This article is coming to the wrong conclusion from the facts given. That the grades given by these algorithms matched the grades given by human essay reviewers on standardized tests almost exactly doesn't mean that our algorithms are great and that they should be rolled out to normal school environments to help students learn to write. It means that our method of grading essays on standardized tests is absolute garbage, because the algorithms can't judge semantic meaning or depth of insight. We're basically grading people on how prepared they are for writing About.com content farm articles.