To explain or to predict?
summary
Argues that predictive models win at prediction while regression gives causal significance, and shows concrete tradeoffs: an SVM classifier (trained on split train/test) and feature‑selection pipeline predicts equity‑crowdfunding success, but the regulator forces a switch to an interpretable rule‑based or linear model that loses accuracy. Explains bias–variance choices, rejects PCA/random‑forest when explanation is required, cites cases where a predictor explains only ~5% of variance, and stresses the ethical need to report how deployed decision models actually behave.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 7:20Explanatory vs predictive modeling
- 10:13Equity crowdfunding example
- 16:13Modeling approach differences
- 25:21Takeaways and ethics
- 28:31Q&A feature selection
- 31:25Applications and ethics
- 34:55Trade-offs explained
- 37:48Closing remarks
related talks
17mExplaining models: an empirical study of how explanations impact fairness judgment
26mStop Explaining Black Box Models, Use Interpretable Models Instead
17mChallenges Towards Production-Ready Explainable Machine Learning
25mOpening Remarks
34mInterpreting Machine Learning Models: Why and How!
37mCreating correct and capable classifiers
48mTBA
28mBeyond Prediction Performance: How Modeling Decisions Shape Fairness Outcomes in Statistical...