Note from Justin: If you are a UMass student and logged into your umass.edu email address, you should be able to see a link at the bottom of this page called “Information for UMass students”. If you can’t see that link please:

1. Double-check that you are logged into Notion with your @umass.edu email.
2. Let me know (either by commenting here or emailing me)

Course: Making Predictions

Course number: COMPSCI 590L

Instructor: Justin Domke

Area: AI

Prerequisites: Undergraduates: Senior and Junior CS Majors who have completed COMPSCI 383 with a grade of C or better. Graduates: No formal prerequisites. Informally, students should be familiar with calculus (e.g. MATH 131, 132) and basic probability theory.

Catalog description: How can we make predictions? The traditional approach in computer science is machine learning. However, this question is addressed in many ways in different fields. One approach is to simply "guess", in which case cognitive biases are important. Another approach might be to identify people who are good at predicting. But do such people exist? And how can we combine their judgment. Economics suggests prediction markets, where people compete for financial reward. This course will cover many different methods for making predictions. The goal is to understand the strengths and weaknesses of each.

Learning Objectives:

• Describe common well-validated cognitive biases in prediction-making.
• Be able to describe the precise guarantees of common inferential statistics (frequentist confidence intervals, Bayesian posteriors) along with the associated guarantees.
• Describe what forecasting tournaments have shown about the difficulties of forecasting future events.
• Create informative visualizations of datasets and draw informal conclusions from them.
• Mastery of the probabilistic view of prediction, including updating and Bayesian inference.
• Describe the ideas behind prediction markets, and for what prediction tasks these would be most successful.

Tentative course syllabus:

• Introduction: Goals of the course, overview of different prediction methodologies -- 2 lectures
• Guessing: Cognitive biases involved in making predictions -- 2 lectures