When I was teaching and dealing with the advent of widespread internet access (before the mobile phone era), I started each term with a new class by asking everyone to write about any topic for a whole class. These "essays" were not graded. Instead, I held onto them to compare student-submitted work with authentic samples of their writing styles and skills. Before anyone developed cheating detection software, I was routinely identifying cheaters because their submissions were always far more sophisticated that any authentic sample I had on file.
I just used ZeroGPT and pasted the beginning of the Declaration of Independence into it. It was 100% AI-generated. I guess Thomas Jefferson had help from AI!
From my experience, AI answers are usually "balanced" and "vague". It avoids extreme stance. The literature would contain too much decoration for a single sentence or scene.
Revision history wont do much if they wrote it on their phone and copied it into word afterwards. I agree these cheat-catching tools are more miss than hit
Yes, very true. These are all data points that teachers can use to understand how a student generated their assignment. None of them should be used in isolation.
I wouldn't trust these AI based AI detectors. Pulling students aside isn't as feasible for everyone -- for example, a professor at a community college, or even high school teacher, with over a hundred students you only see for short periods of time.
This introduces a whole new topic of conversation! If your class load is so large that you can't get to know and support your students, there are a lot of problems that will emerge.
Hey Sowash, (if for some reason you’ll read this 2 days after publication) I would like your opinion on something interesting. -So first, a disclaimer, this is the first video I’ve seen of yours so I’m literally unaware of any other opinions or content you’ve made.- Alright, I’m currently in college myself and I’m finishing up my spring semester soon. However, this semester I’ve ran into the most unique professor I think I’ve ever seen. Majority of his assignments are essays but he actually allows and even encourage the students to use a variety of ai models to help write their paper. He claims that the future involves ai so why try so hard to fight against it? The goal of the class is to prove you can understand the information and not to check if you can write an essay like you’ve been doing since public schooling. If the work with ai can improve your writing than the end result is your writing is better. But to truly test students if they actually understood the materials after you’ve turned in your paper is also a quiz that comes the next class period. It’s a short quiz with no more than 20 questions but it’s actually a majority of the points on the essay rubric. The quiz is simple with no trickery and basically asks the essential material that you would have already known if you had read the textbook and done your essay yourself. Some questions are literally “what was the cause of the event?” If a student failed this quiz section then the total number of points on the essay rubric is like artificially capped because the quiz is worth a large portion of the points. This means students who merely throw the prompt into an ai generator and learns nothing from the paper will fail the quiz and in result get a low grade on the essay assignment. So here’s the question. I’ve personally have found this to be a unique approach to the ever evolving world of ai and I believe ai will eventually become harder and harder to detect. The world runs on new implementation of technology to advance the classroom setting. We’ve seen computers with canvas be integrated in many schools when just a couple decade ago it was alot rarer with a sharp divide on opinion of its use. Do you think we’ll eventually see more professors actively allow the use of ai work along side student’s work like my professor has done? Or do you think schools and professors will continue to resist the change and future utilise more and more advance software to detect possible ai writing?
What your professor has done is not that unique as more and more educators are realizing that they need to update their assignments to align with the realities of the AI age that we're living in.
Thanks for posting such a thoughtful comment! I think the answer to your question is "both." There are always teachers that are reluctant to change and update their teaching methods based on new technology. I have been dealing with this for my entire career...it's one of the reasons I started this channel. But there are also many great teachers who understand that they need to change and evolve with society. It sounds like your college professor is in the second group. As others have commented, this is not a particularly novel way of teaching. I did something similar in my own classroom. The goal is to give students a variety of assessments so that even if they cheat on one it would be difficult to fake your way through all of them. I think you should tell your professor about this conversation and see what he says!
If/when you have the time, try running essays by students who are neurodivergent or those whose first language is something other than english and try those. It's been shown that AI plagiarism detectors have a bias for writing styles consistent with those two cases. Awesome video John!
You can check out part 1 here: th-cam.com/video/MUhByUDzdX4/w-d-xo.html
Spidey-sense of a teacher is built up over time. Worry for new teachers having to deal with this.
Fair point. Hopefully new teachers have some good mentors supporting them.
When I was teaching and dealing with the advent of widespread internet access (before the mobile phone era), I started each term with a new class by asking everyone to write about any topic for a whole class. These "essays" were not graded. Instead, I held onto them to compare student-submitted work with authentic samples of their writing styles and skills. Before anyone developed cheating detection software, I was routinely identifying cheaters because their submissions were always far more sophisticated that any authentic sample I had on file.
That's honestly such a clever idea.
Hey teacher! Leave the kids ALONE!!
