AI Mistakes in the Classroom

Large language models and their rapid, sweeping impact on education have only just begun to be documented in the teaching and learning literature. In quick succession, many publications on the acceptance and use of artificial intelligence in the classroom have appeared. The literature primary fell into two categories: (1) how-to guides for incorporating AI into the classroom (e.g. Mollick & Mollick, 2023; Noy & Zhang, 2023; Winkler & Ross, 2019) and (2) concerns about how it will be used for cheating (e.g. Amani, et al., 2023; Barros et al., 2023; Michel-Villarreal et al., 2023; Okaiyeto et al., 2023). While not explicitly stated in most articles, an underlying learning theory that strongly supports the inclusion of AI in the classroom is Connectivism. This theory can explain why AI in the classroom supports learning. However, what is currently missing from the literature are documented failures caused by incorporating AI into assignments and examples of how such inclusion has hindered learning. This paper attempts to begin to fill that gap. The following is a literature review showing how inclusion of AI supports the Connectivism view of learning, followed by three examples of mistakes when building AI based assignments and how they violated the Connectivist view of learning, leading to frustration by students and professors.