New study reveals students master GenAI creation before basics

New research from Eduardo Oliveira and colleagues at multiple universities challenges the common assumption that GenAI literacy follows a linear progression from foundational understanding to creative application. Using psychometric analysis (Rasch measurement and Guttman ordering) on a self-assessment instrument with 158 participants—including students, academics, and professional staff—the study mapped the perceived difficulty order of various GenAI skills.

The results reveal a stark divergence: academics tend to follow the expected linear path, but students show an 'inverted' profile, often mastering advanced creation and prompting tasks before acquiring basic conceptual knowledge about how generative AI works. The correlation between the skill difficulty rankings of students and academics was remarkably low (r = 0.188). The authors argue that this 'skill bypass' creates a fragile fluency—high self-efficacy in prompting can mask low understanding of AI mechanics. This finding provides empirical grounds for moving beyond one-size-fits-all curricula toward more adaptive, diagnostic-driven modules.