Pedagogical GenAI use and self-reported creative cognition in art education: a three-wave time-lagged correlational study among design students

The integration of generative artificial intelligence (GenAI) into higher education has sparked a profound transformation across creative disciplines, raising critical questions regarding authorship, originality, and the cognitive development of students. As text-to-image engines, algorithmic design editors, and conversational assistants become permanent fixtures in university design studios, educators and researchers alike are racing to understand how these technologies impact student creativity. A recent empirical investigation conducted by researchers J. Zhang and D. Xing explores this dynamic, shedding light on the psychological mechanisms that link classroom GenAI adoption to students’ self-reported creative cognition.

Published in Frontiers in Psychology, the study examines the cognitive and emotional pathways of 389 design undergraduates at a public university in Busan, South Korea. Rather than evaluating general attitudes toward artificial intelligence or simply measuring platform-specific usage, the researchers focused specifically on pedagogical GenAI use—defined as course-related activities such as visual ideation, style exploration, prompt-based revision, alternative comparison, and feedback seeking. The findings suggest that while utilizing generative tools correlates positively with creative self-efficacy and cognitive output, the psychological benefits are primarily channeled through active student engagement rather than mere anxiety reduction.

Chronology and Research Design of the Three-Wave Study

To establish a clearer sequence among the variables than is typically achieved through traditional cross-sectional surveys, the research team implemented a three-wave, time-lagged correlational design. Data collection was structured across a single academic semester, with intervals of approximately two weeks separating each phase to reduce same-occasion response biases while maintaining a stable coursework environment.

At Wave 1, 452 participants reported their baseline pedagogical GenAI usage and demographic background. Two weeks later, at Wave 2, 431 returning participants completed assessments measuring creative self-efficacy and AI-related learning anxiety. By Wave 3, a final pool of 410 students reported on their creative engagement and self-reported creative cognition.

Following rigorous data cleaning—which eliminated respondents failing embedded attention checks or flagged for careless string-and-time responses—the final analytical sample comprised 389 complete student records, representing an 86.1% retention rate. The participant demographic reflected a diverse cross-section of design disciplines, including fashion design (28.5%), visual communication design (26.0%), digital media and animation (24.4%), and industrial and product design (21.1%). Furthermore, 44.2% of the cohort reported over a year of prior experience with generative tools, while the remainder possessed between six months and a year (33.4%) or under six months (22.4%) of familiarity.

Statistical Findings and Structural Equation Modeling

Using covariance-based structural equation modeling with robust maximum likelihood estimation, the researchers evaluated the complex interplay between technology use, psychological appraisals, and creative outcomes. The measurement model demonstrated a robust fit, evidenced by a comparative fit index (CFI) of 0.983 and a root mean square error of approximation (RMSEA) of 0.026.

The structural model accounted for 54% of the variance in self-reported creative cognition, 27% in creative engagement, 24% in creative self-efficacy, and 4% in AI-related learning anxiety. Higher pedagogical GenAI use was directly associated with greater creative self-efficacy ($beta = 0.489$, $p < 0.001$) and lower learning anxiety ($beta = -0.191$, $p = 0.002$).

The total modeled association between GenAI use and self-reported creative cognition reached $beta = 0.456$, with indirect components accounting for approximately 59% of this total. Notably, the serial indirect association operating through self-efficacy and subsequent creative engagement was substantial ($beta = 0.122$), whereas the corresponding anxiety-engagement path played a minimal role ($beta = 0.015$). The direct association from anxiety to creative cognition was not statistically significant, challenging the popular assumption that AI tools primarily enhance student creativity by alleviating academic stress or technology-related fears.

Comparative Analysis of Competence and Anxiety Pathways

The study’s results offer a nuanced perspective on the social-cognitive framework governing human-AI interaction in art education. The data revealed that creative self-efficacy and AI-related learning anxiety function as distinct, parallel appraisals rather than opposite poles of a single psychological spectrum, as evidenced by a near-zero correlation between the two constructs ($r = 0.04$).

While self-efficacy demonstrated both a direct link to creative cognition and an indirect pathway mediated by student engagement, anxiety exerted influence solely through an engagement-dependent channel. The simple, unmediated anxiety route to cognition was negligible. Consequently, the researchers conclude that competence beliefs—specifically, a student’s confidence in generating and refining ideas—play a far larger role in fostering creative cognition when using GenAI than the reduction of anxiety.

Engagement emerged as the strongest concurrent predictor of self-reported creative cognition ($beta = 0.495$). This highlights the behavioral, emotional, and cognitive investment students bring to their creative tasks as the primary vehicle through which technological affordances translate into cognitive gains.

Implications for Design Studios and Curriculum Design

As universities grapple with the integration of artificial intelligence into studio curricula, the findings provide empirical guidance for educators seeking to balance technological efficiency with authentic human creativity. Rather than banning generative tools or granting uncritical access, design programs are encouraged to structure assignments in ways that promote active engagement, critical evaluation, and iterative problem-solving.

The data suggest that coursework requiring students to compare alternative outputs, document prompt-based revisions, and justify stylistic choices naturally aligns with higher creative self-efficacy and engagement. Conversely, relying on GenAI merely to bypass the ideation phase in pursuit of rapid final deliverables may undermine the deep cognitive processing necessary for authentic artistic development.

Methodological Limitations and Future Directions

Despite the rigorous analytical approach, the authors acknowledge several limitations that qualify the scope of their findings. Because the study was observational and lacked experimental manipulation, the paths represent statistical associations rather than proven causal effects. Furthermore, variables such as baseline creative performance, digital proficiency, and instructional quality were not controlled for, leaving room for potential unmeasured confounding.

The reliance on self-reported measures for creative cognition—rather than objective, rubric-based evaluations by independent expert panels—also points to the need for future research incorporating multi-source assessments. Additionally, because engagement and cognition were measured concurrently at Wave 3, the temporal direction of that final link remains theoretically guided rather than definitively identified.

Future empirical investigations are recommended to utilize longitudinal designs with baseline controls, incorporate behavioral log data from AI platforms, and evaluate actual creative outputs across diverse artistic mediums such as music, performance, and fine arts. By doing so, educational researchers can continue to map the complex boundaries where human ingenuity intersects with machine intelligence.