Examining the association between discussion strategies and learners' critical thinking in asynchronous online discussion

A well-designed asynchronous online discussion (AOD) has the potential to encourage learners' critical thinking (CT). Previous studies have shown that the discussion strategies selected by instructors when designing AOD questions can influence learners' CT. However, the associations between different discussion strategies and learners' CT have not been fully explored, and clear guidance for instructors on how to promote learners' CT by selecting appropriate discussion strategies is limited. In this study, a coupling deep learning model named CritiNet was developed to classify and identify learners' CT automatically based on Murphy's CT analysis model as a coding scheme ( recognize, understand, analyze, evaluate, , and create) ) in 15,483 Chinese text discussion posts. These discussion posts were generated by 505 learners in four different discussion strategies: case-based discussion, debate, open-ended discussion, and role play. Then, the associations of the discussion strategies with learners' CT were examined. Results indicated CritiNet had excellent performance in classifying Chinese text discussion posts and identifying learners' CT. Pearson's chi-squared test reported a strong association between the four discussion strategies and learners' CT. Cross-analysis revealed differences among the four discussion strategies in encouraging CT. Specifically, the case-based discussion strategy developed evaluate of CT more effectively, whereas the three other strategies promoted analyze to a greater extent. . By contrast, the open-ended strategy encouraged the least development of CT in create, , and the role play strategy generated the smallest proportion in evaluate. . The implications of these findings for instructors to encourage learners' CT effectively in AOD were discussed.