Abstract
Abstract
In science education, students engage in scientific argumentation to develop evidence-based reasoning and understanding of scientific knowledge. Recent research highlights the potential of generative AI, including Socratic AI, to deepen our understanding of how students develop reasoning during inquiry-based learning, including the development of scientific argumentation and epistemic beliefs. However, these efforts are still in the early stages. This study aims to advance this research by analysing students’ reasoning across Socratic AI-powered Argument-Driven Inquiry (ADI), traditional ADI, and control conditions. Data were collected from 90 secondary students during a randomized classroom intervention designed to foster scientific argumentation and reasoning. Students participated in activities involving argument construction, evidence evaluation, and revision across the three conditions. Students’ written responses were coded for scientific argumentation and epistemic beliefs and analyzed for weighted connections between coded components using Epistemic Network Analysis. The analysis of students’ reasoning revealed distinct patterns of weighted connections between codes and varying degrees of reasoning development across the three conditions. Additionally, the Socratic AI condition showed broader connections involving the evaluation of competing arguments and personally reasoned justification. The findings offer insights into students’ scientific reasoning during AI-supported argumentation and have implications for future research and practice.