Affordances and Challenges of GenAI Feedback in Primary Students’ Scientific Drawing

Abstract

GenAI can be a potential tool to facilitate students’ scientific practices through automated and real-time feedback. Yet, it is unknown how to design effective GenAI feedback to facilitate students’ progress in scientific practices, particularly at a primary age. In our study, we conceptualise a design framework for providing feedback for primary students to improve their drawings of scientific observations and models. Based on the design framework, we created a Science GenAi Feedback for Drawing (Sci-GiFD) application and conducted a mixed-method case study on 65 Year 3 primary students. The application provided feedback to students’ observational drawings of evaporation of salt water across six days, as well as drawing a side-view model of evaporation. We collected various forms of data, including students’ responses to pre- and post-drawing of scientific model, process data of 404 drawings and accuracy of GenAI in scoring drawings. Quantitatively, students’ improvement of drawing of topic-specific scientific model were associated with the application based on the design framework. Qualitatively, as supported by drawing artefacts and the associated feedback from two students, some students were able to act on GenAI feedback to revise their drawings of observation and scientific models. Practical implications related to pedagogical and technical design of GenAI-powered applications will be discussed.