Professor's invisible prompt trap catches 32 students cheating on their midterm with AI

[midian182]

Facepalm: Modern problems require modern solutions – according to the meme. In this case, the issue is students using AI to cheat on assignments. The solution, used to great effect by one professor, was to hide a white-colored font in the assignment prompt telling the AI to talk nonsense about an unrelated subject.

In an online discussion post, Alcorn State University history professor Dr. Jason Gibson posed a question that represented part of his students' midterm. It was about the Industrial Revolution, but knowing how many students now rely solely on artificial intelligence for their answers, he secretly embedded instructions telling the AI to include information about Madagascar in a way that makes no sense.

Being written in white text meant the instructions were invisible. Sure enough, 32 out of the 35 students across two classes simply copied the question, pasted it into a chatbot, then copy and pasted its output into the answer window.

Had any of these cheaters even bothered to check what the AI had generated, they'd likely have noticed the strange references to the island country constantly appearing in an essay about the Industrial Revolution.

In the viral TikTok post below, an exasperated Gibson said all the students he caught using AI failed this portion of their midterm.

@iamjasongibson I can't make this stuff up. #professor #highereducation #college #midterm ♬ Piano famous song Chopin Deep deep clear beauty - RYOpianoforte

In a follow-up video, Gibson showed some examples of how Madagascar was included in the answers. In one case, the AI talked about the Industrial Revolution before adding what seems to be a cryptic message one spy would whisper to another: "Madagascar floats sideways through the afternoon."

@iamjasongibson Part 3 #professor #highereducation #college #midterm #education ♬ original sound - Dr. Jason Gibson

In a similar vein, another answer talked about technology and social inequality before finishing with, "Madagascar purple bicycle whispers to the ceiling."

Gibson did say he wasn't using this as an opportunity to mock his students. He added that after he confronted them about the cheating, he gave them the opportunity to contest their grade. Only two did, though the fact that any of them tried to argue their case is pretty surprising.

Ever since ChatGPT went mainstream, students using AI to cheat has been a growing problem – it was considered endemic even back in August 2024. Teachers have tried to fight back using handwritten work, oral tests, and even AI to come up with questions that would be difficult for students to complete by simply feeding it into a large language model.

But it's a battle that humans aren't winning. Earlier this month, a professor at Brown University saw midterm averages jump from the typical range of 65% to 80% to 96% after he allowed take-home exams. He estimated that all but two of his 86 students used AI to cheat.

Permalink to story:

Professor's invisible prompt trap catches 32 students cheating on their midterm with AI

 

[Hexic]

Now this is funny, nice job by that teacher. Hopefully some kids were embarrassed enough to learn a lesson… but I doubt it.

 

[Neatfeatguy]

I guess the days are gone when pencil, paper and cursive where the way to write and turn in assignments. I'm not even that old, yet. Even back in the mid to late 90's a lot of my teachers required us to write out our papers and most of the time the work was done during a few days of class because we'd go to the library to find books via the card catalog, browsed the books organized by the Dewey Decimal System, browsing old newspapers on microfiche, making use of the encyclopedias and finally, checking out books!

Makes me wonder if any of these kids today even know what a library is or what exactly it is used for....

Good on the teacher making the kids look stupid, but bad on the teacher for allowing kids to simply do everything online.

 

[noel24]

Nice...

 

[Squid Surprise]

This is a brilliant move - it teaches the students HOW to use AI when working.

The lesson isn’t not to use it… it’s to use it properly. Now they know they need to read the question(s) before putting them into an AI - and read the answer(s) as well!

Blindly trusting an AI to do your assignment is foolish - you always need to proof read!

 

[Theinsanegamer]

I guess the days are gone when pencil, paper and cursive where the way to write and turn in assignments. I'm not even that old, yet. Even back in the mid to late 90's a lot of my teachers required us to write out our papers and most of the time the work was done during a few days of class because we'd go to the library to find books via the card catalog, browsed the books organized by the Dewey Decimal System, browsing old newspapers on microfiche, making use of the encyclopedias and finally, checking out books!

