Psychological Mechanisms and Heterogeneity in Art Design Students' Willingness to Use AI: An Empirical Analysis Based on an Integrated Framework

Hypotheses
This study tested whether psychological factors (perceived usefulness, ease of use, AI anxiety, creative self-efficacy, attitude toward AI itself, and attitude toward using AI) predict students' willingness to use AI, and whether students show distinct subgroups with different AI adoption patterns across design stages.

Data Collection
630 Chinese art design students completed an online survey using a novel "task allocation method": indicating what percentage of work they would prefer AI to complete across five stages (creative ideation, material collection, visual design, copywriting, final optimization).

Key Findings
Students' fundamental beliefs about AI's role in art were the strongest predictors of willingness, not perceptions of usefulness or ease of use. When these deeper attitudes were considered, classic technology acceptance variables became non-significant—suggesting identity and values matter more than functional considerations in creative domains.

Three distinct profiles emerged: a pragmatic majority using AI for auxiliary tasks while maintaining human control over core creativity; a low-reliance group resistant across all stages; and a high-acceptance group viewing AI as creative partner across the entire process, including ideation.

Upper-level students were more likely high-acceptance, suggesting experience fosters openness. No differences across majors indicated patterns transcend disciplinary boundaries.

Implications
Effective AI integration requires more than technical training—educators must help students develop nuanced understandings of AI's role in creativity. Different students need different approaches: low-reliance students require foundational experience, pragmatic students benefit from gradual expansion into creative stages, and highly accepting students need guidance maintaining critical perspectives. AI willingness is task-dependent, so interventions should be stage-specific rather than promoting uniform adoption.