2025 ASEE Annual Conference & Exposition

With the release of and widespread availability of generative AI interfaces, higher education is starting to experiment with new learning paradigms around the incorporation of these tools and exploring the dynamic effect they can have on student experiences. This paper presents a quadrant student classification and characterization model based on student engagement and perceptions of generative AI in a first-year “Introduction to Engineering Design” course at a small private university.

Within a semester-long first-year engineering course, themed around using robotics to solve a variety of engineering challenges, a custom-built chatbot platform consisting of five AI tools was created and provided to the students for use throughout the semester. The 30 students in the class (21 of whom consented to research and whose specific interactions are studied here) leveraged this resource as needed to accomplish any of the semester tasks assigned: coding robotic systems, designing and building mechanical solutions, creating digital project documentation, and looking up class-specific logistical information.

For instance, the use of a primed chatbot knowledgeable in the format, syntax, and operation of the robot-specific Python functions, reduced the need for dedicated programming lectures in the introductory course. Independent of programming background, the chatbot enabled students to develop, test, and iterate on code solutions via natural language based conversation with the generative AI system. However, in reality different students leveraged the system in unique ways, varying in duration, intensity, and level of engagement with the chatbot. In this research study, students' variety of interactions are analyzed to create a classification and characterization model that offers a framework for understanding the diverse methods, styles, and levels of intensity in which students interact with AI.

The model introduced here, and backed by student use data from throughout the semester, is based on two primary axes of classification: (1) students’ prior knowledge and experience in coding and robotics (ranging minimal to abundant), and (2) students’ willingness and perception of engaging with generative AI as a learning tool (ranging low to high). The qualitative data used to develop this model was gathered through tracked interactions by students throughout the semester (resulting in over 1,000 different chatbot conversations) and paired with comprehensive pre/post student survey responses to both Likert and open-ended response questions around their technical backgrounds, personal perceptions of generative AI, and self-reported use of generative AI. By analyzing the concrete level of engagement with AI tools as collected by the system along with prior experience, the quadrant model developed is able to identify and categorize students into four distinct types: (a) Overwhelmed Learners, (b) Pioneering Learners, (c) Engaged Learners, and (d) Self-Reliant Learners.

This paper introduces these four new different learner profiles, defines them in the context of using generative AI for solving engineering problems, and highlights the varied and complex ways students interact with generative AI over the duration of a semester. This classification framework offers insights into the diverse ways students navigate AI-assisted learning environments, revealing key trends in learner engagement, learner difficulties (and how they overcame), and learner growth and development throughout the course. The framework encourages educators to reflect on how AI can be further tailored to specifically meet the needs of diverse individual learners, suggesting future developments in generative AI that can further enhance or reshape educational experiences. However, it also underscores the need for careful consideration of student readiness, comfort levels, and the design of AI tools to maximize their effectiveness in the classroom, especially in light of cautious and resistant learners. Finally, this work contributes to the broader conversation on the role of personalized AI learning and pedagogical strategies that influence overall learning efficacy.