The Potential of Using AI to Improve Student Learning in STEM

• In Event: Generative AI in STEM Teaching and Learning: Informing Transformative Uses Through Ethical Research and Development

Abstract

Understanding AI's role, impact, and potential in education is crucial for educators, policymakers, and researchers. This paper overviews how AI-driven research and development can enhance student learning, address challenges, and explores future opportunities. By examining past work and several recently NSF-funded projects, this paper highlights AI's potential to revolutionize STEM learning. It explores how AI can support learners in developing an integrated understanding of STEM, including three-dimensional learning consistent with A Framework for K-12 Science Education (National Research Council NRC, 2012), while enhancing learner engagement and interest. These projects provide insights into AI in STEM education's current state and challenges, and future possibilities.


Learning is most effective when it involves making connections among ideas, facilitating knowledge transfer to new situations. This process, known as "integrated understanding," contrasts with "inert knowledge," where students memorize facts without applying them to new contexts (Pellegrino & Hilton, 2012; Spiro et al., 2019). There is potential for AI to serve as a learning partner for students, supporting the development of integrated understanding in the classroom. By acting as a learning partner, AI can provide cognitive, motivational, metacognitive, and collaborative support, thus enhancing student learning outcomes. We identify relevant cognitive theories (e.g., situated cognition), non-cognitive motivational and social constructs, and collaboration, which are used to ground AI's role in education. We then describe how these theoretical frames lead to various functions that AI can perform as a learning partner and in support of student learning. In previous decades, R&D work has often focused on providing a specific type of support for students within AI-backed systems. However, generative AI opens the possibility of providing multiple support types, even individualized for learners, within a single system.


A major research goal of utilizing AI in education is the development of adaptive learning systems, which are manifested in a range of various applications, including personalized feedback, adaptive learning paths, intelligent agents (such as tutors and chatbots) and virtual and augmented reality. These applications create personalized learning environments, ensuring that each student receives tailored guidance or environments to meet their unique needs. Looking ahead, R&D utilizing generative AI holds the potential to revolutionize these adaptive systems by creating content rather than merely analyzing or classifying existing data. Generative AI can produce text and audio-visual data offering truly personalized learning experiences and automated content generation. This capability enables generative AI to adapt to individual student needs and achievements dynamically, for example creating learning materials or offering supportive comments interactively. These adaptive capabilities allow each learner to receive the support to succeed, making AI an important partner in future STEM education.


However, research is necessary to determine the effectiveness of generative AI in promoting integrated understanding and ensuring that the applications are grounded in learning theory. Addressing cultural and socioeconomic disparities is crucial to prevent widening the educational gap, ensuring equal access to AI, and supporting the goal of AI for all students. Nevertheless, the future of AI in STEM education can be viewed as promising, with the potential to revolutionize student learning experiences and outcomes.

Author

• Joseph S. Krajcik, Michigan State University