Reducing Bias in Artificial Intelligence...
summary
Reproduces a concrete example using a public diabetes patient dataset and a Python script that uses pandas and matplotlib to explore bias. Trains a logistic regression with scikit-learn (baseline accuracy ~0.59), measures disparities using MetricFrame to find groups with higher false negatives, then applies ThresholdOptimizer to reduce the false negative rate from about 0.6 to about 0.44 and shows the before-and-after numbers.
AI-written summary · about
chapters (15)
- 0:00Welcome
- 2:00Fairlearn intro
- 5:07AI bias in healthcare
- 7:31Algorithm unfairness causes
- 9:06Data bias effects
- 11:10Fairlearn explained
- 12:27Diabetes patient data
- 15:09Hospital data bias
- 16:46Patient classification balance
- 18:21Model evaluation before mitigation
- 21:32MetricFrame for injustice
- 23:58False positives and negatives
- 25:48Algorithm bias reinforcement
- 26:10ThresholdOptimizer solution
- 28:10Fairlearn results
related talks
53mResponsible AI with SciPy
23mRemoving Unfair Bias in the Machine Learning Pipeline
47mFight AI Bias With… Bias
01h 00mThe Elephant in your Dataset: Addressing Bias in Machine Learning
47mThe Elephant in your Dataset Addressing Bias in Machine Learning
43mThe Elephant in your Dataset: Addressing Bias in Machine Learning
55mEthical Machine Learning
39mBeyond Bias: Crafting Fairness in Pythonic Machine Learning Systems