Computer-Graded Essays

Once more the news has dropped a story about computer grading into our laps. Now, for most people, this would probably amount to the need for a quick shrug, if anything. Who cares, after all, whether artificial intelligence (AI) starts taking over the behind-the-scenes tedium of academia.

It's not as if I think a computer will replace me any time soon. What I do in the classroom, no computer will achieve in my lifetime. However, what concerns me here is that the best AI, at least in an English classroom, is going to ding each student for each error, whether or not any specific one should be counted as an error. Fragments have there place. In papers, even. While we could improve the software to make educated quesses (and it's true that not all humans would agree on the value of any one case of a broken rule) in such cases, but that still falls short.

Shame on any English instrctor or professor who leaves out of remedial and first-year composition classes a complete explanation of the rhetorical situation! Students must learn to write with an audience in mind—a human audience that cannot be fooled by creative sprinkling of the word "chimpanzee" throught an essay. They have to consider occasion, voice, and purpose. I assume, perhaps inadvisedly, they are aware of the topic when they begin. What does a computer know? Can it identify appropriate use of humor? Does it know the difference between writing for a wedding and writing for a funeral? Will it find, without penalty to the student, a delayed introduction?

It's dangerous ground, and we've been walking along it too long already. Until we can show—and this would take years of testing, revision, and re-norming against human-scored samples—that a computer can do the work of a human in this manner, I will take my long days and night scoring by hand, thank you. It's the worst part of the job, but it is that last one I am willing to hand off.