The Challenge of Improving Critical and Scientific Thinking in Engineering Students Through RAG Systems Based on Validated Scientific Corpora

Developing critical and scientific thinking in engineering students remains a persistent challenge in higher education. Traditional teaching practices in mathematics and physics often prioritize procedural knowledge over analytical reasoning, leading to limited reflection, superficial understanding, and low levels of engagement. This research-in-progress doctoral study explores the potential of retrieval-augmented generation (RAG) systems-artificial intelligence models supported by validated scientific corpora-to enhance critical and scientific thinking in firstand second-year engineering students. Based on constructivist and deep learning theories, the study proposes a mixed-method quasi-experimental design that integrates technological mediation, ethical literacy, and active learning strategies. This paper presents the theoretical framework, research design, preliminary observations from pilot experiences, detailed evaluation instruments, strategies for monitoring AI biases and hallucinations, and expected contributions. The study aims to provide an empirical basis and ethical framework for the integration of generative AI in engineering education.