WIP: Guiding, Not Giving: Analyzing the Effectiveness of Guided Problem-Solving in CS1

Abstract

This Work-in-Progress research paper investigates the role of generative Artificial Intelligence (AI) in computer science education, focusing on guided problem-solving in introductory programming courses. As tools like ChatGPT become increasingly accessible, educators face the challenge of supporting student learning without encouraging over-reliance. This study explores an approach where AI -generated support is limited to providing structured hints rather than complete solutions. A custom-built application, Drona, integrates AI assistance into lab assignments by offering context-aware guidance without directly solving code or accessing student submissions. The research involves undergraduate and graduate students completing programming tasks on the application, followed by surveys to evaluate learning outcomes and engagement. Student learning is measured through task completion, while engagement is analyzed through interaction patterns and self-reported feedback. Data collection is complete, and initial findings from pilot sessions suggest that guided AI can reduce cognitive load and increase persistence, particularly for less experienced students. However, usage patterns vary by experience level, indicating a need for more adaptive feedback. The study aims to inform responsible AI integration in CS education by identifying how such tools can promote critical thinking without substituting for it. While the findings are specific to the Drona application and ChatGPT-4, the research contributes early insights into designing AI -enhanced educational environments that maintain academic integrity and support diverse learners. Future work will explore refining hint mechanisms and extending the study to additional instructional contexts.

Ähnliche Arbeiten