Digital Discrimination: Cognitive Bias in Machine Learning
summary
Describes concrete causes and fixes for algorithmic unfairness, citing IBM contributions and open-source fairness tools that detect bias in training data, measure disparate impact, and apply mitigation like reweighting and post‑hoc calibration. Shows case studies: criminal‑justice risk scores that drive disproportionate arrests and facial/gender classifiers that mislabel darker skin tones, with numbers from deployments and audits used to quantify harm. Demonstrates auditing pipelines, explains evaluation metrics used in production, and reports which interventions reduce measured disparity and which fail to generalize.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:34Digital Discrimination: Cognitive Bias in Machine Learning
- 3:54Examples of Blas in Machine Learning • Solutions to combat unwanted bias. • Tools to combat unwanted bias.
- 33:42AI Fairness 360 Toolkit Open Source Library
- 37:42AI Explainability 360 Toolkit Open Source Library
- 40:32Tool 3: Model Asset eXchange Open Source Pre-Trained Deep Learning Models
- 41:55Data Asset exchange Open Source Data Sets
related talks
37mDigital Discrimination: Cognitive Bias in Machine Learning
38mSocial Implications of Bias in Machine Learning
38mSocial Implications of Bias in Machine Learning
23mRemoving Unfair Bias in the Machine Learning Pipeline
47mEthics & AI: Identifying & preventing bias in predictive models
47mEthical Machine Learning
37mKeynote: Ethical Machine Learning: Creating Fair Models in an Unjust World
35mBuilding Inclusive Machine Learning Models