Grading can overwhelm teachers as the semester ends. Does AI present a better way?

April 14, 2023 3:33 am
an illustration of a robot coming out of the screen of an oversized laptop surrounded by students raising their hands

Learning can remain substantive and vigorous in an age of automation, writes Eric Thomas. Teachers must resist the temptations of efficiency offered by artificial intelligence. (Getty Images)

Exam question No. 1: Multiple Choice

Once you respond to this question on the final exam, who is going to be grading it? 

  1. Your instructor, a human being
  2. A machine, which senses the presence of #2 pencil on a Scantron sheet
  3. A large language model, essentially a labyrinth of computer code
  4. It really doesn’t matter, because I didn’t study for the exam

Exam question No. 2: Essay

With all of the recent talk about students using artificial intelligence in academic work, how should we think about the flip side of the question: What about automated grading? 

A few years ago, Washington Post columnist and Kansan David Von Drehle presented a session to high school journalists in Kansas City. Among other things, he spoke about a defining moment in his education.

A professor at the Denver University, the school Von Drehle attended as an undergraduate, assigned a piece of writing that seemed pedestrian. Von Drehle said that he considered it a simple prompt that he could complete quickly and earn the credit he needed from the class. Von Drehle submitted it and was ready to move along.

However, the professor returned to him a detailed, challenging and extensive critique — one that was longer (and perhaps more thoughtful) than Von Drehle’s submitted assignment. Through the semester that included many revisions and office hours visits, Von Drehle said he learned intellectual curiosity, the grit required to revise and more.

This anecdote pesters me every time I grade and respond to student work as an instructor at the University of Kansas, and before that as a high school teacher. What can we accomplish as instructors when we “grade papers?” And how can we rise above merely grading to change how students learn?

April is a good moment to ask this question. As the end of the semester approaches, Kansas school teachers and college instructors have dozens, if not hundreds, of assignments waiting for their evaluating eyes. (My apologies to some of my students, whose work is queued up as I write these words.)

Pour a cup of coffee for the teacher in your life. They are going to need the caffeine to get through all of that grading.

Unless they don’t do the grading at all.

Automated grading, in one form or another, has swooped in for decades to rescue teachers from hours spent with a red pen and a stack of papers. The multiple choice question, a ubiquitous part of American education, might be seen as the peak of efficiency. Distill a student’s complex understanding about a topic into a simple response: a single letter that is either correct or incorrect.

Commonly, students would have recorded that multiple choice question with a #2 pencil on a Scantron answer sheet. Teachers basked in the time-saving satisfaction that  a tidy stack of Scantrons created. A few minutes of feeding them into a whirring machine, and poof … the test is graded.

The staccato sound of a Scantron machine in the teachers lounge is fading these days as the multiple choice tests have moved online. The antiquated steps of using paper, pencil and machine can be eliminated by Canvas, Moodle and Google Classrooms. And students receive feedback instantly, a principle that educators recognize as valuable.

What about questions that require students to write, rather than selecting from a menu of answers? What about grading the dreaded essay questions — or even a question that prompts students for a few paragraphs?

If we aim to change students’ worldviews and habits, like the VonDrehle example suggests, we surely need to ask for more than multiple-choice thinking. And we need to provide some feedback on that student’s writing.

At universities where classes frequently exceed 100 students, asking for students to write 500 words in their Blue Books is to sentence yourself, the instructor, to 50,000 words of reading — plus critique.

That word count approaches reading and annotating a novel. As one of my students once noted as she stared at my stack of grading, “a really crappy novel.”

Sure, I have made attempts in the past to make my grading quicker. I’ve prewritten comments that I can cut and paste in response to common errors. However, students often can suss out these time-saving tricks. They sometimes see the comments as a mass of cookie-cutter confusion littering their paper. “Did my instructor even write this?” they wonder.

Artificial intelligence, a tool that currently fills many educators with fear because students are using it to avoid writing themselves, can also be used for grading. The companies that offer this software promise many benefits: the elimination of bias, the saving of time and creation of more detailed feedback. Teachers can even scan physical exams created with student handwriting.

The company that created one of these systems, Copyleaks, says that its system can learn the standards that the human graders expect, imitate the human graders and assign scores “all in a matter of minutes.”

The website says, “If you’re looking to assess tens of thousands of standardized tests at the state, national, and university-wide levels, AI Grader is right for you.”

If VonDrehle’s anecdote is the gold standard of education, this might seem absurd. The degree of human learning has been reduced to an electronic transaction of binary data: anonymous yet efficient. An earnest human student expects an earnest human teacher on the other side of the transaction. Anything less feels like an intellectual betrayal.

Even more dystopian, imagine this scenario. A teacher searches for an essay question for a final exam on American history. Why wouldn’t the teacher ask ChatGPT for some suggestions? The question that ChatGPT creates for the teacher goes onto the test.

That question befuddles a student in the class so thoroughly that she relies on Google’s Bard for the answer. She submits “her” answer, and the instructor feeds the student’s essay into an AI grading system.

The result? An echo chamber of artificial intelligence. One large language model asks a question that another answers and yet another assesses. The teacher and student are effectively absent.

Of course, it doesn’t have to be this way. Learning can remain substantive and vigorous in an age of automation. But it won’t happen without action. Lawmakers must provide funding to enable small public school classes, rather than diverting funds to private schools. Teachers must resist the temptations of efficiency.

After all, we should demonstrate to students the value of learning through challenging, and often humbling, attempts at writing something meaningful.

Eric Thomas directs the Kansas Scholastic Press Association and teaches visual journalism and photojournalism at the University of Kansas. Through its opinion section, Kansas Reflector works to amplify the voices of people who are affected by public policies or excluded from public debate. Find information, including how to submit your own commentary, here.

Our stories may be republished online or in print under Creative Commons license CC BY-NC-ND 4.0. We ask that you edit only for style or to shorten, provide proper attribution and link to our web site. Please see our republishing guidelines for use of photos and graphics.

Eric Thomas
Eric Thomas

Eric Thomas directs the Kansas Scholastic Press Association, a nonprofit that supports student journalism throughout the state. He also teaches visual journalism and photojournalism at the William Allen White School of Journalism and Mass Communication at the University of Kansas in Lawrence. He lives in Leawood with his wife and two children.