Topics in Machine Learning: Effective AI Support in Human+AI Settings
COMPSCI 2822R
Subject & Catalog Number
Course Information
Description
Agents are developing capabilities at increasing rates, but do those agent capabilities translate into human capabilities, with respect to creating agents that help us do the things we need to do and be who we want to be? In particular, by doing some short-term task e.g. writing a piece of code, current agents often fail us in longer term tasks e.g. making sure that we can also maintain that code and take responsibility for that code to our teams. As another example, a health assistance agent that is optimized to create engaging visualizations of your health data might be satisfying to look at, but what we really want to evaluate is whether it actually helps you improve your health. In this course, we'll survey the literature and create AI systems that are designed to effectively support human tasks via a combination of reinforcement learning, probabilistic methods, and human factors/design elements. After a few initial assignments, the course will be focused on reading papers, discussion, and a semester-long project.
Course Notes
This course was previously numbered CS 282R.
Recommended Prep
Students are expected to be fluent in basic linear algebra (matrix manipulation), basic statistics (e.g. rules of expectations, importance sampling), basic reinforcement learning (at a CS1810/181 level), and basic software engineering (e.g. programming in Python and numpy, data with 10,000+ rows).
Available for Harvard Cross Registration
