Abstract
The rapid integration of generative artificial intelligence (GenAI) into higher education has created a paradoxical landscape for college students: while technological advancements offer unprecedented convenience, they simultaneously exacerbate the knowledge-practice gap in AI Literacy cultivation. Traditional educational frameworks struggle to address the dynamic interplay between AI-mediated learning environments, ethical dilemmas, and competency development, leaving a critical theoretical and practical void in literacy cultivation models. To bridge this gap, this study pioneered an exploratory sequential mixed-methods design, combining qualitative interviews (n = 30) and quantitative surveys (n = 590, response rate 98.33%) to unravel the mechanisms through which GenAI use enhances students' AI Literacy. Qualitative analysis revealed a spiral-ascending literacy construction model characterized by iterative cycles of cognition-practice-evaluation, wherein 82% of participants demonstrated critical awareness of algorithmic biases and privacy risks. Quantitative results further validated a novel theoretical framework, showing that the social environment indirectly drives application practice via perceived impressions (path coefficient = 0.294, p < 0.001), with group needs fully mediating this relationship (p = 0.439 for the direct path). Structural equation modeling also identified key pathways linking perceived ease of use (β = 0.477) and technological expectations (β = 0.284) to behavioral adoption and future-oriented literacy. These findings challenge linear literacy models by emphasizing ecological dynamics and recursive learning processes, offering actionable insights for designing AI-integrated curricula and policies. Collectively, this research underscores the necessity of multi-dimensional interventions, combining cognitive scaffolding, ethical education, and skill training, to transform passive AI utilization into active literacy cultivation in the digital age.