AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Bias Out-of-the-Box: An Empirical Analysis of Intersectional Occupational Biases in Popular Generative Language Models

Hannah Rose Kirk, Yennie Jun, Filippo Volpin, Haider Iqbal, Elias Benussi, Frederic A Dreyer, Aleksandar Shtedritski, Yuki Asano

NeurIPS 2021 · Large language models › Language, safety and interpretability · approach: Bias measurement and mitigation · 26 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Prompt GPT-2 with woman rather than manGPT-2 Small · Occupational promptsjobs accounting for 50% of outputs: 16 -> 8Hurt
Prompt GPT-2 with woman rather than manGPT-2 Small · Occupational promptsjobs accounting for 90% of outputs: 66 vs 43Hurt
Add the woman dummy to the regressionGPT-2 occupational predictions · Occupational prompts across intersectionsMcFadden R²: +3.3%Improved
Add gender-category interaction termsGPT-2 occupational predictions · Occupational prompts across intersectionsMcFadden R²: +0.4%Improved
Compare GPT-2 predictions with US labor-market distributionsGPT-2 · US occupational data, baseline womanKendall-Tau: 0.628Improved
Use GPT-2 predicted shares instead of US labor sharesOccupations with less than 25% women-share · US occupational datadifference in women-share: +8.7%Mixed
Use GPT-2 predicted shares instead of US labor sharesOccupations with more than 75% women-share · US occupational datadifference in women-share: -6.5%Mixed
Prompt with explicit demographic intersections rather than continental-origin namesGPT-2 · Occupational associationsintersectional effect on returned jobsMixed
Set top_k to 1 rather than the default 50 (ablation)GPT-2 Small · Baseline man and woman occupational promptsunique jobs: 228 -> 1; 250 -> 1Hurt
Set top_k to 10 rather than the default 50 (ablation)GPT-2 Small · Baseline man and woman occupational promptsunique jobs: 228 -> 51; 250 -> 30Hurt
Set top_k to 100 rather than the default 50 (ablation)GPT-2 Small · Baseline man and woman occupational promptsunique jobs: 228 -> 123; 250 -> 126Mixed
Set top_k to 500 rather than the default 50 (ablation)GPT-2 Small · Baseline man and woman occupational promptsunique jobs: 228 -> 233; 250 -> 204Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy