Integrating generative AI into STEM education: enhancing conceptual understanding, addressing misconceptions, and assessing student acceptance

Abstract

Abstract Advancements in artificial intelligence (AI), particularly generative AI models such as ChatGPT, offer transformative opportunities to enhance educational practices in STEM disciplines. Thermodynamics, a fundamental subject in engineering education, presents significant challenges due to its abstract nature and common misconceptions. This study investigates the effectiveness of integrating ChatGPT as a supplemental pedagogical tool, guided by a constructivist inquiry-based approach using the Constructivist Inquiry-Based Learning Prompting (CILP) framework, to enhance conceptual understanding and address misconceptions in an introductory thermodynamics course for first-year Moroccan engineering students. A quasi-experimental design was used, with 120 students equally divided into control and experimental groups. The control group received traditional instruction, whereas the experimental group received ChatGPT-assisted instruction. Conceptual understanding was measured using pre- and post-tests, while student perceptions and acceptance were collected via weekly surveys. Results showed that the experimental group significantly outperformed the control group, exhibiting greater improvements in conceptual understanding and a reduction in qualitative misconceptions, particularly related to entropy and internal energy. However, some quantitative misconceptions persisted, underscoring ChatGPT’s limitations in advanced reasoning tasks, problem-solving, and numerical calculations. Students reported high satisfaction with ChatGPT’s usability and instructional support. Moreover, targeted use of ChatGPT, rather than frequent reliance, correlated with optimal learning outcomes. These findings underscore ChatGPT’s potential to enhance STEM education within inquiry-based, constructivist learning environments and provide evidence for the effective integration of generative AI tools to improve learning outcomes, particularly in resource-constrained settings.

Ähnliche Arbeiten