London, United Kingdom (Enmaeya News) — A recent study finds that the way students use generative AI, or GenAI, significantly affects their learning outcomes in higher education. Researchers examined how different approaches to reflecting on course material with AI tools influenced grades, applied knowledge, critical thinking, and learning autonomy.
The study identified four main approaches. Students taking a constructive approach used GenAI to create new understanding, while those taking an augmentative approach expanded existing knowledge. In contrast, a procedural approach involved following AI instructions without critical thinking, and a regurgitative approach relied on repeating AI-generated content.
Students who used AI constructively or augmentatively earned higher grades and demonstrated stronger learning skills. Those using procedural or regurgitative approaches scored lower and showed weaker applied knowledge, critical thinking, and autonomy.
Researchers said these patterns align with educational theories, including Vygotsky’s Zone of Proximal Development. Students who adopt a mastery goal structure—using AI to fill gaps and deepen understanding—move beyond their current abilities. Students with a performance goal structure, focusing on completing tasks without critical engagement, remain within their existing competence.
The study also draws connections to Bloom’s revised taxonomy and Fink’s taxonomy. Students who engaged constructively or augmentatively reached higher-order skills, including applying, analyzing, evaluating, and creating. Others remained at basic levels, focused on remembering and understanding.
Researchers suggest that educators can leverage AI in curriculum design to foster learning. Assignments can begin with simple AI tasks and progress to exercises requiring critical evaluation. Encouraging students to compare AI outputs with their own reasoning can strengthen logic, problem-solving, and critical thinking.
Limitations include the study’s focus on three marketing units, its exploratory nature, and use of ChatGPT-3. The research did not directly measure goal structures or establish causality. Future studies could use controlled, longitudinal designs to track how AI impacts learning over time and at different knowledge levels.
The findings highlight that GenAI is not a threat to learning when used correctly. Instead, its value depends on students’ approach, offering opportunities to enhance higher-order thinking, problem-solving, and independent learning in higher education.
