WIP: Structured AI Tutoring in Engineering Education

Abstract

This innovative practice Work-in-Progress paper presents a structured approach to integrating AI tutors into undergraduate engineering education at a Hispanic-Serving Institution. The study focuses on two upper-division computer architecture courses and explores how large language models (LLMs), such as ChatGPT-4o, can be scaffolded to support student learning. Our approach emphasizes guided AI interaction rather than direct answer generation. It is grounded in Vygotsky's Zone of Proximal Development and implemented through the ADDIE instructional design model. Two scaffolding protocols—AI Think-Aloud and AI-Driven Pair Programming—are proposed, that foster metacognitive engagement and collaborative problem-solving. The paper highlights the need for differentiated instruction aligned with Universal Design for Learning (UDL) to assist individuals who lack skills necessary to take advantage of LLMs. Preliminary findings suggest that structured AI integration can enhance learning outcomes and engagement, while minimizing risks associated with unregulated AI use. Evaluation methods include SOLO taxonomy analysis of student-AI interactions and short-term longitudinal survey data on student attitudes. The work contributes to emerging frameworks for ethical, inclusive, and pedagogically grounded AI adoption in engineering classrooms.

Ähnliche Arbeiten