Makes me wonder if any of these kids today even know what a library is or what exactly it is used for....

Good on the teacher making the kids look stupid, but bad on the teacher for allowing kids to simply do everything online.

Cursive was totally, utterly, useless. There's a reason its not taught anymore.

I know in the 90s you had to walk uphill both ways, but today, everything is done online. When you work a job, anything you submit is going to be on a computer. The days of physically writing out tests is LONG gone.

Glad the teacher thought of this. Schools have got to get over their technophobia and adapt to modern tech, instead of trying to do things the way their great grandfathers did it.

 

[Dreadcthulhu]

I think more fields should have final competency exams (done in person, with both written sections & verbal interviews by an organization that is separate from the college system) similar to how the Bar exam works for the legal field. Would make trying to cheat your way through college (by AI or other means) a lot more risky, if you have to pass a through exam on your field without an easy way to cheat.

 

[thetimidwildone]

It's a conundrum... in school, you want the kids to learn the material and practice crafting ideas, but then once they get into the working world, management presses everybody to "use AI" for everything because they believe outsourcing human thought will save money.

We really are moving closer to Idiocracy/Wall-E every day.

 

[poohbear]

As a teacher, the reality is the only assessments I place higher value on are the 1:1 face to face assessments I do with students. I wasn't surprised when some of the students scoring very high on the written online assignments didn't have a clue about the subject matter when I did the 1:1 conversational assessments with them.

 

[airbornex]

Besides the obvious, I wanted to say students are making it too easy on teachers.
Sure, use AI, but at least read the answer and make it your own. Otherwise you're too lazy and deserve getting an F.

 

[Cyclone7d]

Does anyone remember the Scantron bubble sheets? Now that was foolproof or until the educator used the wrong key! (LOL) What does this say about our young adults who are choosing the wrong and easy way when nobody is watching them?

 

[Neatfeatguy]

Cursive was totally, utterly, useless. There's a reason its not taught anymore.

I know in the 90s you had to walk uphill both ways, but today, everything is done online. When you work a job, anything you submit is going to be on a computer. The days of physically writing out tests is LONG gone.

Glad the teacher thought of this. Schools have got to get over their technophobia and adapt to modern tech, instead of trying to do things the way their great grandfathers did it.

I wouldn't call cursive "totally, utterly, useless.", just sounds like you happen to hate it and have no use for it. Which is perfectly fine, I'm sure that applies to most people these days. I could understand one suggesting that cursive is no longer needed since the main form of communication is done electronically.

Actually, when I was in elementary school, I did walk both ways up hill. Walked downhill and then about halfway up the next hill to school, so when I left I walked down hill and bit and then all the way back up the hill to get home.

In my original post I wasn't claiming folks need to write in cursive again, they just need to get back to writing in school and get away from the using electronics for so much.There's been information starting to go around how the movement away from actual writing and focusing on electronic use is showing a decline in cognitive and academic development.

 

[Doofus McGoofus]

Makes me wonder if any of these kids today even know what a library is or what exactly it is used for....

As a Gen-Xer, stop with the generational superiority. If the kids today don't know what a library is, it's on their parents for never taking them.

 

[Fizmez]

Universities need to start teaching skills that can't be easily achieved with AI.

And that starts with the abolition of essays - an exercise in time wasting.

Instead, how about having the students research, gain knowledge, use AI if they like and then present live with a Q&A at the end?

