Modelos de lenguaje para la generación de preguntas de programación con diferentes niveles de dificultad

Abstract

Introduction: This study explores the potential of Large Language Models (LLMs), specifically ChatGPT-4, in generating Python programming questions with varying degrees of difficulty. This ability could significantly enhance adaptive educational applications. Methodology: Experiments were conducted with ChatGPT-4 and participants to evaluate its ability to generate questions on various topics and difficulty levels in programming. Results: The results reveal a moderate positive correlation between the difficulty ratings assigned by ChatGPT-4 and the perceived difficulty ratings given by participants. ChatGPT-4 proves to be effective in generating questions that cover a wide range of difficulty levels.Discussion: The study highlights ChatGPT-4’s potential for use in adaptive educational applications that accommodate different learning competencies and needs. Conclusions: This study presents a prototype of a gamified educational application for teaching Python, which uses ChatGPT to automatically generate questions of varying difficulty levels. Future studies should conduct more exhaustive experiments, explore other programming languages, and address more complex programming concepts.