A Teacher's Perspective: Using Automated Feedback in Science Classrooms (Poster 6)

• In Event: Automated Analysis of Students' Science Explanations: Techniques, Issues, and Challenges

Abstract

Introduction
In this presentation, I will discuss how I used an automated natural language processing (NLP) technology, [name blinded] in my classroom, that provided students with real-time feedback for fostering scientific writing. As an integral part of the development process, I was part of co-design sessions (Penuel et al., 2007) in which I worked with the research team on the kinds of feedback that would help students write better science explanations, and provided input based on my pedagogical insights to shape the technological innovation and its application (Cober et al., 2015; Holstein et al., 2018; Kali et al., 2015). Although some co-design studies have explored automated feedback in writing tools (e.g., Bywater et al., 2022; Conijn et al., 2020), few have examined how teachers can inform the design of automated feedback for science classrooms.

Lessons Learned
A design-based Physics unit where students designed a fun yet safe roller coaster was used in this study. I taught the roller coaster unit over 15 instructional days in an 8th-grade science classroom (n=90) in the midwestern US. Students explored relationships between height, mass, and energy before writing an initial essay. After receiving feedback, students made revisions to their essays. They wrote a final essay at the end of the unit.
I found that students arrived eager to view feedback after submitting their essays on both occasions. Traditionally, this feedback would come a week or two later after I read each essay, and it was often accompanied by a final grade. This time, it was an opportunity to use the immediate feedback to make edits and changes during the unit. Students were not only willing to take the time to make the changes but looked forward to the chance to show what they learned and improve their written work. The immediate and specific feedback allowed me to work with each student individually based on what was missing, incorrect, or unclear in the particular essay. It provided the time to reteach and clarify during the unit, rather than reflecting on what to change after the unit was completed.
While the automated feedback was extremely beneficial, it could be improved to provide more specific feedback for individual students. Many reported reading the same sentences in their feedback, resulting in the feeling that it wasn’t personalized and reduced desire for revision. Additionally, the feedback often linked multiple concepts into one suggestion, and students did not know where to start to make the changes. Students also struggle with revising, focusing more on “additions”, adding extra sentences related to their feedback topics, rather than fixing their original statements. As a teacher, I need to focus on direct instruction related to revising for concise and specific scientific writing.
Based on this experience, the research team is working on revising the essay structure and the feedback.

Author

• Jessica North, Waunakee Community School District
  Presenting Author