Before You Get Too Excited About That GitHub Study…

A careful methodology-critique of the viral 'Tech Is Sexist: women's GitHub pull requests rejected more' study. Scott shows the study actually found women get MORE requests accepted overall; the 'bias' rests on one subgroup (outsiders, gender-revealed: men 64% vs women 63%, maybe not significant); BOTH sexes' acceptance drops when gender is revealed (the male drop is downplayed); no Bonferroni correction across subgroups; no control for the study's own finding that women's changes are larger/less-needed; the tiny effect is mostly women-on-women; and the paper is NOT peer-reviewed (an undergrad preprint). Contrasted with the media's wild spin ('a vile male hive mind is running an assault mission against women in tech'). A model of careful study-reading + media-misrepresentation exposé, with constructive peer-review suggestions. Critiques a study rather than relying on one -> no caveat.

Quality 74 · Strong. A sharp, patient debunking that reconstructs what the study actually found vs the media spin, with solid statistical hygiene (subgroups/Bonferroni/uncontrolled confounds/not-peer-reviewed). Strong.

Claude’s paradigm shift 45 · Moderate. Moderate. A careful application of good statistical hygiene to a specific viral study; little new framing.

Real-world impact 2 · Minor. Minor (2) — a study-critique/media-criticism essay with within-sphere reach; no concrete real-world change.