Code Comprehension Problems in Introductory Programming to Overcome ChatGPT

The prevalence of Generative AI (GenAI) technologies like ChatGPT necessitates a transformation in introductory programming education. This change shifts the focus to developing critical thinking among students instead of just learning a programming language. Therefore, this study introduces the Code Comprehension Problem (CCP) approach, incorporating value tracing, code correction, and modification exercises to enhance critical thinking and reduce dependence on ChatGPT. This research also assessed the performance of GPT-3.5 and GPT-4 on CCPs, comparing the effectiveness of full-task descriptions versus individual subquestions as prompts. Our analysis revealed that while ChatGPT can address CCPs with variable accuracy, presenting subquestions individually tends to elicit more precise responses. However, we argue that without more advanced prompting techniques, the answers generated by ChatGPT are not suitable for beginners due to the tendency of these models to use more advanced syntax and objects. These findings underscore the need for instructional strategies that not only leverage AI’s potential but also cultivate students’ autonomous problem-solving abilities, preparing them for a future where AI is an integral part of the learning and problem-solving process.