Gendered Navigation of Advice and Suboptimal Behavior in Matching Algorithms: Evidence from the Residency Match

Key findings from this study

• The study found that men engage in independent advice-seeking about algorithm mechanics more frequently than women, seeking information beyond baseline institutional guidance.
• The researchers demonstrate that enhanced help-seeking behavior correlates with deeper understanding of the matching algorithm and superior matching outcomes.
• The authors report that algorithmic outcome disparities can arise from gender-differentiated navigation strategies rather than from algorithmic bias in the matching mechanism itself.

Overview

This study examines how gender shapes the navigation of two-sided matching algorithms in labor markets, specifically investigating whether gendered patterns in help-seeking behavior contribute to disparities in algorithm outcomes. The research focuses on the National Residency Matching Program (NRMP), which allocates graduating medical students to residency positions in the United States using a matching algorithm. Drawing on theories of gendered agency, the authors hypothesize that men engage more frequently in independent advice-seeking beyond institutional guidance, resulting in deeper algorithmic understanding and improved matching success compared to women.

Methods and approach

The research employs a mixed-methods design combining archival and qualitative data. Quantitative analysis draws from medical students' responses in an incentivized simulation of the NRMP algorithm, capturing behavioral patterns in help-seeking and matching outcomes. Qualitative data derives from 66 semi-structured interviews with medical students navigating the actual match process. This combination permits examination of both measurable behavioral differences in algorithm navigation and the underlying mechanisms through which gender influences decision-making strategies and information acquisition patterns.

Results

The study provides empirical support for the central hypothesis that gender differences in independent advice-seeking contribute to disparities in algorithm outcomes. Men demonstrated significantly greater propensity than women to seek additional information beyond baseline institutional guidance about algorithm mechanics. This elevated help-seeking behavior among men correlated with improved algorithmic understanding and superior matching results. The findings indicate that outcome disparities emerge not from algorithmic bias but from gender-differentiated patterns in how individuals acquire and process information about algorithm operation.

Implications

These results demonstrate that equity concerns in algorithmic matching extend beyond the technical properties of matching mechanisms themselves. Group-based performance disparities can manifest through differential engagement with available informational resources and heterogeneous learning strategies shaped by social identities. This finding suggests that institutional efforts to promote equitable outcomes require attention to the social processes through which market participants develop understanding of algorithmic systems, not merely the formal design of algorithms. The mechanism identified—gendered patterns in agentic information-seeking—may operate across diverse matching markets and algorithmic contexts beyond medical residency allocation.

The research indicates that baseline institutional advice provision may be insufficient to equalize outcomes when individuals have differential propensities to supplement such guidance through independent investigation. Institutions deploying matching algorithms might consider interventions that systematically expose all participants to equivalent informational resources or explicitly encourage supplementary learning independent of self-initiated help-seeking. Understanding identity-based variation in algorithm navigation strategies provides a foundation for designing more inclusive institutional structures around algorithmic market participation. Future research might examine whether observed gender differences persist across demographic contexts or interact with other dimensions of identity.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.