We Hold These Truths to be…AI Generated?
A friend recently sought my advice about a writing contest he was organizing. This contest was for high school students and would offer a scholarship to the winner. Specifically, he was concerned about entrants using AI to write their essays – or worse, someone actually winning the contest while using AI to write their paper!
While they had added a stipulation that all entries must be personally-generated (explicitly prohibiting AI), he wanted to know which AI detector tool was best. They intended to screen every entry through the AI. Together we mulled over the top Google results for AI Detectors, including GPTZero, Scribbr, and Writer’s AI content detector. The results were surprising: All but one of them falsely labeled human-generated content as either partially or completely AI generated. Humorously, we opted to take the Declaration of Independence and paste it in there. The detectors flagged it as being up to 94% AI-created!
We supposedly “hold these truths to be self-evident,” but even one of the most famous documents in American history isn't safe from being mislabeled as AI-written.
It should be self-evident (you see what I’m doing here? 😉), but AI-Detector tools are far from perfect and should not be relied on to determine authenticity. We talked last week about not avoiding or abusing AI, and that same balanced approach should be taken in using AI detectors. Over-reliance on AI detectors can cause unwarranted accusations, leading some college students to strategize ways to preemptively respond to a false accusation. Anecdotally I recall times in college a few decades back and having to run my papers through TurnItIn, frustrated at the inaccurate result.
But these tools wouldn’t exist if there wasn’t a demand for them.
We know that some students will just use AI to generate a paper – some of you commented last week expressing their shock as just how many students simply copy and paste articles or AI outputs and label it as their own original work.
So what do we do?
How do we navigate a world with rapidly evolving AI content, while the tools to identify them are still maturing? Here are a few tips:
Focus on quality, not detection: Rather than looking at the tool, read the document for yourself. Does it meet the standards for quality and content? Can you see if the author demonstrated any sort of critical thinking?
Be transparent and up front: If you don’t want AI to be used, say so. If you do allow it, consider a transparent and explicit approach – for example, if a student uses AI to generate titles or an outline but wrote the paper themselves, that distinction matters.
Use them as a tool, not a crutch: AI detector tools can be helpful, but only as a part of your evaluation. Any output from these tools should be taken into consideration and not be treated as definitive.
Stay informed: We know that these tools are evolving – in fact, in the several weeks between helping my friend and writing this article, the detector tools have improved, and most aren’t labeling the Declaration of Independence as egregiously AI as it did before. These tools will continue to evolve, and the topic of plagiarism is not new.
AI Detector tools are not the end-all answer to solve the issue of plagiarism but are rather one piece of the puzzle. We must approach the use of these tools with caution, focusing on quality and transparency!
My debut book “It’s Not the Tech - Leadership Lessons from the I.T. Field” is out now on paperback and eReader formats.