Sex and Gender Analysis Improves Science and Engineering

Sex and gender analysis improves science and engineering places equality inside scientific method: for whom do findings hold, which differences disappear in averages, and what social relations do technologies reproduce? This 2019 Nature Perspective proposes a methodological roadmap, beyond increasing the number of women scientists.

Start with questions and categories

The authors distinguish biological sex characteristics from gender norms, identities and relations, while stressing their interactions. This requires more than adding a male/female column. Researchers must explain why a distinction matters, how it is measured and to whom results apply. Differences cannot simply be presumed in advance.

One ecological example shows how pooling responses can conceal effects running in opposite directions. An image-captioning example concerns a model labeling a man as a woman because he appears in a kitchen. Aggregation hides information in the first case; stereotyped divisions of labor shape recognition in the second. Gender analysis is not an optional appendix limited to human biomedical research.

Representation and design are not substitutes

More women on a team do not automatically change experimental subjects, product dimensions or training methods. The paper therefore addresses funders, journals, universities and researchers together. Questions, sampling, analysis and reporting need coordinated attention. Findings of no difference also deserve reporting, rather than manufacturing difference to demonstrate gender’s relevance.

The roadmap must still resist squeezing complex relations back into fixed binary categories. Finer grouping is not inherently fairer: sample size, privacy, intersections and the purpose of a study matter. Medical examples illustrate methodology; they should not be extracted as individual treatment advice.

Gender Shades examines the distribution of technical errors across race and gender. Data Feminism connects classification and counting to power, extending the question to who participates in defining the research problem.