TITLE: Integrating Reinforcement Learning into the ASSISTments Platform ABSTRACT: ASSISTments is an online learning platform that focuses on K-12 math education. In the past, ASSISTments has run randomized controlled trials to determine the effectiveness of a variety of different ways to engage and assist students as they struggle to master their coursework. Now, there is an ongoing effort to bring reinforcement learning, and specifically multi-armed bandit algorithms, into the platform to recommend, and hopefully personalize, the tutoring students receive. The Reinforcement Learning Service, which will provide students with bandit-recommended tutoring, is being designed as modularly as possible, and will recommend content to students based on features of the students, problems, districts the students are in, and potential tutoring. There are currently open questions as to how to best implement this reinforcement learning service. Which contextual bandit algorithm is a good fit for exploring for personalization? How should student benefits and statistical power be balanced? The ASSISTments team hopes to gain insight into these questions and more through the reinforcement learning workshop. PRESENTER: Ethan Prihar