This dataset comprises of multimodal data collected from 118 Finnish secondary school students (aged 12–14 years) who participated in a five-week collaborative inquiry-based learning intervention in physics focused on thermodynamics. Students were divided into 32 small groups (2–4 members each) formed based on their self-reported self-regulated learning (SRL) profiles. Group formation aimed to ensure heterogeneity in SRL skills, which were assessed using validated scales covering cognitive regulation, metacognitive strategies and regulation, motivation regulation, and SRL self-concept (Karlen et al., 2021, 2024; Hirt et al., 2021). The collection of this dataset is funded as part of the Research Council of Finland (project 355776) Learning Regulation with AI – Promoting Adaptive K-12 Learners (LEAD) (P.I. Sanna Järvelä). The logistics of the data collection was carried out with the support of LeaF Research Infrastructure, University of Oulu, Finland. The study was conducted in authentic classroom environments during regular science lessons. Before the data was collected, students participated in an introductory session, which included a short video explaining the role of the Metacognitive Artificial Intelligence (MAI) system, how to complete the self-report questionnaires, and a trial collaborative task. The goal was to familiarise students with the study procedures, tools, and data collection infrastructure (e.g., video/audio recording setup, researcher presence). Over the five weeks, students completed weekly 90-minute lessons consisting of guided collaborative inquiry activities supported by MAI. These activities included warm-up questions, hands-on experiments with physical apparatus, teacher demonstrations, and computer simulations. Students worked collaboratively to document their hypotheses, observations, and reasoning in shared worksheets. Teachers facilitated the sessions using scaffolding strategies aligned with SRL principles, promoting independent problem-solving and reflection. All tasks and instructions were standardized across the five weeks of lessons. Data collection was multimodal and included classroom video recordings, group-level audio recordings, and self-reported data. An adapted domain knowledge test on thermodynamics (Vidak et al., 2019; Yeo & Zadnik, 2001) was administered before and after the intervention, covering key concepts such as heat transfer, thermal expansion, and states of matter. SRL skills were measured using four previously validated instruments. The resulting dataset includes group-level video and audio recordings, pre-and post-tests of domain knowledge, SRL self-reports, and student-generated learning artefacts (e.g., completed worksheets). These data allow for in-depth analysis of collaborative inquiry processes, the role of AI-based scaffolding in supporting SRL, and students’ conceptual development in thermodynamics within an ecologically valid school context.

2025