The Plagiarism Pandemic: How AI is Infecting Academic Integrity

As a college professor, I’ve noticed more and more assignments coming in that seem off — too polished for the student’s level, disjointed, or oddly formatted. Some are clearly entirely AI generated.

As a former detective and after a couple decades of clinical practice, I love gathering data and observing patterns. The second I notice an assignment with an odd flow, or over-sophisticated wording, my plagiarism senses start tingling. My trained eye picks up on the “symptoms” of AI writing — no human author crafts their words that smoothly after pulling an all-nighter shift in the ICU.

The implications for academic integrity are concerning. While AI can provide helpful writing support, submitting entirely generated content as original work crosses an ethical line. I’ll be far more impressed if you actually cite ChatGPT when you’ve used it.

While certain AI-generated text stands out clearly to my trained eye, I admit there are times it poses a tricky detection challenge. With continued advances in generative technology, AI writing can often pass as plausibly human. Unlike the over-sophisticated and disjointed essays of old, some cutting-edge AI can craft remarkably coherent arguments and fluid prose. If students take care to finesse and interpolate generated text, tracing its origins becomes quite complicated. Even experts cannot spot every instance or catch those who cleverly disguise the AI’s involvement. And of course, we must also allow some benefit of the doubt to earnest students.

In many cases, I simply cannot be 100% certain if an essay was human-written or not without a confession. But through diligence and experience, professors can still uncover the majority of issues. True excellence comes from a human spark, not artificial intelligence alone.

Automated AI Detection — Not there yet.

Some may suggest using AI writing detection services to analyze student papers for artificial generation. I have tested these systems using my own writing from over a decade ago, well before advanced AI was on anyone’s radar, and found concerning inaccuracies — flagging 50% of my genuine human-written work as AI generated.

While AI detectors may someday be a useful supplemental tool, at this stage I do not believe they are reliable enough to solely determine whether a student’s work is artificially produced. We cannot confront students about AI use based on a flawed algorithm alone. The subtleties of language and critical thinking require a human lens. My process relies on holistic human assessment, drawing on years of expertise in student writing. No automated tool can replace the insights I have accumulated through extensive experience in the classroom. For now, AI detectors provide an interesting data point but should not drive high-stakes conclusions.

The key to my strategy is pattern recognition. After reviewing thousands of pages of AI-generated text and countless student papers each year, I’ve honed my skills in detecting AI assistance. Educators should be able to distinguish when a student has used generative AI instead of their own effort. Here are the red flags I watch for:

Drastic Changes in Writing Style

When a previously average student suddenly writes like a postmodern Shakespeare, I sense some AI trickery afoot.

Don’t try slipping an elegant, persuasive analysis under my door when all your emails are written in monosyllabic text-speak. If a student has turned in previous assignments written in their own voice, a paper with drastically different style raises suspicions. Does it have a more sophisticated vocabulary than past work? Is the tone completely off base for that student? A big swing in quality or style from known writing samples suggests AI generation rather than the student’s own skills improving.

Lack of Consistent Flow

I’ve read enough rambling, directionless papers to know strings of saturated jargon don’t compensate for poor critical thinking.

AI systems are great at mimicking styles on a paragraph level, but often struggle with cohesive narrative flow between sections. Papers written by an AI tend to jump between topics and tones. Arguments go off the rails easily without a human hand guiding the overall structure. If your thesis statement and conclusion seem to be from two totally different papers, you can’t blame me for thinking an AI wrote this.

Over-Reliance on Sources

A good rule of thumb: if the references section is longer than the analysis, you’ve got an AI phoning it in.

Students know to include quotes and references to back up their analysis. But an AI-generated paper often leans too heavily on outside sources at the expense of original argumentation. If something reads more like a compilation of abstracts instead of a unique synthesis, it could be the work of an AI scraping text from the web. Real students have to pad things out a bit more to hit their page minimums. If the number of quotes in your paper exceeds the word count, you can bet it was written by AI.

Odd Errors and Word Choice

I’m sorry Pauline, gonorrhea outbreaks really don’t weave an intricate tapestry.

Repeating the same distinctive phrases across paragraphs, revealing the AI’s limited vocabulary and ability to rephrase ideas. Occasional grammar hiccups are expected, but truly unnatural phrasing is a giveaway of AI writing. Does a paper have contradictions, abrupt topic shifts, or words used inaccurately? Proprietary AI models make mistakes no student would. Also, overuse of obscure vocabulary words incorrectly, as the AI tries too hard to sound academic, e.g. “The physician’s egregious diagnosis was deleterious and mercurial.”

While generative AI holds exciting potential to augment education, allowing it to replace the learning process defeats its purpose. By looking for these clues, I hope to guide students back to meaningful writing and research skills that will serve them beyond the classroom. Academic integrity remains fundamental, even in this time of rapid technological change. Authentic effort still matters.

Could you spot an AI-written paper? Are your institutions implementing policies to combat this?