The New York City Department of Education wants students to do their own writing, not to submit essays written by a computer program.
Michael Elen-Rooney wrote in Chalkbeat:
New York City students and teachers can no longer access ChatGPT — the new artificial intelligence-powered chatbot that generates stunningly cogent and lifelike writing — on education department devices or internet networks, agency officials confirmed Tuesday.
The education department blocked access to the program, citing “negative impacts on student learning, and concerns regarding the safety and accuracy of content,” a spokesperson said. The move from the nation’s largest school system could have ripple effects as districts and schools across the country grapple with how to respond to the arrival of the dynamic new technology.
The chatbot’s ability to churn out pitch perfect essay responses to prompts spanning a wide range of subjects has sparked fears among some schools and educators that their writing assignments could soon become obsolete — and that the program could encourage cheating and plagiarism.
“Due to concerns about negative impacts on student learning, and concerns regarding the safety and accuracy of content, access to ChatGPT is restricted on New York City Public Schools’ networks and devices,” said education department spokesperson Jenna Lyle. “While the tool may be able to provide quick and easy answers to questions, it does not build critical-thinking and problem-solving skills, which are essential for academic and lifelong success….”
The education department’s ban will only cut off access to the chatbot in some settings. Students can still get on the site on non-education department devices or internet networks.
Hmmm. So students could download essays from ChatGPT at home and copy it.
Diane Ravitch's blog 
I wrote a piece about AI in Linkedin, https://www.linkedin.com/pulse/practice-creativity-paul-bonner/?trackingId=9oqoVoV6V0bJPGXYTldj9g%3D%3D, where a friend of mine took what I wrote and asked Chatgpt to rewrite it. My friend found the AI awkward and “personality and meaning was lost in the translation.” It seems to me that such activity would provide an opportunity to show that the mechanistic approach to AI simply doesn’t compare with human thought. Simply banning the platform won’t prevent abuse if this is not the universal feeling about technology. Perhaps New York should promote more time for writing that includes meaningful student reflection versus too much time on reading fluency.
I don’t understand why the Chalkbeat author says that ChatGPT produces writing which is “stunningly cogent and lifelike.” The AI writing sounds nothing like a student of any age. As someone who teaches writing in a university, I am hoping that the availability of the Chatbot signals a turn away from the focus on generic writing prompts and the “Five Paragraph Essay,” in favor of having students write in order to better learn the subject matter they are studying.
Well argued, well stated. Thank you.
Hopefully, this controversy will lead educators to question the worth of a system that demands that most teachers (the public school teachers) see 150+ children daily/yearly. You cannot teach anyone to think with that kind of class load. Writing and getting feedback on writing is the process that teaches. You cannot do that without personal attention. Trying to give 150+ personal attention is giving the sow too many piglets
yes, yes, yes
Thank you, Roy!
Good morning Diane and everyone,
I don’t know about anyone else, but to me writing is hard as the devil! 🙂 If you don’t think writing is damn HARD, you must have great talent. I’ve written a LOT. I do so for different purposes. In this recent discussion about the value of having students learn to write, I tried to think back to how writing helped me in my life. In what way was it valuable to me? First, I had to do a lot of reading and research. It used to be about going to the library and using the research tools there (I’m dating myself here!) I had to learn to find out HOW to get information. That can be a challenge in itself. Then I had to read and digest the information. I had to ask myself what I personally thought of this information. I had to think cogently. It was a challenge to bring together ideas, concepts, and information in a new and creative way. I had to put this all to paper where it could be read and understood by others. That involved a lot of thinking and revision. It wasn’t just a matter of throwing onto paper other people’s ideas, but it made me think about what I thought and the way I wanted to put MY thoughts and ideas into written form. So, for me, writing is a way to elucidate and bring into focus what I am learning and thinking. Learning to write is more than just the writing itself. It involves all the steps I’ve outlined here. So, if we want students to learn to write in this particular way, we have to find value in it. If the value of education is reduced to the amount of money I can make with it, most people won’t need to write in this way. We have to ask what writing is for. What is it in service to? And then go from there.
If I had to guess, I would guess that eventually, AI will be to writing as calculators are to mathematics. Everyone will use it to varying degrees, and the act of writing will become more like the act of editing.
Hmmm. I see what you are saying, but writing is editing, for the most part.
Writing and editing are simultaneous actions.
yes
I think we’re probably thinking the same thing. If there were a continuum, with text generation at one end and “editing” at the other, I imagine AI sliding the act of writing further toward the “editing” side of the continuum. That isn’t necessarily a bad thing.
Let us put away our technophile’s crystal ball for a sec, shall we. If bots got involved in writing, we would all be the loser. No one would be able to think without being autocorrected. Love letters would be lonely acts of emptiness as pornography is to coitus. Dianeravitch.net would be boring and lose all its readership.
Just trying to be a techno-realist. It’s going to happen, and it’s already happening bit by bit, with autocorrect.
Again with the crystal ball, and the sentient computers, and the flying robot cars, and the interplanetary boldly go where no ass has gone before…
Good points, Mamie. I would add that any particular writing takes audience into account. We do not write in life from prompts, we write because we are prompted by subject matter to write (apologies to all who deplore the passive voice). How can AI understand your audience without talking to you about it? What is more important in writing than understanding the audience?
Roy,
You are so right!
I am a professional writer. I always write with an awareness of my audience. Is it other historians? Is it the general public? Is it my family members? Every writer thinks of the audience.
Audience, voice, tone and point of view are what most human writers take into consideration when they work.
I would go a step further, