OMFG. OpenAI's Educator Considerations for ChatGPT

[highplainsdem]

I'd never seen these till today. Learned of this page via an NBC New York article on ChatGPT writing news stories with invented quotes.

A few excerpts from OpenAI's "educator considerations":

https://platform.openai.com/docs/chatgpt-education/educator-considerations-for-chatgpt

We recognize that many school districts and higher education institutions do not currently account for generative AI in their policies on academic dishonesty. We also understand that many students have used these tools for assignments without disclosing their use of AI. Each institution will address these gaps in a way and on a timeline that makes sense for their educators and students. We do however caution taking punitive measures against students for using these technologies if proper expectations are not set ahead of time for what users are or are not allowed.

Translation: We know students will cheat using ChatGPT. Deal with it. But don't punish kids for cheating unless you've given the students exact anti-cheating guidelines first since you should expect them to try to cheat using our handy cheating tool.

Classifiers such as the OpenAI AI text classifier can be helpful in detecting AI-generated content, but they are far from foolproof. These tools will produce both false negatives, where they don’t identify AI-generated content as such, and false positives, where they flag human-written content as AI-generated. Additionally, students may quickly learn how to evade detection by modifying some words or clauses in generated content. In addition, the OpenAI AI text classifier is narrow in scope, and not a tool for detection of other things such as plagiarism from, e.g. the use of copied text from the internet or other sources.

Translation: You're on your own trying to figure out if they're cheatjng with ChatGPT. Have fun!

While tools like ChatGPT can often generate answers that sound reasonable, they can not be relied upon to be accurate consistently or across every domain. Sometimes the model will offer an argument that doesn't make sense or is wrong. Other times it may fabricate source names, direct quotations, citations, and other details. Additionally, across some topics the model may distort the truth – for example, by asserting there is one answer when there isn't or by misrepresenting the relative strength of two opposing arguments. For these reasons, it's crucial that students know how to evaluate the trustworthiness of information using external, reliable sources.

Translation: ChatGPT is a very convincing bullshitter. Better hope your students have the knowledge to catch all its lies, or the time and willingness to fact-check everything it tells them.

ChatGPT may produce content that perpetuates harmful biases and stereotypes, sometimes in subtle ways. This includes generating biased or stereotypical portrayals of groups of people, which can be harmful, particularly in a context where those biases are being taught, learned or otherwise reinforced. The model is generally skewed towards content that reflects Western perspectives and people. One example of this is that the models perform best in English, and some measures that we have taken to prevent harmful content have only been evaluated in English. The dialogue nature of the model also has the potential to introduce or reinforce user biases and preferences over the course of interaction with the model.

Translation: Oh, and it can be a bigot, too. A creative, convincing bigot. Deal with it.


In the past, if someone had shown up near school grounds to sell or freely offer students an almost undetectable way to cheat that would also spread lies and bigotry, they probably would NOT have been welcomed.

There would have been outrage from educators and parents alike.

OpenAI is getting away with this, so far, because they also introduced ChatGPT as a free toy to entertain adults by supposedly making them "creative" as the chatbot writes things for them. A free toy that might also be able to save them some time and effort in their work. A free toy that might even let businesses employing writers get rid of some or all of the writers they're paying if they're fine with publishing bullshit articles.

So even some adults who can see what a disaster this is, in terms of the harm done to education, still would like to play with ChatGPT themselves. And might be willing to pay monthly fees for it, once they're hooked, which was the plan all along with this free release.

I have to wonder how much whoever wrote these "educator considerations" was laughing.

[ColinC]

1. Ummm I'm pretty sure "don't plagiarize" covers it.

Using somebody or something else’s work and passing it off as your own is entirely covered by nearly every institution already. 🤷‍♂️

[cbabe]

2. Reminds me of when Wikipedia as a reliable

source was outlawed. Anyone can post. Posts not verified.

