Describe how computer scoring is used to achieve the purpose of scoring written submissions. What are the risks and benefits of substituting machine scoring for human grading?

Start with the proposition that writing assignments, which employ grammatically complex constructions, use expected words and expressions, include sentences with greater average word counts, and include fewer spelling mistakes are statistically significantly correlated with higher grades, while assignments that are missing one or more of those features are statistically correlated with lower grades. From this starting point design computer programs that assign grades by parsing grammar and counting words. With these kinds of computer programs, written assignments, like those on college admissions tests, can be graded by computer. Sophisticated computer grading programs can generate the same grades as well qualified humans 99.9 percent of the time, but the computer does its processing without comprehending the meaning of what is written and without applying the four tests of the quality of arguments presented. [See more of the answer in the box "Grading Written Submissions" near the end of the "Coincidences, Patterns, Correlations, and Causes" section.]