My son had his portion of a group assignment flagged as "30% likely to be AI".
But it wasn't. I literally watched him type it and turn it in.
WWGRMD if this happened to you?
I've told him to talk to the professor but this particular course is online only and the office hours are very limited. It seems to me like something important enough for a face to face.. not an email.
Can screening tools (that look for AI) provide "proof"?
In reply to OHSCrifle :
So it was 70% likely to be human right?
Those tools are known to be notoriously bad. So bad, in fact, that OpenAI pulled theirs from the market because the false positive rate was so high and they caught so much E36 M3 for it. TurnItIn (god berkeley this company) is on record with WaPo saying that their tool really isn't that great. Oh, and also it might be impossible to create one.
If the professor makes an issue of it, I'd say something. If not, I'd leave it alone.
The models work by data fitting, which just means they primarily focus on the randomness of word choices and "correctness" of the grammar. That's at best problematic. At worst it is highly discriminatory against non-native English speakers.
Well he was given a zero initially but the professor subsequently sent out a note saying "if you think this is incorrect please contact me".. so he did.
Here are some references for him then:
https://www.washingtonpost.com/technology/2023/06/02/turnitin-ai-cheating-detector-accuracy/
Don't be the RollingStone guy. It looks bad for everyone.
Find what tool the prof uses, run some of his papers through it and see what grade he gets. Be prepared to take it to the dean
Those tools are garbage. I've fed them papers written by AI and papers written 25 years ago. There's zero correlation. They'll give different percent odds on the same paper checked two different times. I've had the percent change drastically based on changing one word (out of thousands) or nothing at all.
They shouldn't be used as a measure of anything and any teacher using them is only revealing their ignorance.
I'm dealing with this with my students now. There are a few things that can complicate the diagnosis, but ultimately it comes down to knowing your students. In addition to the obvious problem of students putting a prompt into an AI generator and having it pump out an essay, things like Grammarly and Babbel (mostly an issue for ESL students) can trip the detectors as well. I warn students about all of these things, and give them links to the detectors so they can check for themselves before submission. If something throws flags, I run it through multiple detectors, and if it looks fishy, I withhold a grade until I can speak with the student.
In almost every case I've dealt with, my initial instinct was right. The diligent students clearly wrote their essays (it only takes a few questions to figure out if they know what they wrote), and the slackers clearly didn't.
Apparently this school is using TurnItIn. Probably has the checker built in to their "college" portal where syllabi and assignments get are all hosted and posted.
To be continued.
I think one of the problems you will end up with it is hard to Un-Ai something ,
So if you ask Ai to do a report for you so you get some ideas , even if they write it themselves later , they cannot un-ring the bell and there will be things they copy from Ai without even thinking about it ,
Its a hard problem to solve ,
It's berkeleying English. There is only so many ways to write, "This course is horseE36 M3." berkeleying write it in Spanish or French then google translate it.
We use SafeAssign, but it's the same concept. I routinely get papers flagged by it for plagiarism, but when I look at them I see that the student has simply given the same (correct) citations for a book or other source I assigned, and the institutional database is flagging it because other students in my class use the same source. They're just tools, and they're only as good as the person using them. Hopefully the professor in question is actually paying attention.
My only limited experience with AI so far is with Microsoft copilot/designer and if how that works is typical, I would never let it write a term paper. Don't term papers have to have references and footnotes and such?
In reply to VolvoHeretic :
Normally, yes, though the standard of undergraduate instruction is hardly uniform. There are also assignments like brief reaction statements or other short writing work that typically is not cited (part of a recent push to include lots of low-stakes assignments in an effort to get unprepared students up to speed without failing courses, as well as actually doing the work assigned, which is more of an issue than you might imagine). My practice is not only to require sources and citation, but for survey classes to provide those sources; this makes it far more difficult to plagiarize or apply AI. In upper-level courses, I work with one of our librarians to ensure students know how to do research and cite their sources properly, since in a fair number of cases I've been told by students that they employed academically dishonest methods because they didn't know how to do it themselves, which is a sad commentary on the aforementioned current standard of undergraduate instruction.
so it begins
In reply to 02Pilot :
Like trying to write a piece of uninteresting bs fluff papers in APA format....
In reply to Ranger50 :
Not quite sure I understand what you're getting at.
In reply to 02Pilot :
Outside of a quick 5-10min "discussion" and a equally E36 M3ty handout, you have to write out 3-25 page papers over topics that interest no one in a format few know about.
I wish someone would do a better job at explaining it. Getting a substandard grade on a decently written paper because you didn't cite something "textbook" is insane. APA format is the worst.
Besides the strict grading guidelines/rubic, 80% is a "C", APA format is why I don't pursue my BSN. It's nothing but busy work.
In reply to Ranger50 :
I find that when I'm writing proposals for prospects or training documents, citing my sources and using APA or MLA is something that is extremely important.
