Psychological Mechanisms and Heterogeneity in Art Design Students' Willingness to Use AI: An Empirical Analysis Based on an Integrated Framework
Description
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.
Files
Steps to reproduce
Research Design Cross-sectional survey design combining variable-centered and person-centered approaches. Questionnaire Development Instruments (all 5-point Likert scales unless noted): Comprehensive Attitude Toward AI: 20 items adapted from Alsuwaida (2025) Perceived Usefulness: 3 items from Davis (1989) TAM Perceived Ease of Use: 3 items from Davis (1989) TAM AI Anxiety: 3 items adapted from Zhu et al. (2025) Creative Self-Efficacy: 3 items from Tierney & Farmer (2002) Attitude Toward Using AI: 4 items adapted from Zhu et al. (2025) Novel Outcome Measure - Task Allocation Method: Participants indicated percentage (0-100%) of work they would prefer AI to complete across five design stages: Creative ideation Material collection Visual element design Copywriting Final modification/optimization Demographics: Gender, age, grade, major, AI usage frequency, self-assessed proficiency. Quality control: Trap question embedded requiring "neutral" response. Data Collection Target population: Chinese art design university students. Recruitment: Online distribution via professional communities (WeChat/QQ groups), university program coordinators, social media. Sample: 721 responses collected; 630 valid after removing trap question failures (87.4% effective rate). Statistical Analysis Software: SPSS (descriptives, correlations, reliability, EFA, regression); Mplus 8.3 (latent profile analysis). Phase 1 - Preliminary: Cronbach's α, EFA with varimax rotation, descriptives (means, SDs), normality checks. Phase 2 - Variable-centered (H1): Pearson correlations; hierarchical regression (Step 1: demographics; Step 2: six psychological variables; DV: mean AI willingness across five stages). Phase 3 - Person-centered (H2): Latent profile analysis on five stage scores; models 1-4 classes tested; selection based on AIC, BIC, aBIC, entropy, LMR-LRT, BLRT, class proportions, interpretability. Phase 4 - Profile validation: Kruskal-Wallis (grade), chi-square (major), ANOVA with LSD post-hoc (psychological variables). Quality Assurance All scales α > 0.80 EFA confirmed factor structures Profile validity established via external variables Reproducibility Requirements Questionnaire (available from authors or cited sources) SPSS and Mplus 8.3+ (or R with tidyLPA) Sample size: 500+ recommended Target: art design students