How STEM students think: an AI-powered systematic review of thinking skills in STEM higher education

Abstract In STEM education, the development of thinking skills is increasingly viewed as essential for preparing students to navigate complex and uncertain futures. However, the field remains conceptually fragmented, with diverse and sometimes overlapping terminologies, definitions, and theoretical models. This study presents an LLM-powered systematic literature review, following the ENTREQ framework to ensure methodological transparency and rigor, aimed at identifying and mapping the thinking skills most frequently discussed in STEM higher education over the past decade. Using the AI tool Elicit, 170 peer-reviewed journal articles were analyzed. The review synthesizes definitions, conceptual frameworks, and interrelations among skills such as Critical, Systems, Computational, Mathematical, Spatial, Creative, Design, Reflective, Scientific, and Complexity Thinking. The results reveal eleven core thinking skills and highlight the conceptual overlaps and distinctions between them. AI use not only expedited literature screening and coding but also mitigated selection bias by systematically applying inclusion criteria, overcoming common limitations of manual reviews such as inconsistency and time lag. This overview contributes to greater clarity and consistency in the field and provides a foundation for future curriculum design and research, supporting a more coherent integration of thinking skills in STEM education.