Oh. No, it's the exact opposite of that.
When I went to college back in the 70s I majored for one year in geology enrolled in what they call The Experimental College created by the geology professor but which covered all of the under graduate college general classes.
We would have to go to the library, find a subject to research, write a term paper, hand it to the professor and all of the students who would then have to research your paper so that after you gave your 1/2 hour lecture on your subject, they would ask you questions for another 1/2 hour and if they didn't do a good enough job of showing your ignorance on your topic, the professor sure could.
Too bad all of those credits only lasted if you stayed in that school till you graduated, but the second you transferred somewhere else, they where discredited and useless.
I really did learn to use and like libraries which worked out well after giving up higher education because I couldn't figure out what I wanted to do for the rest of my life and worked construction throughout the western US during the winter and for something to do, I would hang out in the public and University libraries of whatever city I was in. What a geek.
Too bad I somehow managed to get through high school and 4 or 5 years of college without ever taking a grammar class that I remember. When in doubt, punctuate. 
Maybe that's how everybody is getting caught cheating with AI, their punctuation is too perfect?
Somehow, that sure felt like I just wrote a term paper. 
In reply to Ranger50 :
I'm not sure how much of your dissatisfaction is specifically directed at me, or just a sort of general set of assumptions. If it's the former, you misunderstand how I run my classes. I don't do handouts, discussions can go for an hour or more, and essays are assigned on topics that are either specific and relevant to the subject or (in upper level seminars) on topics of the students' choosing. Citation matters simply because it is important for students to understand the role that evidence plays in the construction of an argument, and that making unsupported statements does not constitute valid support for a thesis. Learning to utilize various style of citation (my disciplines use Chicago, not APA or MLA) is just part of the process, the same as learning the appropriate format or structure for any sort of written work. Grade breakdowns are clearly provided in a rubric that students have before submission; citations account for 5% of the available points.
The simple fact is that there is no real standard for instruction at the college level. On the one hand, this is theoretically beneficial, as it allows good instructors extensive leeway to build their courses as they see fit. But with limited oversight, full-time faculty at research institutions that are often more interested in their own work than in teaching, and large numbers of underpaid contingent faculty at smaller schools, the ideal is all too rarely achieved. This, combined with the serious and widespread misapprehension that a liberal arts education is intended to be job training, frequently leads to disappointment and frustration on the part of both students and faculty.
In reply to 02Pilot :
You completely missed that what I listed is what I got for instruction vs what you listed I wish I received.
In reply to Ranger50 :
I don't see where you ever mentioned that you were relating your own experience.
In reply to VolvoHeretic :
That sort of instruction tends to work very well for students who are willing to put in the time and effort, which is not all of them. I use something vaguely similar in my upper-level seminars, where it tends to be successful because the students have already made it through a couple years and have proven a commitment. Because colleges and universities are increasingly dominated by administrators, who are concerned first and foremost with enrollment, retention, and graduation rates - all of which have only a passing relationship (no pun intended) with quality of instruction - there is more and more pressure to adopt methods that keep students coming back semester after semester and graduate them on time (more or less). This has led to a vehemently-denied but nonetheless obvious lowering of standards, particularly in freshman classes. We now have mandatory progress reporting to whole departments of counselors who constantly keep tabs on students and remind them to do their work, which inevitably keeps passing rates higher but teaches them nothing about responsibility. At the same time, there are now thresholds for "DFWI" rates (D/F/withdrawal/incomplete), above which we have to "make adjustments" to our methods of instruction, which basically means dumb it down so more students can pass, irrespective of whether the poor outcomes had anything to do with instruction; I got flagged one semester when I had a class that had otherwise normal rates, but had several students who never attended class at all but stayed registered, and two who had to withdraw for medical reasons (one personal, one family).
My high school chemistry teacher had a bumper sticker on his El Camino that said it all: "When all else fails, lower your standards".
02Pilot said:I'm dealing with this with my students now. There are a few things that can complicate the diagnosis, but ultimately it comes down to knowing your students. In addition to the obvious problem of students putting a prompt into an AI generator and having it pump out an essay, things like Grammarly and Babbel (mostly an issue for ESL students) can trip the detectors as well. I warn students about all of these things, and give them links to the detectors so they can check for themselves before submission. If something throws flags, I run it through multiple detectors, and if it looks fishy, I withhold a grade until I can speak with the student.
In almost every case I've dealt with, my initial instinct was right. The diligent students clearly wrote their essays (it only takes a few questions to figure out if they know what they wrote), and the slackers clearly didn't.
My wife is teaching a 102 level English class this semester, and recently had to send out a "Some of you are using AI to write your papers and you may think you're clever but you're not" email to the class. Like you said, if you know your students it's pretty easy to tell who is cheating. Robot writing is pretty easy to detect just by reading it in most cases.
You'll need to log in to post.
