Machine / Algorithm Bias (Data Science, Machine Learning)

Our lives are, more and more, dependent on decisions that are taken by algorithms: from a meal that we want to order to a potential partner that we want to date. What if these algorithms are biased? What if a completly different meal will never be present amog that the App offers to us? Worse, what if a program decide that a person is ranked as a "high risk" based on color skin, neighbhood where he/she lives... These are small examples of the impact that biased algorithms make in our lives.

Transcript

1. GDG Campinas | DataFest 2018 Machine Bias Campinas, SP, Oct, 31th 2018
2. Curioso e data lover. Trabalha em projetos de desenvolvimento de software a mais de 25 anos. Tem 4 filho/as, 2 cachorros e é jogador (bem amador!) de squash. Davi Carvalho - Developer e GP de plantão na Dextra
3. 2.5 quintillion bytes
4. In God we trust; all others must bring data -- W. Edwards Deming
5. Machine Bias
6. Algorithmic Bias
7. Muhammad ibn Musa al-Khwarizmi
8. Model
12. Algorithmic Bias
13. Where Bias Comes From? • AI systems are taught from training data • People labels much of this data by hand • We choose what AI system will learn • AI field is not diverse
14. "I'm not capable of being wrong" -- HAL 9000
17. Manage data Train models Evaluate models Deploy models Make predictions Monitor predictions Machine Learning Workflow
18. Trained Data Label ML Algorithm New Data Predictive Model Prediction
19. Correctional Offender Management Profiling for Alternative Sanctions COMPAS
20. Bernard Parker Dylan Fugett 10 3
22. https://www.propublica.org/article/facebook-ads-age-discrimination-targeting
24. Cosine Similarity man woman nurse physician θ θ A B A B
25. DEEP NEURAL NETWORK SCAN DETECT SEXUAL ORIENTATION FROM FACES - VGG-Face: DNN to extract facial features - Training classifiers: prediction model with logistic regression + singular value decomposition (SVD), that is similar to principal component analysis (PCA), a dimensionality-reduction approach.
27. "Weapons of Math Destruction" (Cathy O'Neil)
28. INCLUSIVE CODING - Who Codes Matters - How We Code Matters - Why We Code Matters
29. "This is Water" (David Foster Wallace)
30. - [Old fish] "Morning, boys, how's the water?" - [Young fish] "What the hell is water?"
31. davi.silva@dextra-sw.com @davisilva