John Warner renews case against AI grading student writing

John Warner has revived his case against automated grading of student writing, arguing in Inside Higher Ed that no matter how plausible generative AI's feedback has become, the practice violates the basic contract of a writing assignment. "The purpose of writing is to communicate with an audience," he writes. "In good conscience, we cannot ask students to write something that will not be read."

The June 15, 2023 column, part of his "Just Visiting" series, extends an objection Warner first published more than a decade ago against edX's algorithmic grading model. What has changed since then, he acknowledges, is that today's systems can generate readable, prompt-appropriate responses to student prose. What has not changed, in his view, is the pedagogy underneath: reading and responding to a student's writing is the instructor's core work, and handing it to a machine is to "willingly cede some portion of your humanity." Two researchers we follow have pushed the piece into our feed, a sign the argument is still being pulled into faculty debates as models improve.

Warner is blunt about what he sees as a category confusion. If a grading model appears to work, he argues, that says more about how the assignment was designed than about the machine's competence; a task a language model can pass was likely measuring the wrong thing. The systems themselves, he insists, "cannot read. They cannot think. They cannot feel." His conclusion is that educators should draw a firm boundary while the wider ethics of AI in classrooms are still being worked out, and he doubts the vendors pushing the tools will grant that time.