Investigating student engagement with personalised generative AI feedback: Effects on creative problem-solving and feedback uptake

Abstract

Personalised generative artificial intelligence (GenAI) feedback holds potential for supporting students’ creative problem-solving, yet empirical evidence on its effects and underlying mechanisms remains limited. This study addressed these gaps through a within-subjects experiment with 62 Chinese university students, aiming to investigate the effects of personalised GenAI feedback on students’ creative problem-solving and feedback uptake in comparison with standard GenAI feedback. To further reveal the mechanisms underpinning these effects, we examined how students’ interactions and behavioural uptake of feedback during human–AI collaboration were associated with their creative problem-solving performance. Results showed that personalised GenAI feedback improved the quality and elaboration dimensions of students’ solutions, while slightly reducing the amount of ignored feedback compared with the standard GenAI condition. Moreover, higher-level uptake of GenAI feedback, reflected in the amount of applied feedback, positively predicted creative problem-solving performance across conditions, and interaction behaviours such as co-constructing ideas and negotiating inconsistencies with AI were consistently associated with this high-level uptake. This study offers insights for designing GenAI feedback that emphasises personalisation to foster students’ creative engagement, underscoring the critical role of students’ interactive collaboration with AI-generated feedback in cultivating innovation within GenAI-supported learning environments.

Implications for practice or policy:

• When designing personalised agents, AI developers should integrate complementary feedback mechanisms to enhance both student engagement and creative development.
• Educators should prioritise fostering students’ higher-level uptake of GenAI feedback, as this is critical to its effectiveness on creative production.
• Instructional approaches should be designed to cultivate students’ interactive engagement with GenAI tools and develop their creative capacities, thereby fostering productive human–AI collaboration.