You mean I should ignore the students in my classroom?
If it starts with "Certainly! I'd be happy to..."
😅
I just used ZeroGPT and pasted the beginning of the Declaration of Independence into it. It was 100% AI-generated. I guess Thomas Jefferson had help from AI!
The accuracy is significantly impacted by the length of the passage. The longer it is the more accurate the results will be
So just because someone is good at english, automatically means its A.I written? Seems like someone needs their license revoked 🤡
Not at all...my point is that these AI detectors are really inaccurate and shouldn't be used to assign guilt.
From my experience, AI answers are usually "balanced" and "vague". It avoids extreme stance. The literature would contain too much decoration for a single sentence or scene.
Yes, I agree. AI writing is very "Flat"
Revision history wont do much if they wrote it on their phone and copied it into word afterwards. I agree these cheat-catching tools are more miss than hit
Yes, very true. These are all data points that teachers can use to understand how a student generated their assignment. None of them should be used in isolation.
interesting
Just look for the words "tapestry, harmonious, and navigating"
Three words I use everyday...😂
@@jrsowash AI loves them
Thanks for the shout out! Very important and relevant topic.
I wouldn't trust these AI based AI detectors. Pulling students aside isn't as feasible for everyone -- for example, a professor at a community college, or even high school teacher, with over a hundred students you only see for short periods of time.
This introduces a whole new topic of conversation! If your class load is so large that you can't get to know and support your students, there are a lot of problems that will emerge.
as a student, ive already bypassed this and got a high mark on my paper, f the system
Did you get high marks because you worked hard and submitted your best work? Good for you.
Should we be catching students using AI? Or teaching them how to use it?
Depends on the assignment I suppose.
Hey Sowash, (if for some reason you’ll read this 2 days after publication) I would like your opinion on something interesting. -So first, a disclaimer, this is the first video I’ve seen of yours so I’m literally unaware of any other opinions or content you’ve made.-
Alright, I’m currently in college myself and I’m finishing up my spring semester soon. However, this semester I’ve ran into the most unique professor I think I’ve ever seen. Majority of his assignments are essays but he actually allows and even encourage the students to use a variety of ai models to help write their paper. He claims that the future involves ai so why try so hard to fight against it? The goal of the class is to prove you can understand the information and not to check if you can write an essay like you’ve been doing since public schooling. If the work with ai can improve your writing than the end result is your writing is better. But to truly test students if they actually understood the materials after you’ve turned in your paper is also a quiz that comes the next class period. It’s a short quiz with no more than 20 questions but it’s actually a majority of the points on the essay rubric. The quiz is simple with no trickery and basically asks the essential material that you would have already known if you had read the textbook and done your essay yourself. Some questions are literally “what was the cause of the event?” If a student failed this quiz section then the total number of points on the essay rubric is like artificially capped because the quiz is worth a large portion of the points. This means students who merely throw the prompt into an ai generator and learns nothing from the paper will fail the quiz and in result get a low grade on the essay assignment. So here’s the question. I’ve personally have found this to be a unique approach to the ever evolving world of ai and I believe ai will eventually become harder and harder to detect. The world runs on new implementation of technology to advance the classroom setting. We’ve seen computers with canvas be integrated in many schools when just a couple decade ago it was alot rarer with a sharp divide on opinion of its use. Do you think we’ll eventually see more professors actively allow the use of ai work along side student’s work like my professor has done? Or do you think schools and professors will continue to resist the change and future utilise more and more advance software to detect possible ai writing?
What your professor has done is not that unique as more and more educators are realizing that they need to update their assignments to align with the realities of the AI age that we're living in.
Thanks for posting such a thoughtful comment!
I think the answer to your question is "both."
There are always teachers that are reluctant to change and update their teaching methods based on new technology. I have been dealing with this for my entire career...it's one of the reasons I started this channel.
But there are also many great teachers who understand that they need to change and evolve with society.
It sounds like your college professor is in the second group. As others have commented, this is not a particularly novel way of teaching. I did something similar in my own classroom. The goal is to give students a variety of assessments so that even if they cheat on one it would be difficult to fake your way through all of them.
I think you should tell your professor about this conversation and see what he says!
@@jrsowash Cool, thank you for the response!
Isn't this just showing how to fool AI detectors?
It's not hard. In fact, they are wrong more than half the time.
Three words: Google Docs History.
If/when you have the time, try running essays by students who are neurodivergent or those whose first language is something other than english and try those. It's been shown that AI plagiarism detectors have a bias for writing styles consistent with those two cases.
Awesome video John!
I have read those reports. I don't have sample essays that I can test, but I would be interested in hearing the results from anyone who does!