How would that work in a public school setting? Seems like there are too many kids and not enough teachers/time to implement this across all subjects.
If you are right, and oral exams become the best way to evaluate student learning, then I could see smaller private schools becoming more popular as a correlation.
> One complained to the AP about having to repeatedly rewrite their papers so it doesn't get flagged as being AI-generated. And that, we have to say, does suck.
I've had to rewrite the text of a help dialog so that my app didn't get flagged on VirusTotal. Feels kind of the same.
My favorite solution to this is to embrace the technology and have students really learn its uses and limitations.
Tell the students to prompt their LLM of choice to write an essay on a given topic or set of topics. Have the students include the prompt as well as the generated response. The actual assignment is for the students to then edit and critique the LLM-generated essay. This both requires them to demonstrate an actual expertise in the subject matter and directly shows them the limitations of LLMs.
It's the tragedy of the commons, but students using AI are only ruining their educational experience.
The age of computers has brought easy word processing and portable research tools that give you incredible freedom when writing papers or doing homework. No need to purchase book after book or hang out in the library all day if you're doing research.
Now with AI making automated plagiarism trivial, the system will need to change. I believe this can very well be the start of the return of pen and paper and non-digital essays under the watchful eyes of teachers.
Nobody wants such changes to happen, neither the educational staff nor the students, but I think it's inevitable.
AI is like calculators. They're an amazing tool that's essential for certain tasks, like actually applying higher algebra, but they're not a replacement of the basic skills most course and subjects are trying to teach.
Allowing the use of generative AI on essays is like giving primary school kids calculators to teach them arithmetic: the tool takes the place of the skills you're being taught. You can use all the spell and gammer checking AI you want on an essay (unless you're doing an English essay specifically to test your grammatical skills, of course) but don't grab a calculator when you're being taught how to do basic multiplication!
Sounds good to me, we just need to start teaching handwriting better. Have you seen handwriting from kids these days? They can't even read their own writing.
How about requiring less essays but have the essays be better? You know, on the level above GPT.
I see the problem of GPT as a hybrid of two problems:
1. The biggest problem here is not GPT - it's that human work has often been so bad, that a system with an intellect of a retarded cockroach (as Michio Kaku puts it) was able to displace it. People are so used to bad writing, that they don't recognize GPT nonsense.
2. GPT is largely another web search engine, and so it's an old problem repackaged: searching the web and copy-pasting what has been already said, which can be of high quality - maybe there is a solution to that, like being more specific?
> the system will need to change. I believe this can very well be the start of the return of pen and paper and non-digital essays under the watchful eyes of teachers.
Nah, in reality it'll just result in sales of heavily locked-down test-taking kiosks with all manner of UBA enforcement to school districts by shady vendors. The solution to technological problems is only ever more technology.
My solution would be to fuzz the students into producing hallucinated answers. Bullshit the bullshitters-- ask them questions that are impossible to answer and see which ones complete the assignment. Anybody who completed the assignment fails. The ones that couldn't finish, you give partial credit based on their effort, and grade on a curve. We were never supposed to be testing students' ability to regurgitate trivia. They're not training to be Jeopardy contestants.
Can you explain how this would be a "tragedy of the commons"? I'm not sure I understand how there's a limited resource which each individual is incentivized to extract more than their sustainable share of.
I am not surprised. Lots of jobs are going to be done by AI very soon, and it's unlikely that a human will be able to get a job without pretending to be an AI. So it looks like these students are well-prepared for the future.
Students primarily don't study for jobs. They study for baseline competency across a broad range of skills. Writing is not just important for the finished product (a written document) but it's also useful to elaborate on and evaluate a person's thought processes, such as the ability to comprehend and analyze information. Students who fail to learn to write are at a significant disadvantage to their peers who succeed.
Leadership is the art of getting someone to do something because they want to do it, if I might steal from Eisenhower.
As a college instructor, I've spent a of time thinking about how to get students to not want to cheat versus detecting cheating. (I feel the latter is a losing game for me.)
One online school I worked for initially had no grades, back in their early days. There was literally zero incentive to cheat. You had to do the hard work or you couldn't get a job, end of story. It was beautiful, and worked great.
Where I work now requires grades, and I don't have handy job interviewers to act as my graduation gate.
You had to do the hard work or you couldn't get a job, end of story.
Sounds great, but I feel like there's a piece I'm not seeing there. Your school had a stable of interviewers who wanted specifically to work with your graduates? They effectively served as the ultimate graders, assessing how much your students learned? Presumably with your personal recommendation?
That's awesome. I wish more places would hire based on an in-depth understanding of the candidate's capabilities. That's a hard skill, but potentially yields a lot of "bargains" by finding new people with potential but not demonstrable experience.
It wasn't a stable of interviewers that the school had (though later we partnered with Triplebyte for a bit, IIRC)--it was just the real life interviewers we match-made with for our students looking for work. Trial by fire.
