2025 ASEE Annual Conference & Exposition

This paper aims to identify current trends in the use of predictive AI and generative AI in education by examining their application across specific domains. Predictive AI is divided into three levels: degree, course, and individual. At the degree level, models predict broad academic outcomes, such as student retention and overall success, by analyzing institutional data. These insights can help universities design interventions to improve student retention rates. At the course level, predictive models focus on identifying at-risk students and recommending targeted interventions to boost their academic performance. At the individual level, AI offers personalized support, adapting to each student's unique needs, behavior, and progress to foster better learning outcomes.

Generative AI is discussed in terms of student-focused applications and teacher-focused applications. For students, generative AI enhances the learning experience by personalizing educational content, providing real-time feedback, and encouraging creativity and problem-solving. These tools allow students to engage more deeply with their learning materials while simultaneously building crucial skills for navigating an increasingly AI-driven world. For teachers, generative AI automates routine tasks, such as grading, content creation, and feedback, enabling educators to focus more on student interaction and personalized teaching methods.

The integration of predictive AI and generative AI not only delivers more personalized, data-driven learning experiences but also improves administrative efficiency in the classroom. These technologies have the potential to revolutionize how educators approach teaching and how students engage with their learning materials. This paper further analyzes the challenges, ethical considerations, and future directions for implementing predictive and generative AI technologies, taking into account their impact on teaching practices and student outcomes across diverse academic environments.

This work is funded by NSF #2136600.