Profile analysis of college students' AI literacy and cognitive flexibility and its correlation with deep learning ability.

Introduction: AI literacy and cognitive flexibility are essential core capabilities for college students in the age of intelligent technologies. Nevertheless, the heterogeneous group characteristics of college students in these two competencies, as well as their connections with deep learning, have not been sufficiently investigated. This study aimed to classify college students into distinct subgroups according to their levels of AI literacy and cognitive flexibility, explore relevant influencing factors of subgroup categorization, and further clarify differences in deep learning performance across different subgroups. Methods: A cross-sectional research design was adopted. A total of 591 college students recruited from two universities in mainland China participated in this questionnaire survey. Latent profile analysis (LPA) was used to categorize participants based on AI literacy and cognitive flexibility levels. Multiple logistic regression analysis was applied to analyze demographic and behavioral factors correlated with subgroup belonging. The Bolck-Croon-Hagenaars (BCH) approach was utilized to compare deep learning scores among all identified profiles. Results: Three distinct latent profiles were extracted: the Dually Constrained group (30.5%), the Latent Potential group (31.3%), and the Comprehensively Integrated group (38.2%). Urban residential background, natural science major enrollment, satisfactory academic performance, and frequent AI application were positive predictors of membership in the Comprehensively Integrated profile. In contrast, rural residency, humanities majors, poor academic achievement, and infrequent AI use were significantly correlated with classification into the Dually Constrained group. Statistically significant upward trends in deep learning scores were observed from the Dually Constrained group to the Comprehensively Integrated group. Discussion: Obvious heterogeneity exists among college students regarding AI literacy and cognitive flexibility. Students with insufficient AI exposure and unsatisfactory academic performance are more likely to fall into the disadvantaged Dually Constrained subgroup. Targeted educational interventions are required for vulnerable student groups to lift their AI literacy, cognitive flexibility and deep learning outcomes. Notably, given the cross-sectional nature of this research, all detected relationships are correlational rather than causal; causal inference cannot be drawn.