Machine Learning and AI Foundations: Prediction, Causation, and Statistical Inference

Introduction

• Prediction, causation, and statistical inference

1. What Is a Casual Model?

• Lady tasting tea
• Why causation matters in a business setting
• What is a causal model?

2. Healthy Skepticism about Our Data and Our Results

• Skepticism about data: Truman 1948 Election Poll
• Skepticism about results: Is that really the best predictor?
• Skepticism about causes: Is X really causing Y?

3. Correlation Does Not Imply Causation

• What is a strong correlation?
• Pearson on correlation and causation
• Correlation and regression
• Challenge: What is causing what?
• Solution: What is causing what?

4. Prediction and Proof in Statistics

• Using probability to measure uncertainty
• p-value review
• Hypothesis testing checklist
• Taleb on normality, mediocristan, and extremistan
• Challenge: Evaluate significant finding
• Solution: Evaluate significant finding

5. Deduction and Induction

• What are induction and deduction?
• Hume on induction
• Popper on induction and falsification
• Taleb on induction
• Counterfactuals: Pearl on induction and causality

6. Prediction and Proof in Data Mining

• Data mining vs. data dredging
• Train/Test: What can go wrong?
• A/B testing during the evaluation phase

7. The Two Cultures: Contrasting Statistics and Data Mining

• The Two Cultures
• Explain vs. predict
• Comparing CRISP-DM and the scientific method
• Applying the two methods at work

Conclusion

• Review