Strategy-Oriented Feedback for Fostering Systematic Problem-Solving in Machine Learning Education

Abstract

Enabling students to develop systematic problem-solving strategies is a central goal in computing education and of particular relevance in the emerging field of machine learning (ML) education. While exploratory approaches are common in ML learning tasks, fostering the development and persistence of structured problem-solving strategies remains challenging, as these demand considerable metacognitive regulation and persistence, causing learners to often revert to exploratory trial-and-error behavior. To address this challenge, we augmented a digital puzzle-based learning game for decision tree construction with an adaptive feedback module generating individualized messages based on the continuous evaluation of learners' problem-solving strategies. Building on an earlier baseline study, the present work investigates how this strategy-oriented feedback shapes students' problem-solving processes. For this purpose, screencast video data and gameplay logs (N=205, approx. 55 hours of gameplay footage) are used to enable fine-grained insights into learners' strategic behavior, its persistence, and transitions. The findings demonstrate how strategy-oriented feedback can support the development of structured problem-solving skills in decision tree construction and inform the design of ML learning environments that foster transferable competencies in secondary computing education.

This is the author's version of a paper accepted for publication at the 2026 Workshop in Primary and Secondary Computing Education Research (WiPSCE 2026). The final authenticated version will be published in the ACM International Conference Proceedings Series and will be available via the ACM Digital Library