Experiences Teaching a Large Upper-Division Data Science Course Remotely

Abstract

This summer, we introduced a number of structural changes to an existing upper-division data science course to optimize it for remote instruction. With hundreds of students situated in various time zones, a purely synchronous model was not feasible. Our course included a menu of synchronous and asynchronous activities, allowing students to consume content in the way they found most convenient and desirable. We presented lectures asynchronously as a series of short prerecorded videos with conceptual questions following each video. We used the results of these conceptual questions to gauge students' preliminary understanding of the material and to identify common misconceptions. Each week ended with a single synchronous lecture recap session that clarified these misconceptions and provided a summary of the week's material. To supplement lecture content, we offered both prerecorded and live discussion sections, as well as live lab sections. Weekly surveys allowed us to adapt our course to address student concerns in real time. Given the online nature of the course, we were able to depart from the orthodox paper-only exam format used by most courses at our institution. Instead, we held each of our three exams in a format optimized for its respective content. One midterm used a web tool that allowed for both multiple choice and free-response coding questions. The other midterm was on paper and required students to scan their exams. The final exam was a hybrid of the two formats. In this experience report, we present the motivation, implementation details, and effectiveness of these changes.