Learning engagement and creative self-beliefs in generative AI-supported design education: self-reported prompt literacy and dimensions of AI dependence

Generative AI (GenAI) is changing how design students learn, yet little is known about how students’ perceived capability to work with GenAI relates to their learning engagement and motivational self-beliefs, or under which forms of AI dependence this link is weaker. Drawing on social cognitive theory, this cross-sectional study examined self-reported prompt literacy, learning engagement, and two components of the creative self: creative self-efficacy (CSE), a motivational belief in one’s creative capability, and creative personal identity (CPI), the importance of creativity to one’s self-concept. Survey data came from 829 undergraduate art and design students in 40 intact classes at two institutions in mainland China; latent-variable and moderation models accounted for the clustering of students within classes. Self-reported prompt literacy was positively associated with engagement ( β = 0.54), and engagement was positively associated with both CSE ( β = 0.62) and CPI ( β = 0.47). Engagement was more strongly associated with CSE than with CPI (difference = 0.14, 95% CI [0.04, 0.24]) and remained uniquely associated with CPI after CSE was taken into account. When four dimensions of AI dependence were modeled simultaneously, the prompt literacy–engagement association was weaker among students reporting higher cognitive dependence; evidence for loss of control was weaker, and functional and emotional dependence showed no clear unique moderation. All interaction effects were small. The dimensional analysis identified cognitive dependence as the clearest moderator. These cross-sectional, self-reported findings suggest possible boundary conditions for the capability–engagement association and require longitudinal and behavioral testing.