Students' Revision of Text Evidence Use in an Automated Writing Evaluation System

Published in: Assessing Writing (2020). doi: 10.1016/j.asw.2020.100449

Posted on on February 26, 2020

by Elaine Lin Wang, Lindsay Clare Matsumura, Richard Correnti, Diane Litman, Haoran Zhang, Emily Howe, Ahmed Magooda, Rafael Quintana

We investigate students' implementation of the feedback messages they received in an automated writing evaluation system (eRevise) that aims to improve students' use of text evidence in their writing. Seven 5th and 6th-grade teachers implemented eRevise (n = 143 students). Qualitative analysis of students' essays across first and second drafts suggests that the majority of students made changes to their essays that were in line with the feedback they received, though few of these changes resulted in substantive improvement in essay quality. Twenty percent of students did not attempt to implement the feedback; these students generally made small changes to wording or mechanics. In response to the feedback to add more evidence, students whose essays did not improve or showed only slight improvement frequently added in evidence that was not text based or repeated evidence already present in the first draft. When prompted to explain how the evidence they included connected to their claim, many students paraphrased the evidence, added a short conclusion, or explained generally how the evidence supports claims (not how this was instantiated in their writing). Implications for teaching argument writing and for designing AWE systems that support students to successfully revise their essays are discussed.

Research conducted by

This report is part of the RAND Corporation External publication series. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.