Critique of Mark D. Shermis & Ben Hamner, 'Contrasting State-of-the-Art Automated Scoring of Essays: Analysis'
- Author(s): Perelman, Les C.
- et al.
Although the unpublished study by Shermis & Hamner (2012) received substantial publicity about its claim that automated essay scoring (AES) of student essays was as accurate as scoring by human readers, a close examination of the paper's methodology demonstrates that the data and analytic procedures employed in the study do not support such a claim. The most notable shortcoming in the study is the absence of any articulated construct for writing, the variable that is being measured. Indeed, half of the writing samples used were not essays but short one-paragraph responses involving literary analysis or reading comprehension that were not evaluated on any construct involving writing. In addition, the study's methodology employed one method for calculating the reliability of human readers and a different method for calculating the reliability of machines, this difference artificially privileging the machines in half the writing samples. Moreover, many of the study's conclusions were based on impressionistic and sometimes inaccurate comparisons drawn without the performance of any statistical tests. Finally, there was no standard testing of the model as a whole for significance, which, given the large number of comparisons, allowed machine variables to occasionally surpass human readers merely through random chance. These defects in methodology and reporting should prompt the authors to consider formally retracting the study. Furthermore, because of the widespread publicity surrounding this study and because its findings may be used by states and state consortia in implementing the Common Core State Standards, the authors should make the test data publicly available for analysis.