performance, and promote smarter decisions for a better world.

Discover more at www.informs.org

###