Burkett, C., Blake, S., & Junco, E. (November, 2024). A content analysis of ChatGPT lesson plans targeting misconceptions. National Science Teachers Association National Conference on Science Education, New Orleans, LA.

Generative AI chatbots have received a great deal of attention recently with educators divided on the potential usefulness of the technology. Some argue chatbot use has potential to reduce beliefs in misconceptions and the dissemination of misinformation. However, little empirical work has investigated the tool’s potential for educators to create and refine lesson plans targeting misconceptions in science. The current study provides a content analysis of lesson plans created using ChatGPT to address ten common misconceptions for elementary students. Plans were analyzed for types of content included. As well, a relational analysis was performed to investigate how concepts within the lesson plans were connected and were coded for their potential to address the targeted misconceptions. Results highlight the similarities and differences of content and demonstrate the potential for chatbot use in guiding lesson plans to target misconceptions when used with appropriate skepticism.

Burkett, C., Wood, N., McFarland, P., Largest, T. & Stowers, A. (2024, April) The role of graphs in confidence about consistency decisions. Midwestern Psychological Association (MPA) 96th annual meeting, Chicago, IL.

Learners are often overconfident in their knowledge, yet little research has examined confidence in text/graph comparison-based decisions. This study used a 2 × 2 × 2 mixed design (N = 250 undergraduates) to test how graph density, claim complexity, and consistency (contradictory vs. consistent) affect decision confidence. Participants judged whether eight text/graph pairs matched and rated their confidence. Results showed confidence was high across conditions and unrelated to decision accuracy. Instead, graph literacy, self-efficacy, and graph density predicted confidence ratings. These findings suggest learners rely on cues other than performance when evaluating multiple representations, underscoring the importance of graph features and literacy in shaping confidence.

Burkett, C. (2022, May). Something’s not right, but what? Complexities and confidence in the identification of contradictory elements between text and graph. Paper presented at the Annual Meeting of the Association for Psychological Science (APS), Chicago, IL.

Critically evaluating multiple representations is vital in today’s world. Undergraduates (N=66) were asked to identify sources of contradictions between text and graph when graphs and type of contradiction varied in complexity. Results confirmed identification of contradictions was more accurate when complexities were lower, despite high confidence ratings in all conditions.