Abstract:

Assessing SQL exercises in large university courses poses significant challenges due to the diversity of valid solutions, the complexity of handwritten responses, and the variability in how instructors penalize errors. These factors often lead to inconsistencies and subjective grading, particularly when hundreds of exams must be corrected in limited time.

This paper presents EvIS (Intelligent SQL Assessment - Evaluación Inteligente de SQL), an educational innovation project that explores the use of generative artificial intelligence to automate the correction of SQL queries. The system integrates Optical Character Recognition (OCR) with Large Language Models (LLMs) such as ChatGPT to process handwritten SQL statements, identify potential errors, and propose consistent penalizations across student submissions.

The proposed approach aims to support teachers by reducing manual workload, ensuring greater fairness and transparency in grading, and providing students with faster and more informative feedback. Beyond automation, the project seeks to analyze how generative AI can enhance assessment practices in programming-related subjects and promote a more objective and scalable evaluation framework in higher education.

Keywords:

Generative artificial intelligence, sql, correction, fairness.