By Melissa Crow
Information
Among concerns about students relying on AI chatbots to complete statistical tasks, it’s reasonable to ask how trustworthy ChatGPT’s statistical recommendations are, particularly when prompted by introductory statistics students rather than trained statisticians. In this two-part activity, students are first guided through structuring an investigation and choosing appropriate statistical methods for multiple case studies. The included quiz provides useful information on students’ strengths and weaknesses with statistical decision making. In the second part of the activity, students work in pairs to see how well ChatGPT does answering the same questions, using follow-up prompts to pursue specific issues more deeply. Students submit a reflection regarding both their success and their process with ChatGPT, along with their chat transcripts. Instructors can see which types of tasks (if any) students were able to complete with AI assistance, compared to when they worked independently; students benefit by learning both potential limitations of relying solely on ChatGPT’s statistical skills, and also how ChatGPT can successfully augment their existing statistical knowledge. Although developed for a small (<30 student) undergraduate introductory course, there are suggested adaptations for larger classes, alternative formats, and ethical considerations. Downloadable assignment materials are available.