Studying the epistemic value of generative AI in modelling activities
Speaker: Louise Meier Carlsen, ITU
Abstract
Prior research, when examining the epistemic value of digital resources such as dynamic geometry systems or use of the internet and CAS, indicates the lever potential of these regarding modelling activities. Generative AI (GenAI) has become an active tool in students’ educational practice, including modelling activities. When students employ GenAI, it serves as a dynamic media that supports the inquiry process of modelling and introduces new knowledge to be explored. In contrast, we have static media such as textbooks, webpages and videos. The inclusion of dynamic media into modelling activities suggests new possibilities for task design that foster rich modelling and learning activities. This early research on students’ use of GenAI indicates that students use dynamic media to study the modelling process more generally while using static media for knowledge of specific techniques.
Our study is situated in a statistics course. However, recent research has pointed to the close relation and potential benefits between statistics (including statistical modelling) and mathematical modelling.
This study investigates the students’ use of dynamic media and what type of knowledge the dynamic media is used for. We consider a context of data science students in Denmark, where students engage with modelling in data science during the early weeks of a statistics course, where they are asked, “Which exam format is best?”. The students are thus still unfamiliar with standard models and traditions of statistical analysis. Specifically, we examine the students’ use of GenAI when students are asked to both model and conduct a statistical analysis. We therefore ask: What specific knowledge do students seek using dynamic media? How does dynamic media support students’ development of statistical and modelling knowledge?
To address these questions, we analyse students’ logbooks containing GenAI prompts and responses, links to all digital resources consulted, as well as written project reports that were handed in. Our work is guided by the view of modelling from the perspective of the Anthropological Theory of Didactic, which is a process initiated by a generating question which the students cannot immediately answer but must engage in a process of posing derived questions which requires a study and research process where new knowledge is sought and researched in the context of the generating question. The analysis is based on the notion praxeology that lets us explicit knowledge into the categories of type of tasks, technique, technology (discourse) and theory (validation of technology). This framework allows us to explicitly study the knowledge studied using dynamic media and consider the relations and disconnects between the elements.
Our analysis shows that GenAI can contribute knowledge on all levels, including norms and traditions for conducting statistical analysis. Furthermore, the students also used GenAI to develop derived questions to pursue in the modelling process, as well as how to achieve answers to these questions.