I stopped cheating in my schoolwork many many years before I graduated, it’s probably not important but I wanted to say it anyway.
But, when I was in secondary school (high school for some), I wasn’t quite as principled. So, back then, I cheated on my tests and homework as much as the next student.
This article is about artificial intelligence and its use in submitting work that’s not actually yours, better known as plagiarism. I’m sure most have heard, read, or contemplated how it’s possible to get an AI to generate a whole project, assignment or thesis with just a few prompts. While there are many ways in which AI will change how we study (or pretend to), I’m only looking at plagiarism here.
But there are also many tools and websites, including from the current AI leader, Chat-GPT, which detect such text to degrees up to 96% or more. While this might seem good enough, I’m not so sure, for two reasons.
But before I go into that, let’s briefly explore how the AI plagiarism prevention software works. As this article explains, the software is itself another AI tool, but one designed to recognize how human and AI writing differs. In other words, they also use pattern recognition, just in a slightly different way. They look for things like a certain sentence structure being applied uniformly across the text, because humans are more random and unpredictable in their sentence constructions.
Importantly, AI detection tools have been found to be biased against people whose first language isn’t English (the tools were tested in English). This reduces their reliability, opening up room for doubt about their conclusions.
Gaming The System
First, the current model is that students who do not wish to do their work on their own engage what some in the educational field call “mercenaries.” These are people who hire out their services to students, doing everything from writing theses to impersonating the student at exams. This is important, keep it in mind as we proceed.
Second , while thinking about how it would work in the future, and I remembered an episode from my school days.
My class was writing a test and I, for whatever reason, wasn’t adequately prepared. So I asked my classmate and friend to position his hand and body in a manner more favorable to my view, a request to which he acquiesced. I then read everything he wrote and, naturally, rephrased it in my own way.
Now, I am, and have always been, a writer, so I wrote the hell out of that paper because we tend to say a lot more when we have so little to say.
As punishment, I have recently been forced to mark exam scripts where many students are doing just that and I keep shaking my head, thinking, “If you had stopped after the third sentence, I’d have graded you higher.” Truly , less is more.
Anyway, when the results came back, I was one of the highest in the class while my mate was marked as failed. The reason? Teacher was convinced that he’d copied my work. Obviously, he couldn’t reveal the truth because then we’d both have been failed, so he had to take the hit (sorry, Japhet).
Combine this with the proliferation of mercenaries and you already know where I’m going – AI, as it currently stands, would enable me to take on many more student projects, from ever increasing fields because all I’d have to do is have Chat-GPT (or an equivalent) write out the entire thesis. Next, I would check its sources to confirm validity before finally rewriting each paragraph in my own words and the current checkers would be none the wiser. Indeed, if I was particularly good at AI prompting, I may not even have to rewrite anything – I’d just tweak my prompts to make the AI generate more human sounding language.
The problem is, as I will continue to point out, most of the discussion around the effects of AI on work, study and play has been ignorant of the disruption factor. Most articles I’ve read focus on how it’ll change work and study habits, not how it’ll change work and study itself which, for me, is key.
Some disruptions will change the status quo, some will simply ingrain it further as in this case where, instead of 20 mercenaries per department, you have one group of 10 mercenaries to rule them all.
I have some more thoughts on the future of artificial intelligence in our lives and will be writing more on the subject soon. Until then.
Who needs a Publisher 