I remember Middlebury being first:

https://www.historians.org › research-and-publications › perspectives-on-history › january-2007 › wikipedia-banned-by-middlebury-college-for-history-students

Wikipedia Banned by Middlebury College for History Students

Wikipedia Banned by Middlebury College for History Students Vernon Horn | Jan 26, 2007 Students at Middlebury College will no longer be able to cite Wikipedia when writing history papers, according to an Inside Higher Education report.

[hunter]

20. How is "anyone can post" different from the the larger world?

What's the difference between Wikipedia and, say, MSNBC?

[cbabe]

23. Drawing the line on verifiable sources can be

tricky.

An immigrant student asked me about using an article on the Harvard website.

I said: it’s Harvard so it’s probably ok.

He replied: what’s Harvard?

[hunter]

29. I'm a University of California guy.

Anything those Harvard guys say is suspect.

[cbabe]

32. All kinds of universities and profs receive private funds, making their

research questionable. Need to always dig a little deeper.

https://www.huffpost.com › entry › ray-hilborn-funding_n_57365012e4b077d4d6f33238

Prominent Fisheries Scientist Under Fire For Seafood ... - HuffPost

The $3.56 million that Hilborn received from industry groups from 2003 to 2015 is just 22 percent of all the funding the scientist brought to the university, according to the Seattle Times. Hilborn's funding sources included companies like Trident Seafoods and Peter Pan Seafoods, as well as the industry group National Fisheries Institute.

…

https://publicintegrity.org › politics › why-the-koch-brothers-find-higher-education-worth-their-money

Why the Koch brothers find higher education worth their money
Essentially, to inculcate the next generation with a philosophy like their own: As we've reported throughout this decade, education grants from the Koch brothers' network of foundations routinely support academic programs or centers that teach theories and principles aligned with the Kochs' convictions about economics and public policy.

[Scrivener7]

3. It will soon be that assignments will need to be written in class, and classrooms

will not be able to allow computers.

Really, we are hitting a turning point.

The great source of information, i.e., the internet is becoming useless as a source of information.

The great teaching tool, i.e., computers are becoming detrimental to children's learning.

The great simplifier of financial transactions is becoming too susceptible to fraud to be of any use in financial transactions.

Sharpen your pencils, folks. We're going to be needing them. And that's not necessarily a bad thing.

[edisdead]

4. That isn't going to happen.

[Scrivener7]

5. What isn't going to happen?

[edisdead]

6. In class only pencil/pen writing.

[Scrivener7]

7. Who said it was?

[edisdead]

8. You did.

In the post I replied to.




Star Member Scrivener7
3. It will soon be that assignments will need to be written in class, and classrooms
will not be able to allow computers.

Really, we are hitting a turning point.

The great source of information, i.e., the internet is becoming useless as a source of information.

The great teaching tool, i.e., computers are becoming detrimental to children's learning.

The great simplifier of financial transactions is becoming too susceptible to fraud to be of any use in financial transactions.

Sharpen your pencils, folks. We're going to be needing them. And that's not necessarily a bad thing.

[Scrivener7]

9. OK. If that's what you took from the post.

[edisdead]

22. I am not trying to offend you

It is actually what you wrote. Not sure what else to take from it.

[Ms. Toad]

16. You might be surprised.

Starting at least a decade ago, at the college/grad school level there has been a slowly rowing movement to ban computers from the classroom.

The argument is that computers encourage students to trasnscribe lectures, rather than synthesize the information being conveyed. Students who wrote were forced to process the information and synthesize it in smaller chunks as they were receiving because it was impossible to transcribe it using pen and pencil. In addition, many of those who aren't transcribing are surfing the net.

I've taught alongside at least one professor who banned computers from their law school classroom.

Personally, as a student, I found the opposite. I took typing in 7th grade, and can type as accurately and as quickly as any administrative assistant I've ever had). This speed allowed me to synthesize the information and organize my notes in a way that I never could with handwriting. But I was in law school (with a bachelor and masters degree under my belt) - and had had years of practice synthesizing informaiton to the best I could with the slow (handwriting) tools I had. So I took the synthesizing skills I already had and used my computer to implement them more effectively.

