Abstract:

The growing adoption of Generative Artificial Intelligences (GAIs) has significantly impacted Computer Science education, expanding possibilities to support learning through adaptive explanations, content generation, and automated feedback. In the context of Database Systems education, persistent challenges arise from the abstract nature of topics such as Entity-Relationship modeling, SQL queries, functional dependencies, and normalization, highlighting limitations of traditional teaching approaches. Despite the potential of GAIs, empirical evidence regarding their effectiveness in this domain remains limited, particularly concerning student performance, cognitive effort, and acceptance of these technologies as educational support tools.

This study investigates the use of generative AI models, specifically ChatGPT and Google Gemini, as teaching assistants capable of supporting students in solving tasks and understanding fundamental concepts. A controlled experiment was conducted with 24 Computer Engineering students, organized into three stages (ER Modeling, SQL Queries, and Normalization), with participants divided into a control group (without AI support) and an experimental group (with GAI support). The evaluation combined quantitative metrics, based on task accuracy rates, and qualitative measures grounded in the TAM (Technology Acceptance Model), UTAUT (Unified Theory of Acceptance and Use of Technology), and NASA-TLX (NASA Task Load Index) models, including perceived usefulness, ease of use, satisfaction, intention to use, and mental effort.

The quantitative results indicated no statistically significant difference between groups in terms of performance, although the experimental group showed signs of improvement over the course of the activities. In contrast, the qualitative findings revealed high acceptance of the technologies, with 80.9% of participants expressing intention for continued use and recognizing their usefulness in the learning process. However, a high level of mental effort associated with the use of GAIs was identified, along with limitations related to error frequency and the need for critical interpretation of generated responses. It was also observed that, although interaction with the tools was considered relatively easy, challenges remain regarding perceived reliability and effective use.

As contributions, this study provides empirical evidence on the integration of GAIs into Database Systems education and makes available a public repository containing materials and experimental protocols to support replication. The findings suggest that GAIs can act as effective teaching assistants, fostering engagement and supporting learning, provided they are accompanied by strategies that promote prompt engineering skills and critical thinking. In terms of impact, this work supports the informed adoption of these technologies in Computer Science education, highlighting the need for pedagogical approaches that minimize cognitive effort, prevent overreliance, and encourage reflective and ethical use of such tools.

Keywords:

Database systems education, Generative AI, empirical study.