Structured AI-tutoring for Computer Architecture Courses

Abstract

This paper presents a structured approach to integrating AI tutors into undergraduate computer architecture courses at a medium-sized Hispanic-Serving Institution. We propose a system in which AI functions as a scaffolded learning partner designed to promote articulation, reflection, and independent reasoning, rather than as an answer-generating tool. The approach relies on instructor-authored system prompts and controlled interaction patterns to regulate AI behavior, grounded in Vygotsky’s Zone of Proximal Development and implemented through the ADDIE instructional design framework. An assessment structured using an AI-driven pair programming protocol was deployed across multiple course sections, with evaluation based on attitudinal survey data analyzed using ordinal logistic regression. Results indicate that students increasingly perceived the AI tutor as supporting critical thinking, problem solving, and effective feedback over time. While participation attrition limits causal claims, the findings suggest that pedagogically grounded, instructor-mediated AI integration can enhance student engagement while mitigating risks associated with unstructured AI use in engineering education.