How Can I Tell if My Child is Cheating with AI?
With the advent of generative AI (artificial intelligence), parents and educators have become increasingly concerned about students utilizing the technology to cheat on assignments. AI programs like ChatGPT are capable of not only giving answers to questions but also generating entire essays and even term papers in an instant.
While we all want to think the best about our children and presume their academic honesty, we cannot forget that human nature is what it is and that your child may be tempted to turn to AI to complete homework writing assignments. In this article, I will address some strategies to help you determine if your child has started using AI to do his or her writing.
AI Detectors
First, there are several AI content detectors available for free online. These are programs designed to detect the presence of generative AI text. Two of the most popular are Content at Scale and Scribber, but there are many more. You can paste a student’s writing into a field and run a scan on it. The program will then report the likelihood of the assignment being generated by AI.
There are a few caveats to these programs, however. First, they are not completely accurate. There have been documented cases of AI detection tools labeling human-created text as AI-generated (this article from Ars Technica has an excellent overview of why this happens and the various considerations surrounding AI detection). So, they cannot be entirely trusted. Furthermore, such programs often do not keep pace with the latest updates in AI software, increasing their margin of error.
You should, therefore, never make a case solely based on the use of an AI detector. Rather, you can use the AI detector as a starting point for further investigation.
Shortened Time Devoted to Writing
When I recently asked ChatGPT to generate essays for me on various historical subjects, it took about three seconds to generate the essay. Now I know kids are at different places with their writing. Some can write an essay in twenty minutes while others might struggle for well over an hour—but no human can plan and write an essay in three seconds!
If your child has started using AI to complete writing assignments, you can expect their homework to take notably less time than usual. Does your child who used to struggle for two hours with an essay now come down after ten minutes and say they are done? Have you noticed that you no longer see them studying, making notes, and composing drafts for their writing? These are signs they may have started relying on AI for their writing assignments.
Foreign Voice
When I caught a student using AI recently, I showed the paper to the student’s parent for review. The parent said, “This doesn’t sound like something my son would write.” This is an example of a writer’s “voice,” which is simply the stylistic element of an individual’s writing. It can include things like manners of speech, tone, style, word use, and mood. People have specific ways they tend to write. This style can be like a fingerprint identifying the author, sometimes literally. Ted Kaczynski, the famous Unabomber who eluded the FBI for twenty years, was captured when his brother recognized his prose style in an anonymously published manifesto and tipped off the authorities. When we know someone well, we tend to recognize their character in their writing.
AI programs like CHatGPT do not know your child’s vocabulary, tone, or any of the other personal indicators that make up their “voice.” Consequently, an AI-generated text will not sound like something your kid would write. You will not recognize your child’s “voice” in the writing. Now, we have to be careful here because most young people are still developing their writing style, so take that into account. But an AI-generated text will have nothing of your child’s voice in it, which should be noticeable.
Misaligned Vocabulary
In my experiment with ChatGPT, I found that it struggled to write at the appropriate grade level, despite my specific instructions to “write at a 7th-grade level” or “write at a high school level.” A child’s mastery of vocabulary is always very individualized, and an AI-generated text will often use words or phrases that your child would not normally use. In general (at least as of now), the AI seems to overestimate the average child’s vocabulary at a given age. Look, therefore, for words and phrases that you consider outside the range of your child’s knowledge, words that are misaligned with your child’s age and maturity.
I remember busting a student once because he said that a certain historical scholar was “one of the notable luminaries of the Carolingian epoch,” and I doubted that a 12-year-old used words like “luminaries” and “epoch.” I confronted him, and he fessed up.
Formulaic Structure
AI is programmed to deliver answers in an incredibly formulaic way. This is because it wants to cover all points that it is asked to in an organized, concise format. It may, therefore, feel too formulaic, “pre-packaged,” like the literary equivalent of Wal-Mart’s precooked bacon. Every point asked will be covered, generally with equal space divided for each. It will be very “symmetrical” in its structure. Real students, by contrast, tend not to be as balanced. Many students do not thoroughly answer a question or address every given point. Or, if they do, points are generally not addressed equally.
Students tend to write more on stronger points or things they know more about while pushing other points into the background or “fluffing” a subject if they don’t know as much about it. Sometimes, they do not write strong thesis sentences or solid conclusions. So keep an eye out for a structure that seems a little too formulaic.
Too Perfect
Related to the formulaic nature of AI writing is its freedom from error. Even the best student will occasionally make typos and grammatical errors, use superfluous words, use incorrect punctuation, repeat oneself, etc. AI-generated text will be free of these errors. Now, obviously, a student can reduce errors by proofreading and paying attention to detail. However, if there are never any errors of any kind, you are right to suspect something is up.
Lack of Examples
We must remember that when answering a question, all AI has to go on is your prompt. Its answers, therefore, will rely upon the information provided in your prompt and will generally omit anything not deducible from the prompt itself. Students, by contrast, have access to previously memorized, outside information not contained within the question. They are more likely to use examples and anecdotes to illustrate their points. Conversely, AI will be notably vague in this regard. A total lack of examples, quotes, or anecdotal stories could, therefore, suggest AI was involved.
Conclusion
These pointers should equip you to approach your child’s work critically, knowing the signs of AI involvement. “But,” you may ask, “aren’t there times when use of AI can be good and helpful for students? Can it be used ethically?” This is a fantastic question, which I will dive into in the next article!
What do you think about ChatGPT and AI? I invite you to join me and other Catholic homeschooling parents at our Homeschool Connections Community or our Facebook group today.
