Investigating the Validity of Using Automated Writing Evaluation in EFL Writing Assessment

摘要

This study aims to follow an argument-based approach to validation of using automated essay evaluation (AWE) system with the example of Pigai, a Chinese AWE program, in English as a Foreign Language (EFL) writing assessment in China. First, an interpretive argument was developed for its use in the course of College English. Second, three sub-studies were conducted to seek evidence of claims related to score evaluation, score generalization, score explanation, score extrapolation and feedback utilization. Major findings are: (1) Pigai yields scores that are accurate indicators of the quality of a test performance sample; (2) its scores are consistent across tasks in the same form; (3) its scoring features represent the construct of interest to some extent, yet problems of construct under-representation and construct-irrelevant features still exist; (4) its scores are consistent with teachers' judgments of students' writing ability; (5) its feedback has a positive impact on students' development of writing ability, but to some extent. These results reveal that AWE can only be used as a supplement to human evaluation, but can never replace the latter.