Enhancing Python Programming Through ChatGPT (GPT-5) Study Mode as a Learning Tutor: A Study on Learning Achievement, Motivation, and Self-Regulated Learning in K-12 Serious Game (CodeCombat) Programming Education

This study examined a three-week intervention in a Grade 5–6 Python course that combined CodeCombat with OpenAI GPT-5 Study Mode. The Study Mode included the “Study and Learn” interface, which provided step-by-step Socratic guidance rather than full solutions. Twenty-five novice learners in an elementary school used Study Mode in class and at home. Data came from semi-structured interviews and were analyzed using grounded theory. For RQ1 (Learning Achievement), ChatGPT’s Study Mode supported a short try, feedback, and revise cycle in which hint-level prompts helped students identify error sources such as variable name mismatches and loop stop conditions. This guidance shifted them from unguided trial and error to stepwise planning with small tests. Struggling learners advanced through minimum viable goals and brief checklists, while faster learners, prompted by Study Mode suggestions, modularized patterns such as basic syntax, loops, and variables, then transferred them to the multiplayer arena. For RQ2 (Learning Motivation), Study Mode’s on-demand guidance reduced waiting time and frustration, which kept students engaged in class and motivated them to continue practice at home. The chatbot’s calm and supportive tone helped students manage emotions during errors or time pressure, and the experience of solving problems with guided hints strengthened their self-efficacy by allowing them to credit success to their own effort. For RQ3 (Self-Regulated Learning), the interaction rules of Study Mode, which withheld direct answers and required students to provide context before receiving hints, prompted them to set near-term goals, perform quick self-checks, and monitor progress. Repeated prompts about naming, loop stops, and variable resets evolved into internalized pre-flight checklists. In the arena, Study Mode mediated team discussions, aligned vocabulary, and encouraged rapid strategy adjustments in response to leaderboard feedback. By the third week, several students had developed structured question scripts containing context, current behavior, and a request for hints only, demonstrating early self-regulation routines. Overall, GPT-5 Study and Learn acted as a hint-first coach that maintained momentum, deepened reasoning about code, and supported the development of independent problem-solving while preserving student responsibility for solution development.