Imagine you are in class waiting to get your essay back from your teacher. You put in the background research, followed your sources, brainstormed your thoughts, and, through several rounds of revision, wrote something you are proud of.
All of this, to see at the top of the page, what may become a sign of impending doom:
“AI Generation Detected.”
This feeling may become all too real for academic writing. There seems to be an existential threat on the horizon: How do we safely safeguard against academic fraud?
Plagiarism has been a scourge in academics for as long as there has been academics. Whether it is deceitful stealing of others work, or the propagation of power imbalances of individuals, plagiarism crashes us short of the idealistic environment of the triumph of ‘good ideas’ in a civil discourse.
For those of us still in the academic world, Generative AI gives us another layer to this question. “What if the ideas that you are taking do not belong to anyone else?” This kind of thinking was not necessary 30 years ago. We could simply assume that if an idea was not your own, it was someone else’s.
First, what even is plagiarism? The University of Pittsburgh’s Department of English uses the following definition: “to steal and pass off (the ideas or words of another) as one's own; use (a created production) without crediting the source; to commit literary theft: present as new and original an idea or product derived from an existing source. —Webster's Ninth New Collegiate Dictionary”
Under this definition, not everything needs to be cited. For instance: what is the name of the American Football team based in Pittsburgh? The Steelers, of course. This factoid is common enough knowledge that nobody could present it as their own. If I were to make a claim like “The name of the Pittsburgh football team is The Steelers” in a paper, I would not need to cite. This is not really a word or idea that I am trying to pass off as my own.
Now, what if I was writing a piece on why the team has the name “The Steelers”, which is steeped in the industrial history of Western Pennsylvania, notable the steel industry whose remnants are still scattered around the region. Well, I would quickly exhaust my knowledge of the subject. So, I would turn to articles, libraries, or even interviews. The information and ideas I collected, I just admitted, would not be on my own. So, they would need a citation.
These two examples are ‘classical’ in the sense that we, as an academic community, have the wherewithal to handle these based on our academic tradition. Now, consider, what would you do if you saw these statements at the beginning of my piece on the history of the name “The Steelers”
-
The Steelers are named for the steel-industry that dominated the region of Western Pennsylvania, and surround, so-called, steel-belt. Taking their emblem, colors, and culture from the working class, who would become their storied fanbase, was a way to endear the newly formed team, of a newly regulated sport, and attract fans. By outwardly and deliberately signifying the team as an outgrowth of community, culture, and working-class spirit, The Steelers, to today, remain an emblem of working-class football.
-
The Steelers are named after the people who worked in the steel mills in and around Pittsburgh. The owners of the team hoped that this would make the team seem like part of the community and more people would come to the games.
If you pop the two of these into AI detectors (which are dubious at best, in any case), which would you guess would be more likely to suggest AI generation?
Maybe there are some ChatGPT-isms that stand out to you in the first: the embedded clauses, the expanding of points, and the positive uplifting tone.
But now, imagine that this was a serious piece about a medical discovery, or perhaps your crowning article required for your doctoral defense. A much more serious situation. And someone claims to see these markers of ChatGPT in your work – even though you did not ever open it, or any other Generative AI software. (Check out ‘Obviously ChatGPT’ — how reviewers accused me of scientific fraud (nature.com) for a real example!)
These false positive detections of fraud could indeed be a very real ‘witch hunt’ that we find ourselves in. If one can call into question work simply by asserting that Generative AI was used, what defense mechanism do we have? Other than laboriously tracking updates to manuscripts, what responses will we have to reviewers who say that this work was not our own. Reviewers who stay anonymous? Reviewers who stay anonymous and compete for funding in the field?
These questions are cropping up here and there and would receive help from being headed off early. Academia is built upon a cornerstone of sincerity. Being genuine is critical for us to be able to work for the betterment of society.
The time to answer these questions is here.
And if you need more convincing, look back at my blurbs on The Steelers, and double check which you think was written by AI...
Not knowing can be scary, threatening, and call into question everything I’ve written in this article.
Fortunately, I wrote them both.
Daniel Crawford
Pitt GAINS