The students were motivated to learn the material in part because they weren't competing against the school (e.g. for higher grades). They knew we were just a tool to help them learn good skills as rapidly as possible so they could get paid in a job as rapidly as possible. There was no concept of "I have to get good grades and that will translate to a good job". There were no grades. They had their eye on the interview and the skills, period.
The idiots getting caught cheating should be instantly kicked out of the class and possibly the school, but what the hell is this going to accomplish?
> MacKellar, fearing plagiarized programming, requires her intro course students to write their code on paper — which seems like a pretty Draconian measure, given the subject matter.
Unless they're writing it in front of her in class . . . .
35 comments
[ 4.6 ms ] story [ 330 ms ] threadIf you are right, and oral exams become the best way to evaluate student learning, then I could see smaller private schools becoming more popular as a correlation.
If you are doing it in a controlled environment (i.e., classroom), there’s no reason for it to be oral, written works just fine.
https://pointsixtyfive.com/xenforo/threads/overview-of-initi...
(see Day 19)
It's black-mirror-ish, but I'm thinking maybe have it somehow give small zaps to the muscles in your wrist.
I've had to rewrite the text of a help dialog so that my app didn't get flagged on VirusTotal. Feels kind of the same.
Tell the students to prompt their LLM of choice to write an essay on a given topic or set of topics. Have the students include the prompt as well as the generated response. The actual assignment is for the students to then edit and critique the LLM-generated essay. This both requires them to demonstrate an actual expertise in the subject matter and directly shows them the limitations of LLMs.
The age of computers has brought easy word processing and portable research tools that give you incredible freedom when writing papers or doing homework. No need to purchase book after book or hang out in the library all day if you're doing research.
Now with AI making automated plagiarism trivial, the system will need to change. I believe this can very well be the start of the return of pen and paper and non-digital essays under the watchful eyes of teachers.
Nobody wants such changes to happen, neither the educational staff nor the students, but I think it's inevitable.
AI is like calculators. They're an amazing tool that's essential for certain tasks, like actually applying higher algebra, but they're not a replacement of the basic skills most course and subjects are trying to teach.
Allowing the use of generative AI on essays is like giving primary school kids calculators to teach them arithmetic: the tool takes the place of the skills you're being taught. You can use all the spell and gammer checking AI you want on an essay (unless you're doing an English essay specifically to test your grammatical skills, of course) but don't grab a calculator when you're being taught how to do basic multiplication!
Sounds good to me, we just need to start teaching handwriting better. Have you seen handwriting from kids these days? They can't even read their own writing.
How about requiring less essays but have the essays be better? You know, on the level above GPT.
I see the problem of GPT as a hybrid of two problems:
1. The biggest problem here is not GPT - it's that human work has often been so bad, that a system with an intellect of a retarded cockroach (as Michio Kaku puts it) was able to displace it. People are so used to bad writing, that they don't recognize GPT nonsense.
2. GPT is largely another web search engine, and so it's an old problem repackaged: searching the web and copy-pasting what has been already said, which can be of high quality - maybe there is a solution to that, like being more specific?
I wish the answer to this would be to improve education so that it interests people, but it's not gonna happen
Nah, in reality it'll just result in sales of heavily locked-down test-taking kiosks with all manner of UBA enforcement to school districts by shady vendors. The solution to technological problems is only ever more technology.
My solution would be to fuzz the students into producing hallucinated answers. Bullshit the bullshitters-- ask them questions that are impossible to answer and see which ones complete the assignment. Anybody who completed the assignment fails. The ones that couldn't finish, you give partial credit based on their effort, and grade on a curve. We were never supposed to be testing students' ability to regurgitate trivia. They're not training to be Jeopardy contestants.
I see all these claims of "AI will take our jobs!" but nobody can enumerate the actual jobs that will be lost.
To be fair, they link to the source.
As a college instructor, I've spent a of time thinking about how to get students to not want to cheat versus detecting cheating. (I feel the latter is a losing game for me.)
One online school I worked for initially had no grades, back in their early days. There was literally zero incentive to cheat. You had to do the hard work or you couldn't get a job, end of story. It was beautiful, and worked great.
Where I work now requires grades, and I don't have handy job interviewers to act as my graduation gate.
It's a struggle, for sure.
Sounds great, but I feel like there's a piece I'm not seeing there. Your school had a stable of interviewers who wanted specifically to work with your graduates? They effectively served as the ultimate graders, assessing how much your students learned? Presumably with your personal recommendation?
That's awesome. I wish more places would hire based on an in-depth understanding of the candidate's capabilities. That's a hard skill, but potentially yields a lot of "bargains" by finding new people with potential but not demonstrable experience.
The students were motivated to learn the material in part because they weren't competing against the school (e.g. for higher grades). They knew we were just a tool to help them learn good skills as rapidly as possible so they could get paid in a job as rapidly as possible. There was no concept of "I have to get good grades and that will translate to a good job". There were no grades. They had their eye on the interview and the skills, period.
> MacKellar, fearing plagiarized programming, requires her intro course students to write their code on paper — which seems like a pretty Draconian measure, given the subject matter.
Unless they're writing it in front of her in class . . . .