Q&A with
summary
Concrete claims about methods and deployments: a published algorithm gives fair classification with approximation guarantees across arbitrary group partitions (paper on arXiv and code available), practical tradeoffs where adding constraints degrades accuracy, and a planned experiment measuring search-engine manipulation to quantify feedback loops. Concrete tools include open-source libraries and a proposal for a user-facing diversity slider plus recommendation tweaks that inject cross-community items. Operational tips: sample some denied loan applicants to reduce training bias.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:07Media focus on problems
- 2:17Open-source fairness tools
- 3:20Non-binary fairness handling
- 7:13Freedom versus diversity
- 12:45User control in fairness
- 17:00Personalization vs diversity
- 20:40False equivalence in fairness
- 24:47Editorial responsibility
- 28:01Recommendation algorithms
- 31:30Fair lending constraints
- 35:00Credit scoring challenges
- 37:52Data bias approaches
- 38:34Fairness research vision
- 40:21Closing remarks
related talks
35mOpen Source Fairness
40mQuestions and Practices to Make Algorithmic Decision Making More Fair
27mCrouching Harm, Hidden Bias—Addressing issues of fairness....
47mEthical Machine Learning
35mBuilding Inclusive Machine Learning Models
11mAddressing Privacy, Bias, and Fairness in AI Systems
47mAlgorithmic bias is everywhere (especially at Breeze) - what can we do about it?
27mThe ethical algorithm