The power duo: unleashing cognitive potential through human-AI synergy in STEM and non-STEM education

Abstract

AI-driven education tools are expected to impact over 2 billion learners worldwide in the coming years, transforming both STEM and non-STEM disciplines in unprecedented ways (Louly, 2024;Sandhu et al., 2024;World Economic Forum, 2024) Artificial Intelligence (AI) is revolutionizing education through personalized tutoring, real-time feedback, and adaptive learning experiences (Akavova et al., 2023). AI enables teachers to create individualized development plans according to the needs of the students. Its impact on intellectual tasks such as critical thinking, emotional intelligence, and moral reasoning is, however, a debatable topic (Çela et al., 2024). Greater dependence on AI-driven tools is a cause for concern with surface learning and minimal engagement with complex problem-solving and debates (Çela et al., 2024).While AI enhances education in all subjects, it does so unevenly between STEM and non-STEM fields, particularly in its engagement with structured logic-based learning versus interpretative, abstract reasoning (Nagaraj et al., 2023;Singer et al., 2023). Within STEM education, AI's analytical and structured logic nature provides excellent benefits in problem-solving, simulation, and automation of complex calculations. However, non-STEM fields, such as the humanities and social sciences, require more interpretative, ethical, and creative engagements that AI is less likely to be able to provide. This paper explores these differences while advocating for an even-keeled integration of AI that augments, rather than replaces, human teaching.Theoretical Perspectives on AI in Education: Learning theories offer a framework for understanding AI's application in education. In the context of Bloom's Taxonomy, AI can support lower-order thinking skills (knowledge, comprehension, application) in STEM education but lags in supporting higher-order skills like evaluation and synthesis (Essien et al., 2024). Within the context of Vygotsky's Zone of Proximal Development, for non-STEM, AI can be seen as a form of scaffold that can assist students with guided learning tasks but that still needs to be closely mediated by a human to facilitate the growth of abstract thought and creativity (Xue, 2023). Both theories emphasize the ability of AI to guide users in solvable problems however suggest the need for human intervention to achieve a deeper learning outcome.Intelligence in Education (AIEd) entails a broad range of varied tasks, ranging from adaptive learning systems to automated grading. AI, in STEM disciplines, is used to accelerate problemsolving and the grading process; for example, AI tools facilitate personalized tutoring to improve engagement and course performance in computational material, such as mathematics and engineering (Gupta et al., 2024;Mustafa, 2024). AI use in grading has reduced grading inconsistencies up to 44% compared to human grading (Gobrecht et al., 2024) which means less bias across STEM exams. Notwithstanding these prospects, however, there are apprehensions about bias and AI failure on tasks that entail subjective grading like grading for engineering ethics (Orchard & Radke, 2023).AI is also being utilized as a teaching assistant in STEM classrooms. Squirrel AI, an adaptive learning system in China, has been used to provide real-time feedback to students in mathematics and physics, significantly improving their problem-solving efficiency and retention rates (Luo, 2023). AI-powered teaching assistants like Jill Watson, developed at Georgia Tech, have demonstrated the ability to answer student queries efficiently in online courses, reducing educators' workload while increasing student engagement (Taneja et al., 2024). Such applications highlight AI's potential to serve as a valuable assistant in structured, logic-driven subjects.Beyond the classroom, AI is transforming STEM research via predictive analysis and automation. In biomedical engineering, for example, AI-powered