Improving students’ scientific argumentation through AI-supported feedback

Abstract

Scientific argumentation is a core practice in science education, yet many students struggle to construct arguments that effectively integrate claims, evidence, and reasoning. With the growing use of artificial intelligence (AI) in education, AI-supported feedback has emerged as a potential tool to scaffold students’ argumentation processes. This study examined the effects of AI-supported feedback on students’ scientific argumentation using a quasi-experimental, explanatory sequential mixed-methods design. Two intact groups participated: an experimental group receiving AI-supported formative feedback on written arguments and a control group receiving conventional teacher feedback. Quantitative data were collected using a validated rubric based on the claim–evidence–reasoning (CER) framework and Toulmin’s argument pattern (TAP), while qualitative data from student interviews and written responses provided contextual insights. Results showed that the experimental group achieved greater improvements in overall argumentation quality, particularly in evidence use and reasoning. Qualitative findings further indicated that AI feedback supported iterative revision and strengthened students’ understanding of evidence–claim relationships.