The relationship between AI and creativity may depend partly on what learners are trying to achieve. Kwan and Hung (2025) found that students with a stronger reported preference for AI assistance produced less creative designs, with the negative relationship more pronounced among students whose goals leaned toward learning and mastery. Their study raises questions about creative development, although it cannot establish that AI caused the differences.
I am particularly interested in the distinction between completing a creative assignment and developing the ability to create. An assignment can serve both purposes. When AI assistance enters the process, we need to examine which opportunities for learning remain available to the student.
What are students trying to achieve?
Goal orientation describes how people approach achievement. A learning or mastery orientation involves wanting to develop skills, understand something more deeply, and improve. A performance orientation involves wanting to demonstrate competence or do better than others.
These orientations can coexist. A student may want to become a better writer and receive a high grade. In this study, the researchers examined the relative balance between the two orientations, rather than treating students as belonging to fixed categories.
That distinction matters for teaching. We should be cautious about turning a finding involving motivational goals into a label for individual learners. Goals can also depend on the task and the expectations surrounding it.
The broader concern fits within my discussion of AI and cognition: what can a finished product tell us about the thinking that produced it? In creative work, the decisions made along the way may be especially important.
AI and creativity in a university design task
Kwan and Hung recruited 77 undergraduate students at one university. Students completed questionnaires about their creativity, goal orientation, emotional attachment to AI, and preferences for AI assistance one week before a creative task.
The AI measure compared their reported willingness to use AI with their willingness to use human alternatives across everyday situations. Students then had 45 minutes to design a T-shirt for a university student union. Three judges assessed the designs for novelty and usefulness.
A stronger preference for AI was associated with lower creative ratings. The negative association was more pronounced among students whose goals leaned toward learning and mastery, and weaker among those whose goals leaned toward performance.
The evidence for goal orientation changing this relationship was clearer for novelty than for usefulness. A weaker negative association should not be read as proof that AI improved creativity for performance-oriented students.
The design was observational: students were not randomly assigned to different levels of AI use. Reported preferences also do not establish actual usage habits. The researchers acknowledge that causality remains uncertain, the findings may not generalize beyond this type of design task, and the goal-orientation results need replication in larger samples. They did not track changes in creativity over time or separately measure idea generation and idea selection.
Keeping creative decisions with the learner
The authors suggest that teaching and training should attend to creative habits and learners’ motivational goals. I would approach that recommendation through the design of assignments. The study offers a reason to examine students’ opportunities to develop ideas; it does not validate a particular classroom intervention.
One possible adaptation is to ask students to record an initial idea before consulting AI. This could be a rough sketch, a few sentences, or an audio explanation. The purpose would be to preserve a starting point that students can later revisit. It need not become a lengthy additional submission.
After using AI, students could explain one decision: which suggestion did they accept, reject, or substantially change, and why? A useful response would connect that choice to the assignment’s audience, purpose, or constraints. Merely listing prompts would tell us much less about their judgement.
I would also make room for exploring alternatives. Students could develop two possible approaches and explain what each allows them to communicate. AI might enter later to question an assumption or suggest a constraint they have overlooked. These are proposed teaching routines, whose value would need to be examined in practice.
There is no basis here for encouraging students to pursue performance goals at the expense of learning. Nor does the study establish that motivated learners are inherently more vulnerable to AI. Its findings concern a small sample and a particular measure of relative goal orientation.
For teachers, the immediate task is to identify which creative decisions the assignment is intended to develop and ensure students get to practise them. Their explanations and revisions can help us see those decisions alongside the final work.
Reference
Kwan, L. Y. Y., & Hung, Y. S. (2025). Does AI usage diminish human creativity? How goal orientation theory moderates the negative effects between AI usage and creative output. Social Science Computer Review. Advance online publication. https://doi.org/10.1177/08944393251389125