But as a small counter - I am back in the classroom as a student - in a class in which I am actively transcribing (computer programs being written on the fly by an instructor). During those sessions, I am learning less than the ones in which the professor is teaching - so there is some truth to the concern about transcription rather than synthesis.

ChatGPT may accelerate that largely static movement.

[highplainsdem]

12. I wouldn't bet it won't. Saw an article in a British paper a few

weeks ago on a school planning to shift homework to reading and research, with class time shifted to assessing how much the students have learned, since they won't be able to assess them properly otherwise. So it could well mean more homework and then pop quizzes, in whatever form, almost every day. They don't think all kids will cheat if they take assignments like essays home, but they know some will with ChatGPT.

[andym]

13. In class essays are a good idea: They can have/bring in computers and tablets to write them

no need for pencils

[Scrivener7]

14. Exactly. Schools will need to move to where assignments are done

where they can be monitored against use of AI.

[friend of a friend]

18. In junior high I would have loved using a pencil instead of a fountain pen and ink well.

[Thomas Hurt]

10. This is how Skynet begins, with its ancestors spreading disinformation.

[Prairie_Seagull]

11. "Sharpen you pencils' Need to teach cursive again first. I argued against It's stoppage in the-

Seattle School district. Lost of course but I tried.

The onus should be on OpenAI to figure it out. For instance: they could fairly easily make the output have ChatGPT printed in light gray (watermarking) around all text. They should have to fight this battle. Not educators.

[Scrivener7]

15. Lol! True about the cursive.

About the watermark, that is a fabulous idea, but it would reduce the use of ChatGPT, so I bet they never agree to it.

[hunter]

21. Fuck cursive. The devil created that.

My brain and my hands simply don't work that way no matter how hard I've tried. And I've tried hard and been beaten for my failures.

When I was a university student I'd bring at least three blue books to fill with my chicken scratching for every blue book that was expected.

I've got a couple of blue books I've kept where I reached the end, went back to the first page, turned them ninety degrees, and continued writing across from the beginning.

[edisdead]

24. I suspect this will be part of the steategy

But I also think that in large part the way we learn, digest and process information is going to fundamentally change and not necessarily for the worse. Its a big change and many people will fear it and proclaim the end of times over it but largely I feel it will not be much of a big deal.

We are so completely success driven as a species that it sort of boggles my mind. We are developing technologies to perform mundane tasks, which eventually may lead to people being less busy doing the work rather than being or learning for the sake of curiosity and personal fulfillment. The people that are going to cheat are going to cheat. Regardless of chatgpt, wikipedia, internet, etc…. That is never going to change. They cheat because we are so success driven. When there is less to cheat for there will be less cheating.

[ret5hd]

17. A possible (partial) solution:

“All written assignments will be written with using software that has a “track changes” function, and such function activated throughout the process. Any submissions suspected of being written by any AI will closely inspected. Any person found to have used AI to submit any coursework will receive an F in the class and will be ejected from the program.”

[Ms. Toad]

19. So as a retired teacher (high school and law school)

The "closed book" anti-cheating rules in my class were general enough so that anyone using ChatGTP would have been violating the rules. My "open book" anti-cheating rules might have needed to be revised - since the assistance which was expressly prohibited was human assistance. Using the internet was allowed, but the work needed to be their own. The good news is that all of my syllabi had a provision that changes could be implemented at any time, going forward. So the timing of ChatGPT would have prompted me to change it before the fall semester - but even had it come in the middle of the semester I could still implement rules to address it from that date forward.

I would use the fact that ChatGPT lies fluently in two, maybe 3, ways:

(1) teaching internet literacy - an extension of vetting websites. I might find an obscure area of the law and ask ChatGPT to write an essay on that area of the law and have students learn how to fact check it.

