By Julie Deeke (University of Illinois Urbana-Champaign)
Information
In a class where generative AI (like ChatGPT, Claude, Copilot, etc.) is allowed as a tool, how do we support students in growing to be critical AI users? For my large, second course in data science at a large, public, midwestern university, I allowed students to use generative AI as an assistant to help them code on course assignments during the previous academic year. As this is the first course in the data science sequence that allows students to use generative AI, I wanted to help students learn that there are limitations and practical considerations for using AI, including that it can make mistakes. My overarching goal is that students will feel comfortable using AI as an assistant when coding while critically evaluating the results that they receive to ensure that the code is achieving the intended task. Therefore, I introduced a set of questions on a Homework 0 to encourage students to construct experiences working with AI where the results might not be accurate or achieve their desired tasks. The Homework 0 assignment was designed as a beginning of semester assignment to orient students to course tools and expectations. In this Beyond session, I’ll share the questions asked and the pitfalls that I experienced in posing these questions to students about working with AI tools. This assignment is used in a large second course in data science with around 500 students enrolled each semester at a large, public, R1 university. This is the first course in the data science sequence that allows students to use AI as a tool to help with coding. Students who enroll in the course are often intending to study data science or statistics as a minor, certificate, or major. They range from second-semester first-year students to seniors. Because of this, the approach is scalable and intended to be inclusive (does not require the use of AI), although I imagine that it could be modified to many different types of students.