Individual Submission Summary
Share...

Direct link:

Download

Poster #45 - Not All Students Use AI Equally: A Multidimensional AI Usage Divide Among College Students

Friday, November 6, 5:00 to 6:30pm, Property: Boston Marriott Copley Place, Floor: 4th Floor, Room: Grand Ballroom

Abstract

Background and Purpose
Generative artificial intelligence (GenAI) is increasingly embedded in college students’ academic work and daily lives, yet students differ in how and why they use these tools. Emerging research suggests substantial variation in AI use across tasks and purposes, pointing to potential inequalities in AI engagement beyond access to technology. However, existing studies have often focused on specific academic tasks or single dimensions of AI use, providing limited insight into how academic and everyday uses come together within the same student. Little is also known about whether distinct patterns of AI use are systematically associated with sociodemographic, academic, geographic, and anxiety-related characteristics. To address these gaps, this study seeks to identify patterns of GenAI use among South Korean college students and examine factors associated with class membership.

Methods 
Survey data were collected from 207 college students in Korea (34.8% male, 65.2% female) enrolled in nursing (32.4%), health administration (28.5%), and social work (39.1%) programs. Latent class analysis (LCA) was conducted to identify patterns across 18 AI-use activities, with model fit indices and entropy examined to determine the optimal number of latent classes. Multinomial logistic regression analysis was subsequently conducted to examine factors associated with class membership.

Results
Three classes were identified: Everyday Life-Integrated AI Users (24.6%), Academic and Information-Seeking AI Users (30.0%), and Passive AI Users (45.4%). Passive AI Users and social work majors served as the reference categories. Metropolitan residence (B= .822, p < .05) and nursing major (B=1.305, p < .01) were associated with greater odds of membership in the Everyday Life-Integrated class. Metropolitan residence (B=1.226, p < .01) and elevated anxiety symptoms (B=1.579, p < .05) were associated with greater odds of membership in the Academic and Information-Seeking class, whereas health administration students (B= -1.052, p < .05) had lower odds of membership in this class than social work students.

Conclusions and Implications
Students differed in both the breadth and purpose of their GenAI use, and these patterns varied by geographic context, academic discipline, and anxiety symptoms. These findings suggest an emerging AI usage divide among college students and indicate that structural, institutional, and psychological factors may be relevant to how they use GenAI, beyond individual preferences and skills. These findings highlight the potential value of tailoring AI support to students’ patterns of use rather than relying on a uniform approach. Priorities may include foundational AI literacy for students with limited engagement, discipline-specific guidance for academic applications, and equitable access to AI training and resources beyond metropolitan areas. Student-support services may also benefit from recognizing that psychological factors are associated with patterns of academic and information-seeking GenAI use.

Authors