Student Responses to Cognitive Conflicts in GenAI ‐Assisted Learning: A Covariate‐Dependent Markov Chain Analysis on Human‐ AI Interactions

Background As generative artificial intelligence (GenAI) reshapes the learning process, students will inevitably experience cognitive conflicts when their existing knowledge and beliefs clash with information provided by GenAI. Improper handling of cognitive conflicts may lead to a decline in students' learning efficiency and even hinder the development of critical thinking skills. Objective This study aims to understand how students respond to cognitive conflicts during their interactions with GenAI, with a focus on comparing high‐performing and low‐performing student groups. Methods To identify relationships between cognitive conflict intensity and students' responses, we used a covariate‐dependent Markov chain (a sequential data modelling method) to analyse the response states of 411 students in a simulated academic writing task and observed transitions in student response states across different task performances. Results The results revealed that the intensity of cognitive conflict had a significant impact on students' state transitions. Overall, as the level of conflict increased, students' state changes exhibited a phased pattern, shifting from fluctuation (transitioning to a variety of different states) to inertia or convergence (maintaining the original state or transitioning uniformly to a negative response state). Student response state transitions varied across different task performance levels. Conclusions This study provides a novel methodological perspective for modelling sequential response dynamics in Human‐AI interaction. The findings offer new insights into the cognitive conflict dynamics in GenAI‐assisted learning, providing implications for designing GenAI‐assisted learning environments.