Examining the relationship between peer feedback classified by deep learning and online learning burnout

Students are prone to experiencing learning burnout while engaged in online learning due to the lack of interaction. However, peer feedback can play an important role in alleviating learning burnout. This study uses a combination of deep learning (DL) and statistical analysis to investigate the relationship between online peer feedback and learning burnout. Several DL algorithms, including Bidirectional Encoder Representations from Transformers (BERT), Bi-directional Long Short-Term Memory (Bi-LSTM), and BERT-Bi-LSTM, are compared to classify peer comments in terms of cognitive content and affective state. Moreover, by using the feedback classified by DL and survey data from 116 participants, multiple linear regression (MLR) analysis is employed to explore the relationships between online learning burnout and feedback messages. The results show that the BERT model achieves the best performance in terms of classification performance and agreement with manual coding. Further, we find that student burnout is significantly influenced by peer feedback. Individuals who receive more suggestive feedback experience a greater reduction in emotional exhaustion. However, when receiving more reinforcing feedback without guidance, learners' behavior tends to deteriorate. From an affective perspective, the more positive feedback the learners receive, the more they tend to exhibit higher levels of emotional exhaustion and a diminished sense of achievement. And when receiving more negative feedback, learners tend to have a worse emotional experience and demonstrate poorer self-learning behavior. This study contributes to our understanding of the application of DL in peer assessment and the impact of received peer feedback on preventing and alleviating learning burnout.