Decentralizing AI Fairness Decisions
summary
A Netherlands case study shows a municipal ML variable‑selection profiler used to flag welfare recipients, where a mixed expert commission plus affected-person input produces a red/green variable list that excludes blatantly illegal proxies; the municipality pauses the system after audits find unlawful parameters. The panel pushes algoprudence—publishing FRIAs and rationale for included/excluded features—recommends randomized sampling to validate models, and requires combined algorithmic rules plus human oversight rather than blind automation.
AI-written summary · about
chapters (9)
- 0:00Decentralizing AI fairness
- 2:12Risk profiling case study
- 9:18Deliberative fairness panel
- 18:52Institutional perspectives
- 24:56Power in algorithmic governance
- 33:00Oversight versus decentralization
- 41:40AI Act fairness assessments
- 55:26Q&A: Process vs outcomes
- 1:04:14Final remarks
related talks
29mResponsible AI Challenges and Recommendations
47mAI Ethics, Strategic Decisioning and Game Theory
49mAn AI with an Agenda
47mEthical Machine Learning
27mCrouching Harm, Hidden Bias—Addressing issues of fairness....
19mMoral Outsourcing: Humanity in the age of AI
36mKeynote: Ethics in AI: Building Fair and Just Intelligence Systems
29mThe Political Roots of AI Policy