TITLE:
From Support Tool to Learning Partner: A Systematic Review of GenAI Integration in University Science Labs
AUTHORS:
Michail Kalogiannakis, Nikolaos Papakonstantinou, Dimitrios Sotiropoulos
KEYWORDS:
ChatGPT, Generative Artificial Intelligence, Higher Education, Laboratory-Based Instruction, Science Education, STEM Education
JOURNAL NAME:
Creative Education,
Vol.16 No.9,
September
11,
2025
ABSTRACT: This Systematic Literature Review (SLR) investigates the integration of generative artificial intelligence (GenAI) tools, such as ChatGPT, into science education (SE) and STEM education at the higher education level, with a particular focus on experimental and laboratory-based learning environments. Adhering to the PRISMA protocol, 164 initial records were screened, and 20 empirical studies published between 2023 and 2025 were selected based on strict inclusion criteria: relevance to generative AI, application in SE or related STEM fields, involvement of students or pre-service teachers, integration into hands-on instruction, and empirical validation. This selection process ensured both conceptual relevance and methodological rigor. The thematic analysis was structured around five key domains: cognitive benefits, pedagogical practices, student attitudes, teacher perceptions, and broader interpretive convergences. The findings indicate that GenAI supports conceptual understanding, problem-solving, and the development of metacognitive strategies when implemented within well-designed pedagogical frameworks. While students generally held positive attitudes toward the use of GenAI, they also expressed concerns about the reliability and critical interpretation of AI-generated content. Educators recognized GenAI’s educational potential but underscored the need for guidance, scaffolding, and professional development. The review highlights two major interpretive tensions: the simultaneous presence of acceptance and uncertainty among learners, and the dual role of GenAI as either a metacognitive facilitator or a superficial answer generator, depending on its use context. Despite methodological limitations related to sample size and study duration, the selected studies offer valuable insights into the responsible and effective use of GenAI in science education, informing both future research and instructional design. It also emphasizes the need to align GenAI integration with evidence-based teaching strategies to create meaningful and equitable learning experiences in science education and in broader STEM contexts.