Be good (and don't be evil): how to audit your work for fairness and inclusion
summary
Walks through concrete audits: case studies (Latanya Sweeney ads, COMPAS risk scores), legal pressure in the EU and US, and hands‑on fixes that actually change outcomes. Shows actionable checks—counterfactual testing (change race/gender for one profile), reweighting samples and flip labels for harmed groups—plus cautions about sampling, loss functions and hidden model constraints. Reports measurable results (bias drops from ~16% to ~2% with minor accuracy loss) and workflow steps to implement.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:01Algorithmic discrimination cases
- 8:20Biased algorithms awareness
- 12:59EU and US regulations
- 16:30Why discrimination happens
- 21:04Detecting bias in datasets
- 22:47Legal discrimination definitions
- 27:02Under-representation in data
- 30:36Fairness auditing workflow
- 36:12Q&A
related talks
27mCrouching Harm, Hidden Bias—Addressing issues of fairness....
40mQuestions and Practices to Make Algorithmic Decision Making More Fair
31mFlagging *and* Fixing Bias in ML
37mKeynote: Ethical Machine Learning: Creating Fair Models in an Unjust World
42mFairness in Machine Learning: From Theory to Practice
35mOpen Source Fairness
47mEthical Machine Learning
38mSocial Implications of Bias in Machine Learning