Huican Huo: When AI gives advice, who decides what happens next?

This study followed 27 learners through a 12-week academic writing course in New Zealand. We analysed 1,127 episodes of AI interaction alongside journals and interviews to understand learners’ decisions.

Explicit checking occurred in 50.1% of episodes. Learners described comparing AI advice with assessment criteria, teacher guidance and prior knowledge, often when they felt unsure. These accounts show why understanding learners’ judgement requires looking at their reasons for checking and what they did next.