Tích hợp Vibe Coding vào giáo dục STEM: từ cú pháp lập trình đến ngữ nghĩa bài toán

Tóm tắt

The advancement of large language models has given rise to Vibe Coding, a programming approach that lets users generate source code from descriptions of their intent rather than from syntax. In STEM education in Vietnam, the syntax barrier still prevents many students from engaging in simulation and computational activities, raising the question of whether Vibe Coding can lower this barrier without weakening computational thinking. This study proposes a five-phase instructional model integrating Vibe Coding into STEM teaching, in which an algorithm-design step using flowcharts or pseudocode is a mandatory phase before using the language model, aiming to maintain opportunities for students to develop the components of computational thinking. The model is illustrated through three scenarios whose technical feasibility was verified by the research team (not yet pedagogically tested in classrooms): elastic collision simulation (Physics, Grade 10), Monte Carlo area estimation (Mathematics, Grade 12), and automatic text classification (Informatics, Grade 10). The findings indicate that Vibe Coding can support STEM teaching when organized within a rigorous pedagogical process guided by the teacher. The study identifies limitations including the code-verification paradox, the risk of diminished algorithmic thinking, model hallucination, and academic-integrity challenges, and proposes a four-component teacher competency framework and a process-based assessment strategy.