Toward Automated Support for Teacher-Facilitated Formative Feedback on Student Writing

Formative, content-level feedback on student writing has been shown to have positive impacts on both writing and learning outcomes. However, many teachers struggle to provide this type of feedback to large classrooms of students. This paper takes an initial step towards supporting teacher-facilitated feedback through the use of automated and user-directed topic discovery. 114 student essays were collected from a local underperforming middle school as part of a pilot study for Write Local, a digital repository and workspace for authentic problem-based learning activities. Predictive models were built and evaluated to explore the impact of different topic discovery approaches as well as correction of student spelling errors on model accuracy. The resulting models provide promising direction for scaffolding teachers in providing formative feedback on content-level features of students’ problem-based writing.