Finding Fairness - Incorporating Societal Values in Computer Algorithms

Explore the intersection of computer algorithms and societal values in this 48-minute lecture from the Radcliffe Institute for Advanced Study's Fellows' Presentation Series. Delve into the critical concerns surrounding privacy, fairness, and statistical validity as computers and algorithms become increasingly integrated into our daily lives. Learn about differential privacy, data analysis, and the fundamental law of information recovery. Examine the concept of fairness in classification algorithms and its application to group dynamics. Discover how conditional probability and metrics play a role in fair affirmative action via rankings. Gain insights from Cynthia Dwork, a distinguished professor of computer science at Harvard, as she addresses these pressing issues in the field of computer science and its impact on society.

Syllabus

Intro
Sample Space of an Experiment
Differentially Private Data Analysis
Fundamental Law of Info Recovery
Privacy Preserving Data Analysis?
Differential Privacy
Teachings vs Participation
Did You Floss Last Night?
Conditional Probability
Classification Algorithms
Defining Fairness for Groups • Group fairness properties are statistical requirements
Defining Fairness for Groups Group fairness properties are statistical requirements
What should the Metric Capture?
Fair Affirmative Action via Rankings
FAA via Metrics (highly simplified) Sage