Academics

  /  

Courses

  /  

Descriptions

COMP_SCI 496: Preference Aggregation for AI Alignment

VIEW ALL COURSE TIMES AND SESSIONS

Prerequisites

CS PhD or Permission of Instructor

Description

The purpose of this course is to understand how techniques and ideas from (computational) social choice can be used for AI Alignment, notably for Reinforcement Learning from Human Preferences (RLHF). The students will read and present a number of recent papers that cover this topic (mostly papers from NeurIPS/ICML/ICLR published from 2023 onwards), and work on research projects that extend the ideas in the papers presented in class.

• This course fulfills the Technical Elective area.

COURSE COORDINATORS: Prof. Elkind

COURSE INSTRUCTOR: Prof. Elkind