ChatGPT in Statistics and Data Learning

ChatGPT is a powerful artificial intelligence (AI) tool that excels at generating human-like text for a wide range of applications including content generation and creative writing.  ChatGPT has been found to be a useful learning tool in classrooms by acting as a virtual tutor with personalized lessons, summarizing lengthy content and writing computer code. However, there are still concerns with using this tool in the classroom including the accuracy of conceptual ideas and ethics of not citing proper sources. In this study, researchers investigated the application of ChatGPT to specifically statistics and data analytics courses and provided guidance for teachers and students to successfully use this AI tool in data science classrooms.  

Responses from ChatGPT were assessed on topics related to linear regression, since linear regression is a widely recognized and popular method for data analysis used by professionals and non-experts.  Ten topics were examined through twenty-nine different prompts into ChatGPT relating to conceptual ideas, data analysis and teaching preparation regarding linear regression.  If an incorrect or unclear response was generated from ChatGPT a new discussion was started, or the conversation was continued to provoke an improved response.  An example of an initial prompt was “What is the reason for using adjusted r-squared in multiple linear regression instead of r-squared? Write a short answer.” The consistency of ChatGPT responses were evaluated through slight modifications of the same question.  

“Overall, ChatGPT should be considered a teacher whom we gain a broad understanding from, a peer with whose help we develop critical skills, an assistant we supervise, and a student in whose work we identify and rectify errors.” 

Conceptual statistics: Instructors and students should verify qualitative statistical information from ChatGPT.  For example, when asked “How do the two metrics MSE and ASE in predictive modeling differ?”, ChatGPT confused MASE with ASE in the response.  

Data analysis: ChatGPT can be a good tool for finding or creating synthetic datasets and guide data analysis. Writing code, such as Python or R, for statistical applications was found to be a strength of ChatGPT.  

Teaching and assessment: ChatGPT can be an efficient tool to help instructors create problems for assessments and design lesson plans.