Lecture 5
Bias, Discrimination, and Fairness
Individual Risks
We dive into technical definitions of fairness and examine algorithmic bias through a computational lens. Students will learn about fairness metrics (e.g., demographic parity, equalized odds), how bias can emerge from training data, and how interventions such as reweighting or post-processing affect outcomes. The session also explores limitations of current fairness audits and where technical and legal frameworks diverge. We will further discuss how fairness concerns intersect with civil rights law, public trust, and social legitimacy.