Data Bias: The Danger Lies in the Details
summary
Examines how big data and machine learning systems can inherit and amplify social biases based on the datasets used, questioning how to measure result quality and whether more data alone solves bias problems. Discusses risks coming from detailed correlations (correlation ≠ causation) and explores issues around data sources and algorithmic prejudice in search, classification and identification systems.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 5:30Spurious correlation
- 8:30Confounders
- 12:00Social bubble
- 14:10Biased dataset
- 15:42Bots trained on Twitter
- 17:08Algorithms are not racist
- 18:27Shaving everything
- 19:14Biased data
- 22:01Final considerations
related talks
31mData Discrimination
38mSocial Implications of Bias in Machine Learning
37mDigital Discrimination: Cognitive Bias in Machine Learning
51mDigital Discrimination: Cognitive Bias in Machine Learning
40mData and Algorithmic Bias in the Web
35mZero Bias AI
43mThe Elephant in your Dataset: Addressing Bias in Machine Learning
38mSocial Implications of Bias in Machine Learning