When universities advocate GAI but practice falls short: student appraisals and creative process engagement in art education

Synthesis: Chen examines how the gap between universities' public advocacy of generative AI (GAI) and their actual classroom practice shapes art students' engagement with creative process, drawing on the transactional theory of stress and coping. A two-study design tested the effects of GAI advocacy–practice inconsistency: Study 1 was a classroom-based between-subjects scenario experiment with 160 art students, and Study 2 surveyed 425 students who had recently used GAI in art learning, analyzed with PLS-SEM. In both studies, perceived inconsistency raised both challenge and hindrance appraisals. In Study 2, inconsistency had a significant negative indirect association with creative process engagement through hindrance appraisal (β = −0.060) and a positive indirect association through challenge appraisal (β = 0.217). Pathways thinking strengthened the inconsistency–challenge path but did not weaken the hindrance path. The study frames institutional word–deed inconsistency as a demand students can read as either an obstacle or a manageable developmental challenge.

Key Findings

1. Both appraisals respond to inconsistency. In Study 1, the high-inconsistency scenario raised challenge appraisal (M = 4.58 vs 2.96) and hindrance appraisal (M = 4.71 vs 3.30), both with large effects (Cohen's d = 1.22 and 1.06).
2. Inconsistency cuts both ways at the engagement level. In Study 2, inconsistency indirectly reduced creative process engagement through hindrance appraisal (β = −0.060, p = 0.004) but increased it through challenge appraisal (β = 0.217, p < 0.001).
3. Pathways thinking amplifies only the challenge route. The interaction of pathways thinking with inconsistency significantly predicted challenge appraisal (β = 0.172) and its indirect path to engagement (β = 0.075, p < 0.001), but did not significantly weaken hindrance appraisal or its negative indirect pathway.
4. The model has predictive relevance but modest explanatory power. PLS-SEM accounted for 18.9%, 28.9%, and 20.8% of the variance in the three endogenous constructs, with Q² values supporting predictive relevance.
5. No control variable mattered. Gender, education level, and GAI familiarity did not significantly predict creative process engagement.
6. The experiment was a broader inconsistency manipulation. Because guidance clarity, support availability, and assessment predictability were embedded in the scenarios, the manipulation captured a broader experience of inconsistent implementation rather than an isolated word–deed discrepancy.

What GAI advocacy–practice inconsistency means

The paper begins from the observation that universities increasingly promote responsible GAI use in art education through instructional planning, guidelines, and public commitments, yet actual implementation, instructor requirements, assessment approaches, and the handling of AI-related issues often diverge from those statements. The author defines this perceived discrepancy between institutional commitments and subsequent practices as GAI advocacy–practice inconsistency. Drawing on the transactional theory of stress and coping, the study positions this inconsistency as an environmental demand that students may appraise either as a hindrance (conflicting rules, unpredictable assessment, unfulfilled commitments) or as a challenge (an opportunity to adjust strategies and explore alternatives).

Study 1 used a classroom-based between-subjects scenario experiment. Participants in the high-inconsistency condition reported significantly higher perceived inconsistency (M = 4.78 vs 2.83, d = 2.10), validating the manipulation, and stronger challenge and hindrance appraisals. Study 2 then tested the full structural model with 425 art students who had used GAI at least once in the preceding six months, measured on 7-point scales and estimated in SmartPLS 4.1 with 5,000 bootstrap resamples.

Why appraisals diverge

The structural model revealed that the same institutional condition can operate through opposing channels. Inconsistency positively predicted hindrance appraisal (β = 0.431) and challenge appraisal (β = 0.496). Its indirect effect on creative process engagement through hindrance was negative and significant, whereas the indirect effect through challenge was positive and larger in magnitude. In other words, students who interpret inconsistent institutional support as an obstacle disengage from the creative process, while those who read it as a challenge sustain engagement. Pathways thinking — the tendency to generate multiple viable routes to a goal — strengthened the inconsistency–challenge appraisal link and its positive indirect effect on engagement, but it did not buffer the hindrance route, meaning an institutional mismatch could not be fully reframed away.

The author emphasizes that obtaining GAI-generated content is not equivalent to creative engagement. Meaningful art learning depends on critically interpreting the task, evaluating and selecting information, considering alternatives, and iteratively refining outputs. This is why the appraisal distinction matters: it determines whether students remain invested in these cognitive activities or withdraw from them.

What this means for practice

• Institutions. Align public GAI commitments with actual teaching, assessment, and support practices; word–deed inconsistency is not neutral, because it both raises hindrance appraisals and leaves some students to cope alone.
• Instructors. Make GAI expectations and assessment criteria concrete and predictable so that institutional advocacy does not translate into student uncertainty about how work will be judged.
• Instructors. Help students develop alternative routes for coping with implementation uncertainty — the data show pathways thinking supports challenge-oriented interpretation but does not eliminate the hindrance experience.
• Researchers. Report both challenge and hindrance appraisals when studying institutional GAI policy, because a single inconsistency can produce opposing effects on engagement.

Limitations

• Study 1's manipulation embedded guidance, support, and assessment features in the scenarios, so it does not isolate a purely word–deed discrepancy effect.
• Study 2 relies on self-report cross-sectional survey data; the indirect effects are associational, not causal.
• The sample is Chinese art and design students, so findings may not generalize to other disciplines or national contexts.
• Creative process engagement was measured as a reported cognitive activity, not observed creative behavior or output quality.

Citation

Chen, Y. (2026). When universities advocate GAI but practice falls short: student appraisals and creative process engagement in art education. Frontiers in Psychology, 17, 1955192.