Reimagining the Statistics Classroom for the AI Era: Opportunities and Challenges

Artificial intelligence (AI) tools -- particularly large language models such as ChatGPT, Gemini, and Claude -- are rapidly transforming how students learn and instructors teach. Three years ago, they were mostly a novelty; today, they are forcing instructors to rethink assignments, grading, and even the role of the instructor itself. This panel brings together educators from diverse areas of statistics and data science, at both research universities and primarily undergraduate institutions, to explore both the opportunities and challenges of AI. In terms of opportunities, AI has the potential to lower barriers to entry for students and alleviate the administrative burden on instructors. Jacob Bien will illustrate how coding assistants such as Github Copilot have enabled his students, particularly nonmajors, to "write" R code from English prompts. This allows students to focus on statistical reasoning, rather than coding syntax. Andrew Bray will address the potential of AI to replace human grading of exams, which not only can be laborious, but is also remarkably inconsistent. AI grading can make it possible to scale high quality feedback to massive classes. Dennis Sun will discuss a collaboration with science educators to build an AI problem solving tutor, which guides students towards building a problem-solving plan for homework problems. At the same time, AI threatens the way we motivate and assess learning. Even though the *process* is the most important part of learning, we traditionally evaluate the *product*. AI breaks this model by allowing a student to obtain the product without going through the process. The panelists will discuss creative and thoughtful strategies to put the focus back on the process of learning. Harrison Li will talk about how he has implemented an oral exam early in a probability course to encourage foundational conceptual understanding, and transformed a final project in a linear models course to emphasize presentation skills while allowing LLM use for report and code generation. Dennis Sun will share results from an experiment to adapt Oxford-style tutorials to an American setting. Like their British counterparts, students discuss homework problems in small groups (3 students) with an instructor each week, but unlike Oxford tutorials, these are graded and replace the homework grade. It is important that students understand both the opportunities and the challenges of AI. Julia Costacurta will talk about how she involves her class in the creation of AI norms at the start of each semester, which provides an opportunity to discuss the risks of bias, misinformation, and overreliance on AI.