 

[Aolish]

I'm speaking of myself also, we all better watch over our health from here on out, the next doctors, surgeons and rehab technicians are going to be the next generation. And it is these people that will be diagnosing the next diseases and fixing them. Doctors from previous generations are going to be a lot more intelligent (in my opinion), compared to the next doctors that are coming up now. Current doctors and surgeons didn't have that luxury of having AI and the internet from back in the days. They had to sweat blood to get where they're at, but this generation has got it way to easy with AI, social media and the internet. Can you imagine all the misinformation that is out there? We will be next up in line when we run into issues with our health. Your next heart surgeon that relied heavily on ChatGPT to pass his exams could very well be the one that will open you up and work on you. If that isn't scary enough as it is, I don't know what is. I'm not saying using AI to get better on your skillset is a bad thing, not at all. But there are some people who rely on it HEAVILY to the point where they wouldn't know what to do if they didn't have it. That's what I am worried about. Also I'm not getting political or anything, but combine with what is happening now with kids relying heavily on AI and the whole DEI thing when it comes to prioritizing diversity over meritocracy, your going to end up having a real problem where doctors won't know what they're doing 10-20 years later from now. People are going to start dying if this doesn't get fixed.

 

[GuitarJimbo]

The two kids who weren't caught cheating probably knew about "paste without formatting".

This isn't the first time I've heard of professors doing something to bury text so that the students wouldn't see it, but the AI bot would. Changing the font color is one way. Another is changing the font size to shrink the text into a point smaller than one pixel. AI bots are designed to expect copy/pasted text from the internet, so they will filter out any HTML formatting instructions as a matter of course. This makes it fairly easy for the professor to slip in text the student won't see, but the AI bot will. Any student who spends a few minutes a day reading news related to university study will already know this.

What the student wants is the ability to see the text, and only the text, being pasted into the prompt box. Most browsers on most operating systems can do this for you. On Windows and Linux you would press Ctrl-Shift-V to paste only plain text. On MacOS you'd press Cmd-Shift-V (or Cmd-Shift-Option-V). Alternatively, you might have a pop-up menu option like "paste without formatting" or "paste special". In the worst case, the old school way of accomplishing the same thing is to paste the text first into a simple text editor, manually edit out the HTML formatting code, and then copy the remaining text into the AI bot prompt box.

The students who weren't caught probably also used AI, but found the professor's AI trap and removed it before sending the prompt to the AI bot.

 

[GuitarJimbo]

Cursive was totally, utterly, useless. There's a reason its not taught anymore.

Cursive was, and still is, extraordinarily useful. When you have to put text on paper by hand then cursive is dramatically faster than writing block letters. Writing block letters requires lifting the pen from the paper after every stroke. Writing cursive only requires lifting the pen occasionally, such as crossing "t"s and dotting "I"s.

Personal computers have only been around for roughly 50 years. We didn't reach the point where at least half of western households had a computer until about 25 years ago. Prior to that, if you needed to make any text permanent (school assignment, for example) then you had to apply a pen to a sheet of paper. To anyone who still needs to do this even occasionally, cursive is still extremely useful. Even more useful in today's world is the ability to touch type; I.e., to type quickly without looking at the keyboard. Sadly, most students today can do neither.

 

[Neatfeatguy]

As a Gen-Xer, stop with the generational superiority. If the kids today don't know what a library is, it's on their parents for never taking them.

As a person that doesn't need to describe their self as some kind of "generation" to try and make a point, I do feel superior!

 

[acrossthestars]

Look, this is going to rub a lot of people the wrong way, but the facts don't lie: schools teach academic rigor and integrity, including avoiding plagiarism, but modern business teaches "make money, by any means necessary." Whether we want to admit or not, the "lesson" you learn from, watching all of these AI companies IPO is, "cheating works". Schools need to be able to explain to students why they are not allowed to cheat on tests and quizzes, when Anthropic successfully achieved $1 trillion valuation on the stock market, by stealing millennium's worth of intellectual property, scanning books and paintings indiscriminately, and then making money off the derivative productions their models confabulate. "If multi-billion dollar companies make money by breaking the law, why am I forced to obey it?" The answer, of course, is "money". "Rules don't apply to the wealthy", is the takeaway.

According to the outcomes of these corporations, "Plagiarism is only a problem, if you don't have enough money to fight it in court and WIN."