(2) teaching students to imitate ChatGPT's style. I've been teaching students to write essays that would pass the bar exam by making up (or reverse engineering) the law for about a decade. There's a formula to passing the bar exam (at least as to essays). If you can master the formula, even if your specific knowledge of the law is weak, you can pass the bar exam. ChatGPT is an absolute master at exactly what I've been teaching.

(3) Editing - both for bias and inaccuracies. I'd give them a ChatGPT essay which included both biases and inaccuracies and have them revise it.

Detecting cheating, in a proveable way, will be more challenging. If I started to consistently get essays which appeared out-of-sync with the students' abilities I would probalby implement (or boost the weight of) in-class work which I could verify was independent work.

[hunter]

26. My wife has often taught internet literacy, statistics, and medicine in her career...

... whenever she's not practicing medicine herself.

We were science teachers when we met. Then she went back to school, hard core.

I've bragged here I could identify ChatGPT crap in three paragraphs.

I'm guessing my wife could do it in one or two paragraphs.

When I was teaching science I was also a science fair judge at the county level. It only took a few minutes talking to a kid to see what was what, whether they were actually into the science or had followed some simple recipe they or their parents had found on the internet to tic that box on the college applications.

[Ms. Toad]

27. I'm not convinced I could identify it without the context

of interactions with the students. From friends of mine who have spent more time than I have studying it - it is remarkably smooth at writing. And since it spouts BS without a conscience, stuff which is inaccurate is likely to earn a higher grade than poorly written stuff that is accurate. I've watched a number of objectively less knowledgeable students stroll through the bar exam because they write BS with confidence, while my more hesitant - but more knowledgeable students require multiple attempts to pass.

A lot of students have picked up what feels to me like very stilted, artificial writing styles. Which makes it more likely their writing would score lower than the smoothe ChatGPT style.

But once you add the interactions, it is very easy to tell who is doing their own work.

What science did you teach? (Early in my career I taught Physics - then shifted to math and computer science until I burned out.)

[hunter]

30. I taught Biology and Intro to Physical Science in an underfunded overcrowded big city school.

Hardest job I ever had. It was in my head 24/7. So many problems, so few resources.

When my wife was accepted to graduate school in another state I happily followed her.

My wife's sister has been teaching Biology in the same sort of environment for more than thirty years. She's some kind of superhero.

[Ms. Toad]

31. Similar here.

In my remedial classes by the end of the year 2/3 of my students had moved on and been replaced by a similar nunber of new students. Most moved (evicted for non-payment of rent). Some were incarcerated. Some got pregnant. I was considered successful because, as a general rule, more than 50% of the students in m sections of those classes passed. And, since I was "successful," I was rewarded with more sections of that class.

I was burned out after 11 years. My spouse got a job in the metro area just south of where I'd been teaching so I took the opportunity to take a break. (Stay-at-home mom for 5 years, law school, law clerk in the state appellate court, IP practice for 13 years, then back to teaching to end my career.) But leaving after 11 years meant that even though I'm grandfathered in (and don't require CEU to keep my license), as a practical matter I was done in the public schools. Because I earned my masters degree in applied math and had been teaching 10 years, union rules would have required most schools around here to pay me twice as much as a teacher straight out of college. Not only that, but because I had tenure, under state law they would have had less than a year to decide whether to keep me on and grant me tenure. Not a system designed to ensure our schools have the best teachers.

[honest.abe]

25. Here is an alternate view point.

Don’t Ban ChatGPT in Schools. Teach With It.

OpenAI’s new chatbot is raising fears of cheating on homework, but its potential as an educational tool outweighs its risks.

https://www.nytimes.com/2023/01/12/technology/chatgpt-schools-teachers.html

[anarch]

28. just wait 'til people start having generative AI capable chips implanted in their brains

why bother thinking when you can have a machine do it for you?

Hell, I've already given up spelling and math with just MS Office to do it for me.