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

This course will explore the intersection of machine learning (ML) and human behavior: both how ML can help us to better understand human behavior and how human behavior introduces new ML challenges. The format of the course will be a mix of paper presentations, lectures, and a final project. In the first half of the course, we’ll focus on how ML can help us to better understand human behavior. First, we’ll cover large-scale behavioral analyses, where ML can infer real-world behaviors from novel data sources (e.g., social media, mobility data) and analyze these behaviors at scale across millions of individuals, leading to new insights in public health, social sciences, transportation, and more. We’ll also cover the growing literature on developing generative AI models to simulate human behaviors. We’ll discuss opportunities in simulation, from predicting individual-level survey responses to simulating entire societies, as well as challenges uncovered in this new literature, including validation, bias and ethics, and scalability. In the second half of the course, we’ll focus on how human behavior introduces new ML challenges, focusing on ML systems that seek to predict human behavior or outcomes affected by human behavior. We will cover a non-exhaustive set of these challenges, possibly including topics such as (1) limits to prediction, exploring fundamental limits in predicting human behaviors and life outcomes, (2) missing data and selective labels, where behaviors or outcomes are selectively observed for some individuals and not others, sometimes as the consequence of past human decisions (e.g., hiring decisions), and (3) distribution shifts, performative prediction, and feedback loops, where the deployed ML model’s predictions causes changes in individuals’ behaviors and the target outcome that the model is trying to predict.

Requisites

• Students not in the Master of Engineering Program

Repeat Rules