Assessing students' answers is a labor-intensive task that could significantly burden educators. Technological advances have led to the development of automatic scoring systems. However, handwriting recognition errors significantly impacts the deployment of automatic scoring in the context of paper-pencil assessment, which is still the most widely used format of academic exams. Large language models (LLMs) have proven to be effective in scoring tasks and are promising for automatic short answer grading. Nevertheless, LLMs' capability to grade the short answers with handwriting recognition errors is underexplored. The current study addressed this issue in the context of paper-pencil Chinese tests in elementary schools. The LLM used was ERNIE 4.0 because of its outstanding capability in Chinese language comprehension. We compared the grading accuracy of the model on the raw data extracted from handwritten answers by optical character recognition with that on the preprocessed data where the recognition errors were corrected. We found a substantial accuracy difference between raw data and preprocessed data, indicating that LLMs have not yet achieved the precision for grading short answers with handwriting recognition errors. Nevertheless, LLMs showed an interesting characteristic during grading: awarding points to incorrect answers as an acknowledgment of student's effort.

Reliability quantifies the extent of measurement errors in the observed feature scores and is an important quality indicator of measurements in educational research. However, the effect of feature reliability is underexplored in educational studies that use machine learning techniques. Understanding this effect is critical because the most of common features in education are contaminated by measurement errors. Recent research has revealed that the low reliability of features damages the prediction accuracy of machine learning models. The current study proposes that feature reliability also influences the generalization of machine learning models. This paper provides mathematical proof for the notion and further supports it via analyses on two empirical educational datasets. The results of data analyses also indicated that the effect of feature reliability on model generalization was moderated by the model complexity but not related to the model accuracy. Approaches to mitigate the impact of feature reliability are discussed.

Students in computerized learning environments often direct their own learning processes, which requires metacognitive awareness of what should be learned next. We investigated a novel method of measuring verbalized metacognition by applying natural language processing (NLP) to transcripts of interviews conducted in a classroom with 99 middle school students who were using a computerized learning environment. We iteratively adapted the NLP method for the linguistic characteristics of these interviews, then applied it to study three research questions regarding the relationships between verbalized metacognition and measures of 1) learning, 2) confusion, and 3) metacognitive problem-solving strategies. Verbalized metacognition was not directly related to learning, but was related to confusion and metacognitive problem-solving strategies. Results also suggested that interviews themselves may improve learning by encouraging metacognition. We discuss implications for designing computerized environments that support self-regulated learning through metacognition.

Sleep staging is fundamental for sleep assessment and disease diagnosis. Although previous attempts to classify sleep stages have achieved high classification performance, several challenges remain open: 1) How to effectively extract salient waves in multimodal sleep data; 2) How to capture the multi-scale transition rules among sleep stages; 3) How to adaptively seize the key role of specific modality for sleep staging. To address these challenges, we propose SalientSleepNet, a multimodal salient wave detection network for sleep staging. Specifically, SalientSleepNet is a temporal fully convolutional network based on the $U^2$-Net architecture that is originally proposed for salient object detection in computer vision. It is mainly composed of two independent $U^2$-like streams to extract the salient features from multimodal data, respectively. Meanwhile, the multi-scale extraction module is designed to capture multi-scale transition rules among sleep stages. Besides, the multimodal attention module is proposed to adaptively capture valuable information from multimodal data for the specific sleep stage. Experiments on the two datasets demonstrate that SalientSleepNet outperforms the state-of-the-art baselines. It is worth noting that this model has the least amount of parameters compared with the existing deep neural network models.

Confusion has been shown to be prevalent during complex learning and has mixed effects on learning. Whether confusion facilitates or hampers learning may depend on whether it is resolved or not. Confusion resolution, behind which is the resolution of cognitive disequilibrium, requires learners to possess some skills, but it is unclear what these skills are. One possibility may be metacognitive strategies (MS), strategies for regulating cognition. This study examined the relationship between confusion and actions related to MS in Betty's Brain, a computer-based learning environment. The results revealed that MS behavior differed during and outside confusion. However, confusion resolution was not related to MS behavior, and MS did not moderate the effect of confusion on learning.