#then frankly you have no position to be commenting on school practices based soley on your memory of being a student
Explore tagged Tumblr posts
Text
High school teacher but these are how I detect AI writing/ prevent it: One: Make a prompt that requires in class knowledge. Something along the lines of: "Explain how authors demonstrate character development using three examples from the short stories we read in class this week." Now- could the kids just look up what those stories were and edit the prompt? Yes. But it still weeds out a few. Two: Just like on a math test I make the kids show me their work. I break my essays up into paragraphs (I teach ninth grade so many of my students are learning how paragraphs are structured anyways so its good practice) I try not to teach cookie cutter five paragraph essays, but we will discuss in general and start off by writing our intro paragraphs together and even that process often gets broken down sentence by sentence. I wont grade essays until I see and grade the paragraph assignments, and its a big tip off to me that they had AI write the final essay if I cant find any correlation of their ideas from their brainistorming/chunked assignments to their final essays. Again, kids could get around this, but by this point ive weeded out most AI. What's more common is kids just copynig and pasting from these assignments onto their final essays and ignoring the editing phase lol. Three: Trojan horses. Hide a trojan horse in the prompt, search for it in the essay, pretty big give away. Just make sure the trojan horse is something not likely to show up organically in an essay. I usually use words like "banana" or unrelated nouns. This is the most common way I catch AI essays tbh. Four: Integrate AI into the lesson. One thing me and my kids did was use AI to write topic sentences, then we would rank, compare them to topic sentences from articles weve read and so on. For grammar once I had kids correct AI body paragraphs they generated and grade the AI on its use of syntax. It demystifies them, shows that often even if they do use AI and I dont catch it they probably arent going to be getting a great grade since more often then not we find these AI generated pieces to be mid at best. Five: Draftback. It's a google docs extension that lets me view a document being written in real time. Honestly this is the most full proof way, and its funny to watch kids make the typos then go back and erase them. I also catch a lot of plagiarism this way since I can see kids copying and pasting sentences then editing them to try and hide it. Six: I'll be honest, AI writing tends to just be kind of obvious. Most kids that I've caught with AI didn't even bother to change the font size/style so that tends to be a give away. And AI likes to write a lot of nonsense that means nothing and gets very repetitive. Sure a kid could hide this by reading the paper it produced and making corrections before sending it through the AI again, but at that point they are being critical enough of writing that I don't really care and most AI cheaters are frankly not that motivated. Seven: Just talk to the damn kid about their essay. If they can explain their thoughts, ideas, quotes, and so on then yeah they probably wrote it themselves. If a kid can get through al that and STILL I don't catch it, fuck it you get the grade you get. Congrats you made cheating harder than just writing the actual essay itself I hope you enjoyed it. So yes, teachers CAN catch AI, but using AI detectors isn't the way to go. And a good teacher uses these more comprehensive methods not just to catch AI, but it also safeguards kids from being falsely accused. For example I have english language learners that often use googel translate for help, or will copy and paste things that might trigger things like Draftback for example, but then if I can go back and be like "Oh I see they did the worksheets, hmm theres no trojan horses in here, and if I look at their brainstorming I can see their ideas developing from there to this final essay" then it also protects the student from being falsely accused and facing consequence's they dont deserve. Finally, for us teachers, it's important to accept we wont catch them all. You can lead a horse to water but they dont have to drink. We can give all the kids the tools to learn but if they still refuse then that's their choice. It is our duty to try, it is our duty to provide them all the accommodations and opportunities we can, but you cant control every factor of your classroom. And up until recently we actually had virtually no way to know if a kid was just cheating or copying. As my mother used to say: "Back in the 70s if I had an essay about Abe Lincoln I'd just go to the library, find a biography, and copy whole paragraphs. How as the teacher ever going to know they couldn't check against every book in the library." The era of us being 1000% a student wrote their work was an anomaly not the norm. Do I go through these factors every time? No. Usually you can get a pretty good idea when its time to start digging (like for example a kid who refused to do all the pre work for an essay suddenly turns in a ten page paper using the word acquiescence correctly in a sentence. Or if when I'm looking at an essay and draftback says that during the writing process the keyboard only was used for 170 keystrokes on a 1000 word essay. It's important we look for AI cheating in our students work, because AI writing (while its means are unethical thanks to scraping from works with out authors permission) is a tool. It's not going away anytime soon, and if were going to accuse kids of using it we need to have robust methods of detecting it that are HUMAN. Simply using an AI tool (which is what AI checkers are) to detect AI writing is hyprocasy. The most accurate AI checker is old fashioned human investigation.
I hate so much that professors who still can't figure out how to send messages on Zoom think they're capable of spotting AI writing. Professors are just feeding essays into AI detectors with massive fail rates with absolutely zero critical thought about the tools they're using. I moved across state lines. I've spent years of my life trying to get this degree. But at any moment I could be expelled because I got a false positive from a detector that tells you ChatGPT wrote Anna Karenina.
11K notes · View notes