Transforming Student Learning with Generative AI: Insights from a Constructivist Perspective

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

Abstract

AI has gained significant attention for its potential to make authentic learning experiences more accessible, equipping learners to become collaborative problem solvers prepared to address national and global challenges. While there is widespread interest in exploring AI applications in learning, there remains a lack of scholarly exploration on integrating research-based theory with AI, particularly from a constructivist perspective (Vygotsky, 1978). Constructivist learning theory emphasizes that knowledge construction processes and cognitive activities in learning environments should mirror the ways in which experts engage in real-world settings (Jonassen, 1994). To effectively transform educational practices and achieve comprehensive goals, AI should align with established learning theories, with consideration of how AI can bridge the gap between learning theories and practices in education.


Based on constructivist learning theories, a key goal of AI is to provide personalized learning support tailored to individual learner characteristics. AI has been applied to enhance personalized learning by augmenting human intelligence to improve adaptation and precision, incorporating intelligent feedback, agents (e.g., tutors, robots, chatbots), and adaptive learning systems. However, despite the strong belief in constructivism in AI-empowered personalized learning, current AI applications in K-12 school systems are still lagging in fully implementing theory-based learning practices.


This paper aims to explore 1) how integrating Generative AI through the lens of constructivist learning theories can support meaningful learning experiences, and 2) the importance of aligning AI interventions and associated research with these established theories to promote student learning. The approach focuses on Generative AI applications that empower learners to collaborate with AI as a partner in finding solutions to the challenges they face. Two research-driven examples demonstrate how Generative AI can be used to develop dynamic, interactive learning materials and provide real-time feedback, enhancing constructivist teaching and learning practices to accommodate a wide range of learner backgrounds.


The first example is designed for scoring and providing feedback on learners’ responses to formative assessment tasks. It employs various prompt-engineering approaches such as chain-of-thought prompts (Wei, et al., 2022). Chain-of-Thought prompting encourages the AI model to generate explanations that lead to the final scoring. These approaches generate explanations for its scoring decisions, and if learners disagree with their scores and request an explanation, the AI identifies specific statements in their responses to justify its scoring. This process encourages learners to refine their responses iteratively until they are satisfied with their performance. The second example involves developing project-based learning materials (Authors) based on constructivist learning theories using CustomGPT (https://customgpt.ai). Learners interact with a chatbot agent to modify personalized materials tailored to their specific needs. Through these interactions, both the learners and the AI learn from each other, enhancing the personalized learning material to better meet their specific needs.


Research employing generative AI-empowered personalized learning offers the potential to enhance student learning and cognitive engagement by addressing educational inequity and providing meaningful learning opportunities. To realize this potential, AI applications need to be researched and developed based on constructivist learning theories, and iteratively reviewed to ensure alignment with the theories.

Author

• Namsoo Shin, Michigan State University