Reflective Dialogue or Prompt Refinement? Effects of Tutor Scaffolding on Students' Independent LLM Use for Programming
2026-07-03 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors studied two types of AI tutors that help students learn using large language models (LLMs). One tutor used questions to guide thinking (Socratic-Guidance), while the other helped students make better prompts (Prompt-Refinement). Both tutors led to similar results during use, but the Socratic-Guidance tutor helped students learn more deeply and use better prompting strategies later on. Even though students found the Socratic method slower, it improved their ability to learn with LLMs over time. This suggests that guiding students to think is important when designing AI tutors.
Large Language ModelsSocratic MethodPrompt EngineeringDialogic QuestioningLearning GainsPrompting StrategiesEducational TechnologyHuman-Computer InteractionGraduate-Level EducationMobile Robotics
Authors
Jerome Brender, Laila El-Hamamsy, Kim Uittenhove, Aitor Perez, Patrick Jermann, Francesco Mondada, Engin Bumbacher
Abstract
While Large Language Models (LLMs) can provide personalized support in learning, several studies have raised concerns regarding their use in education. Importantly, learning depends on how students engage with LLMs. This study examined how two types of LLM-based tutors shape students' prompting practices, learning, and subsequent LLM-use: a Socratic-Guidance (SG) tutor, which structures interaction through dialogic questioning, and a Prompt-Refinement (PR) tutor that guides the formulation of effective prompts. We conducted a two-phase study in a graduate-level mobile robotics course: 66 students used either the SG or PR tutor during a 6-week intervention, followed by 52 students using an unconstrained LLM during a 3-week course project. Results show that while the SG- and PR tutors led to similar task performance and prompting patterns during guided use, they differ in learning outcomes and later LLM-use. SG-students, relative to PR-student, achieved higher learning gains in later sessions, and were more likely to adopt understanding-driven prompting strategies, which are predictive of higher understanding, when using an unconstrained LLM. Although learners perceived the SG tutor as less efficient, the findings suggest that Socratic guidance supports the development of students' capacity to learn with LLMs over time, highlighting its importance for LLM tutor design.