CS 6828 - Modern Prediction Paradigms: Responsible Machine Learning

Fall. 3 credits. Student option grading.

Prerequisite: CS 3780 or equivalent, and CS 4820. Recommended prerequisite or corequisite: CS 4814, CS 4783 and CS 6810.

M. Kim.

Predictive algorithms influence and shape society. The use of machine learning to make predictions about people raises a host of basic questions: What does it mean for a predictive algorithm to be fair to individuals from marginalized groups? On what basis should we deem a predictive algorithm to be valid? And when should we trust (or distrust) a predictor’s output? This course surveys recent developments in the theory of responsible machine learning. We overview new paradigms for formulating learning problems and highlight key algorithmic tools in the study of fairness, validity, and robustness. Topics covered include: Multicalibration and Outcome Indistinguishability, Omniprediction, Performative Prediction, Distributional Robustness, and Verification of Learning.

Outcome 1: Identify common patterns and assumptions underlying modern prediction problems.

Outcome 2: Evaluate, given new settings, whether using machine prediction is appropriate.

Outcome 3: When appropriate, apply principled frameworks for reasoning about prediction (e.g., outcome indistinguishability, performative prediction) to reason about machine learning responsibly.