Scaffolding Metacognition in Programming Education: Understanding Student-AI Interactions and Design Implications

TL;DR

The paper investigates how AI copilots influence metacognition in programming education by analyzing over 10,000 student–AI dialogues across three years, supplemented with student and educator surveys. Using a mixed-methods approach, it finds that students predominantly engage AI during Monitoring, with limited Planning and Evaluation, and that AI responses are generally correct but vary in usefulness, sometimes risking metacognitive offloading. The authors derive design principles for metacognition-supportive AI that balance guidance and autonomy, advocating structured prompts, indirect scaffolding, and integrated phase transitions to foster planning, monitoring, and reflective evaluation. These insights offer practical implications for developing educational AI tools that strengthen rather than bypass learners’ metacognitive engagement in computing education.

Abstract

Generative AI tools such as ChatGPT now provide novice programmers with unprecedented access to instant, personalized support. While this holds clear promise, their influence on students' metacognitive processes remains underexplored. Existing work has largely focused on correctness and usability, with limited attention to whether and how students' use of AI assistants supports or bypasses key metacognitive processes. This study addresses that gap by analyzing student-AI interactions through a metacognitive lens in university-level programming courses. We examined more than 10,000 dialogue logs collected over three years, complemented by surveys of students and educators. Our analysis focused on how prompts and responses aligned with metacognitive phases and strategies. Synthesizing these findings across data sources, we distill design considerations for AI-powered coding assistants that aim to support rather than supplant metacognitive engagement. Our findings provide guidance for developing educational AI tools that strengthen students' learning processes in programming education.

Figures (8)

• Figure 1: Metacognitive demand (learner) and support (AI) in the student–AI interaction loop.
• Figure 2: Metacognitive Phases of Student–AI Interaction in Programming Tasks
• Figure 3: Distribution and sequential patterns of student prompts across metacognitive phases and specific categories: (a) distribution of phases, (b) distribution of strategies, (c) sequential phase patterns, and (d) sequential strategy patterns.
• Figure 4: Sequences of prompt during problem solving. Each path shows a per-problem trajectory from Start to End. Columns are successive turns. Nodes are grouped as P (Planning), M (Monitoring), and E (Evaluation).
• Figure 5: Markov Chain Analysis of Phase Transitions.