Privacy, Fairness, and Validity Through the Lens of Theoretical CS

COMPSCI 1261 Subject & Catalog Number

Description

Imagine you are a developer at -- or regulating, from your position as CTO of the Federal Trade Commission -- a company that capitalizes on large amounts of personal data for use in a wide range of settings, from analyzing markets to advising judges on parole decisions, to selecting candidates to interview, to testing drugs. How might you think about incorporating societal values, such as privacy, fairness, and statistical validity? What mathematical guarantees are achievable, and what is impossible? How does privacy differ from cryptographic secrecy, and which concept is appropriate for which setting?

This class will provide an introduction to the theoretical underpinnings of algorithmic fairness, differentially private data analysis, cryptography, and ensuring statistical validity, emphasizing common underlying themes and conceptual breakthroughs.   

Course Notes

CS 1261 is a CS-only variant of CS 1260, and, unlike that course, is not interwoven with a course in at Harvard Law School. Students who have taken CS 1260 may not take CS 1261.

Recommended Prep

Experience writing proofs, familiarity with basic statistical inequalities (Markov, Chebyshev, Chernoff), exposure to theoretical computer science