TITLE:
AI-DRS: A Dialogic Generative-AI Instructional Model for Scientific Reasoning in Primary Education
AUTHORS:
Michail Kalogiannakis, Theodoros Spasopoulos, Nikos Papakonstantinou, Apostolos Xenakis
KEYWORDS:
Generative Artificial Intelligence, Micro:bit, Claim-Evidence-Reasoning, Scientific Reasoning, Primary Education
JOURNAL NAME:
Creative Education,
Vol.17 No.8,
August
25,
2026
ABSTRACT: Cultivating scientific reasoning in primary education involves not just making measurements, but also turning them into evidence and explanations. Physical-computing activities often do not support this process, and using generative artificial intelligence (GenAI) as a tool for answers can skip this step. This article suggests a teaching method for primary education that limits GenAI’s role so it can support pupils’ reasoning and then step back. In our research-and-development design, we created the AI-DRS (AI-supported Dialogic Reasoning Scaffolding) Physical Inquiry Model. This model pairs a Nezha-micro:bit soil-moisture investigation with sensor data generated by pupils. It follows a Claim-Evidence-Reasoning framework and includes a dialogic control layer based on three principles: grounding in data, ensuring proper epistemic oversight, and being responsive to pupils. The model incorporates fading based on established criteria, teacher-led interventions, and specific protections for each child. The article specifies the scientific target of the investigation, the sensor-calibration and data-quality rules, the assignment and balancing of inquiry variables, the technical configuration and rule precedence of the system, a teacher escalation protocol with explicit stop rules, and a cluster-aware analysis plan. Next steps include expert validation and a controlled pilot study. The model has a seven-phase process, an architecture that responds to pupil input, a scoring system, and a coding scheme that is sensitive to how pupils engage. It separates the tool’s dialogic moves from pupil responses, allowing for individual analysis. This proposal is theoretical, with no empirical results presented. It establishes a structured approach where GenAI can ask follow-up questions while the teacher maintains control over the knowledge. Validating the model’s effectiveness and understanding how it works is the essential next step.