Topics will vary from semester to semester. See Computer Science Division announcements.

This graduate seminar is an interdisciplinary introduction to the individual, social, economic, and political effects of AI-driven media, with particular emphasis on design approaches that grapple with these issues. Social media is an important application, but we are also concerned with LLMs and systems like news recommenders, online shopping, job listings, and “trending” algorithms. We will start with the basic question of “who should see what when?” and from there to the social science of recommender effects, algorithmic approaches to grapple with them. Topics include a taxonomy of benefits and harms, models of preference and utility, alignment and optimization, causal inference of user effects, well-being, addiction, polarization and conflict, diversity, filter bubbles and rabbit holes, fairness in information retrieval, incorporation of non-engagement signals into ranking, and the design of user controls. Some topics require technical background (computer science or quantitative social science) but students from all related fields are invited to attend.

Requisites

• Graduate students NOT in the Master of Engineering Program other those in EECS

Repeat Rules