Bias in evaluating code contributions

on YouTube ↗10m

" summary

Claim: inferred race and ethnicity from names using NamePrism on about 2.5 million pull requests across 46,000 GitHub projects yields concrete disparities. The dataset maps ~493,000 developers and shows less than 10 percent non‑white contributors. Regression controlling for experience finds Hispanic contributors ~6% lower odds and Asian/Pacific Islander ~10% lower odds of acceptance; same‑ethnicity integrators dramatically raise acceptance (Black contributors up to nine times higher). Authors release data and scripts and recommend using names to proactively support diverse contributors.

AI-written summary · about

" chapters (10)

1. 0:00BIAS IN EVALUATING CODE CONTRIBUTIONS IDENTIFYING DIVERSITY IN ONLINE GROUPS AND CREATING AWARENESS
2. 0:58BUT, DIVERSITY ASPECTS MIGHT AFFECT THE ACCEPTANCE OR REJECTION OF THE CONTRIBUTIONS
3. 2:09INTEGRATORS' PERCEPTION OF RACE AND ETHNICITY OF SUBMITTERS MIGHT ACTIVATES WHATEVER BIASES THE INTEGRATORS HAVE ABOUT PEOPLE WITH THE SAME RACE AND ETHNICITY OF SUBMITTERS
4. 2:59A LARGE SCALE STUDY ABOUT THE PERCEPTION OF RACE AND ETHNICITY IN PULL REQUEST EVALUATION
5. 3:3557.2% OF GITHUB DEVELOPERS HAVE A PERCEPTIBLE RACE AND ETHNICITY, BUT 42.8% DEVELOPERS ARE CLASSIFIED AS UNKNOWN
6. 5:18DOES PERCEPTIBLE RACE AND ETHNICITY HAVE A RELATIONSHIP WITH THEIR CONTRIBUTIONS? Dependent
7. 5:52THE RELATIONSHIP OF HISPANIC AND API DEVELOPERS ON PULL REQUEST ACCEPTANCE IS LOWER WHEN COMPARED TO PERCEPTIBLE WHITE DEVELOPERS
8. 6:45NON-WHITE DEVELOPERS ARE MORE LIKELY TO GET THEIR PR ACCEPTED WHEN THE INTEGRATOR IS FROM THEIR SAME RACE AND ETHNICITY (COMPARED TO PERCEPTIBLE WHITE DEVELOPERS)
9. 8:39SO WHAT NOW? AUTHOR BLINDED EVALUATION?
10. 8:50MAYBE? INSTEAD WE CAN ACTIVELY SUPPORT A DIVERSE GROUP OF CONTRIBUTORS.

" related talks

• 39mState of diversity in open source communitiesHolden Karau · Berlin Buzzwords 2019
• 38mKeeping Your Open Source Community Accessible to AllTreva Williams · Community Leadership Conference @ Open Source Summit North America 2022
• 55mDiversity in OpenSource … show me the data!Denise Dumas, Serena Chechile Nichols · DevConf.CZ 2020
• 45mBias in Artificial IntelligenceNjoki Kamau · Inclusive Design 24 (#id24) 2020
• 53mCoding Like a GirlGabriela D'Ávila · DjangoCon US 2015
• 30mCommit to Inclusion: Building Better Open Source CultureAshley Sametz · FOSS Backstage 2025
• 22mCreative Approaches To DiversityKatharina Borchert · Open Source Leadership Summit 2017
• 47mDiversity and Inclusion for Algorithmic Bias ConsiderationsAnsgar Koene · PyCon India 2020