大规模开放在线学习学生互评效果实证研究

An Empirical Study on the Effect of Peer Assessment in Massive Open Online Learning

Peer Assessment has the potential to overcome the limitations of teacher-grading and machine-grading， and therefore has become an important assessment method in the context of massive open online learning. However， there is a lack of empirical research that investigate the accuracy and effects of online peer assessment. To address such research need， this study analyzed the empirical data of peer assessment， teacher-grading and self~grading from a massive open online course (MOOG) ， and provided preliminary findings regarding the reliability， validity， influencing factors and perceived benefits of peer assessment for this specific context. The results show that although peer assessment suffers from inter-rater reliability， the composite score from all student graders can provide a reliable and consistent final result. The correlation coefficient of 0. 619 indicates a high convergent validity for peer assessment. In addition， this study also explored the influence of factors such as training and calibration， number of graders and score calculation methods. Moreover， the post-course survey results reveal that online learners in general have positive attitude towards peer assessment， acknowledging benefits such as feedback acquisition， higher engagement and development of higher-order thinking. The findings of this study are expected to inform the revision and improvement of assessment models for massive open online learning.