Gender Differences in Research Topic and Method Selection in Library and Information Science: Perspectives from Three Top Journals

Researchers studied thousands of academic papers to see if men and women choose different topics and methods. They found that women often prefer topics like health information and methods like interviews, while men favor topics like information retrieval and theoretical approaches.

The study explores a subtle tension in how scientific knowledge is produced. Traditionally, the choice of a research method is thought to be dictated by the research topic. If you want to know how a search engine works, you use an algorithm. If you want to know how a person feels, you use an interview. This implies that research methods are gender-neutral tools. However, emerging evidence suggests that gender may influence the methodological design itself.

Beyond the topic-method trap

Current understandings of gender disparities often focus on the "what"—which topics are being studied—rather than the "how." It is well-documented that men and women gravitate toward different research fields. For example, women in management often emphasize human-centered research. Men may focus on more technical sub-fields. Because research topics naturally dictate certain methods, these differing interests can create a statistical illusion. If women study people more often, they will naturally appear to use qualitative methods (methods involving non-numerical data like text or speech) more often.

The authors argue that previous research failed to decouple these two variables. Earlier studies could not determine if men and women were simply studying different things. They could not tell if scholars were approaching the same topics with different methodological lenses. To solve this, the researchers analyzed 5,281 articles from three top Library and Information Science (LIS) journals. They wanted to see if gendered preferences for specific methods persisted even when the research topic was held constant.

Scaling classification with CogFT

Manually labeling thousands of papers for their specific research methods is a labor-intensive task. To bypass this, the authors developed CogFT (Cognize Full Texts), an automatic classification model. It extracts methodology from the entire body of a paper rather than just the abstract.

The CogFT architecture operates in two distinct stages, as illustrated in :

1. Full-text Summary: The model segments the paper into 128 blocks. It utilizes SciBERT (a language model pre-trained on scientific text) to create embeddings (mathematical representations of text meaning). A self-attention based global encoder then determines the relationships between these blocks. It outputs a probability score for how critical each block is to describing the research method. This acts like a high-speed highlighter for the most relevant text.
2. Automatic Classification: The model takes the four most probable blocks and passes them through a classification network. This network uses SciBERT and a Sigmoid layer (a function that maps values to a probability between 0 and 1) to assign the paper to one or more of 16 specific research method categories.

By focusing on the full text, the authors addressed a common pitfall. Abstracts often omit the granular details of how a study was actually conducted.

Persistent patterns in method selection

The authors report that gender differences in method selection remain significant even after adjusting for research topics. While the gender distribution of the sampled first authors was roughly 60% male and 40% female, the methodological choices diverged sharply.

The paper finds that female authors show a marked preference for interactive and observational methods. Specifically, the authors report that female authors' shares of papers using "Interview" (65.60%), "Questionnaire" (54.18%), and "Observation" (64.22%) were substantially higher than the overall gender ratio. Conversely, male authors showed a clear preference for more detached or structural methods. The study finds that male authors were more likely to utilize a "Theoretical approach" (30.16%) and "Bibliometrics" (25.17%).

Crucially, the authors used a two-dimensional heat map to visualize these preferences across different topics .

They found that even in topics typically dominated by one gender, certain methodological biases persisted. For example, female authors displayed a strong inclination toward using interviews across five different topics. This occurred even in the "Information organization/information management" topic, which is traditionally more male-favored.

Cognitive styles and educational gaps

The study identifies several potential drivers for these disparities. These range from social epistemology (the study of how we know what we know) to early educational experiences. The authors suggest that these differences might reflect "associative" versus "independent" cognitive styles. In associative cognition, individuals acquire knowledge through close interaction and active listening. This style aligns well with the "Interview" method. Independent cognition involves applying reasoning and analysis to authoritative sources. This aligns with a "Theoretical approach."

However, the authors also point to systemic factors. The availability of certain methods often depends on mathematical confidence. Because of prevailing stereotypes regarding mathematical aptitude, women may be less encouraged to pursue mathematically intensive research designs.

The researchers acknowledge several limitations. First, the study is limited to three specific LIS journals. This may not represent the entirety of the field. Second, the gender inference process relied on matching names to datasets. This resulted in a binary gender label that does not capture the full spectrum of gender identities. Finally, the authors note that their quantitative approach reveals correlations but cannot definitively prove causality.

A verdict on methodological diversity

The study provides evidence that research methods are not gender-neutral in practice. The authors conclude that these disparities are not just about what is being studied, but how scholars perceive and interact with knowledge.

For academic institutions and practitioners, the goal should be to broaden proficiency. To bridge the gap in scientific impact and innovation, the authors suggest promoting cross-gender collaboration. They also recommend encouraging students to master a wider array of methods. This helps challenge the gender-based anxieties that may limit their toolkit. Whether these patterns are rooted in innate cognitive styles or systemic educational biases remains an open question. However, the existence of the pattern is now clearly documented.

Figures from the paper