Can artificial intelligence assistance enhance creativity in university students? An experimental study

Abstract

This study investigates how artificial intelligence (AI) assistance—using the DeepSeek-R1 large language model—influences the novelty and usefulness of creative output among university students, and examines whether different interaction modes moderate this relationship. Two experiments were conducted: Experiment 1 (n = 130) employed a blank-control design to compare a manual control group with an AI-assisted group on tasks of case analysis and entrepreneurial planning, while Experiment 2 (n = 136) adopted an inter-group design to compare single-round simple interaction and multi-round deep interaction under AI assistance. Results showed that AI assistance was associated with enhanced novelty (M = 4.95 vs 3.09, p < 0.05) and usefulness (M = 4.92 vs 2.92, p < 0.05) compared with the manual approach. Furthermore, multi-round deep interaction was linked to higher scores in novelty (M = 6.74 vs 4.44, p < 0.05) and usefulness (M = 6.71 vs 4.49, p < 0.05) than single-round simple interaction. These findings suggest that AI assistance may enhance both novelty and usefulness in student creativity, and that multi-round deep interaction could amplify this effect. The study thus provides evidence consistent with a role for interaction depth in human–AI co-creation, offering a theoretical perspective and potential design principles for developing AI-enhanced creativity tools in education.

Ähnliche Arbeiten