Search for: All records

Open-text responses provide researchers and educators with rich, nuanced insights that multiple-choice questions cannot capture. When reliably assessed, such responses have the potential to enhance teaching and learning. However, scaling and consistently capturing these nuances remain significant challenges, limiting the widespread use of open-text questions in educational research and assessments. In this paper, we introduce and evaluate GradeOpt, a unified multiagent automatic short-answer grading (ASAG) framework that leverages large language models (LLMs) as graders for short-answer responses. More importantly, GradeOpt incorporates two additional LLM-based agents—the reflector and the refiner—into the multi-agent system. This enables GradeOpt to automatically optimize the original grading guidelines by performing self-reflection on its errors. To assess GradeOpt's effectiveness, we conducted experiments on two representative ASAG datasets, which include items designed to capture key aspects of teachers' pedagogical knowledge and students' learning progress. Our results demonstrate that GradeOpt consistently outperforms representative baselines in both grading accuracy and alignment with human evaluators across different knowledge domains. Finally, comprehensive ablation studies validate the contributions of GradeOpt's individual components, confirming their impact on overall performance. 

Abstract BackgroundWhat and how teachers learn through teaching without external guidance has long been of interest to researchers. Yet limited research has been conducted to investigate how learning through teaching occurs. The microgenetic approach (Siegler and Crowley, American Psychologist 46:606–620, 1991) has been useful in identifying the process of student learning. Using this approach, we investigated the development of teacher knowledge through teaching as well as which factors hinder or promote such development. ResultsOur findings suggest that teachers developed various components of teacher knowledge through teaching without external professional guidance. Further, we found that the extent to which teachers gained content-free or content-specific knowledge through teaching depended on their robust understanding of the concept being taught (i.e., content knowledge), the cognitive demand of the tasks used in teaching, and the lesson structure chosen (i.e., student centered vs. teacher centered). ConclusionsIn this study, we explored teacher learning through teaching and identified the sources leading to such learning. Our findings underscore the importance of teachers’ robust understanding of the content being taught, the tasks used in teaching, and a lesson structure that promotes teachers’ learning through teaching on their own. 

This article explores three attributes of teachers’ understanding of fraction magnitude: the accuracy and reasonableness of teachers’ estimations in response to fraction arithmetic tasks as well as the alignment of the estimation strategies they used with the concept of fraction magnitude. The data were collected from a national sample of mathematics teachers in Grades 3–7 in which fraction concepts were taught (N = 603). The results indicated the teachers’ estimations were only partially accurate and reasonable, particularly when fraction division was involved. Furthermore, teachers’ credentials and the grade level at which they taught mathematics were significantly related to teachers’ understanding of fraction magnitude.