Essays used to be an effective way to gauge a student’s ability, but in the age of AI it can be hard to identify who even wrote the work handed in, writes Jonathan Malloy.lisapics/iStockPhoto / Getty Images
Jonathan Malloy is a professor in the department of political science at Carleton University.
Belts are tightening in my university department. People retire and aren’t replaced. Free coffee is a distant memory. But one expense is steadily rising in our office: paper exam booklets.
In the age of generative artificial intelligence, instructors are bringing back the traditional handwritten exam, the only surefire way to ensure students are actually producing their own work. To assign a traditional essay or take-home exam now is to create endless work for yourself, puzzling over whether you are grading the efforts of the student or ChatGPT.
We realize the image of professors furiously trying to hold back the rising tide of AI makes us look even more Luddite. But the underlying problem is not hidebound academics resistant to change. The problem is that no one has yet figured out how to teach students on a mass scale in the age of AI.
It’s easy to come up with AI activities. The trick is how to figure out if the student is actually learning anything. The best way is working iteratively and closely with students; coaching them on the use of AI tools to write, analyze data, and much else; assessing and refining the results; encouraging group collaboration; and using check-ins for evaluation at each stage.
Jonathan Malloy: Canada’s universities are in crisis as money gets tighter
But we need to scale this up. The Canadian postsecondary system has classes with dozens, often hundreds, of students in first- and second-year courses. This is good for accessibility, but also anonymity. Only in small upper-year courses can instructors even learn every student’s name, much less personally monitor their intellectual growth. Those who want to fly under the radar can easily do so.
The essay has been the best vehicle to get around this. Professors don’t enjoy marking essays any more than students enjoy writing them. But they are the most effective way to gauge a student’s ability to organize and express their thoughts cohesively, even if you don’t really know the student.
But now AI can do much of that, and without knowing the student in any depth, we have no idea of their own thinking ability, and who wrote the essay they’ve handed in. Nor, in the absence of incontrovertible evidence, can we prove the use of AI enough to incorporate it into grading. So we turn back to the artificial controlled environment of the written exam.
AI has led to innovation in its own way here, as instructors now hustle to come up with ways around traditional essays and written assignments, and develop their own methods to discourage AI use. But this produces its own problems and inequities. For example, an instructor could require a five-minute interview with each student to make sure they actually understand the essay they supposedly wrote. That has its own risks, turning instructors into interrogators and leaving things up to their subjective judgment. The challenge remains how to scale up; to match seemingly-infinite AI capabilities with finite human time and energy.
My own approach has been to try and engage AI, such as asking students to use generative AI to generate short essays and reflect on the results. Yet it’s hard to build a course around this. AI is also a surefire topic for class discussion: Students are often willing to discuss its use and are well aware of its issues. But turning that into marketable skills is difficult. As a few observers have pointed out, there may be a growing need for more attention to AI ethics, but it’s hard to find a job ad that says “AI ethicist wanted.”
AI can do amazing things, but it can’t substitute for underlying human qualities. A student can use AI tools to accelerate their productivity and efficiency. But if they struggle to even make it to class on time, AI is not going to help. AI can gather and analyze data, and summarize long complex texts. But it can’t help students grasp the paradoxes, irony and wicked problems that characterize human society. Making the link between using AI and actually growing one’s intellect and talents is the challenge of our times. Until we solve it, we will keep stocking up on exam booklets.