2026 ASEE Annual Conference & Exposition

Providing personalized academic support is an effective way to improve student learning and promote equity and inclusion. However, instructors at Primary Undergraduate Institutions (PUIs) and Hispanic-Serving Institutions (HSIs) often face large class sizes and heavy teaching loads that limit their ability to provide individualized support. Large language models (LLMs) have emerged as widely used tools for real-time learning assistance that can be a potential supplement for individualized support, but their use raises concerns related to accuracy, bias, academic integrity, and over-reliance, particularly for underrepresented students in engineering.
Missing from this discussion are student perspectives on whether and how LLMs can provide effective personalized instruction. This paper reports findings from an online survey and qualitative interviews with undergraduate engineering students at an urban PUI and HSI, to understand student experiences with AI tools, identify student needs, and generate actionable guidance to inform the iterative design and implementation of student-centered AI tools for personalized learning. Interview results show that while students regularly use LLMs for support with their coursework, they are also regularly dissatisfied with the results. Students described significant frustrations, including receiving responses that were misaligned with course content, insufficiently scaffolded explanations, and uncertainty about academic integrity. Students also expressed concerns about over-reliance and trust, and a desire for instructor-sanctioned AI tools.
Importantly, students articulated specific features they believed would better support their learning, including course-aligned explanations, guided problem-solving feedback, support for study skills and time management, and content drawn exclusively from instructor-provided materials. Their perspectives point to the need for the development of AI teaching tools that are aligned with learning goals and designed for specific courses. By centering student perspectives from an HSI/PUI context, this study contributes empirical insight into the limitations of current AI tools and offers actionable guidance for the future design of equity-oriented, student-centered AI learning supports in engineering education.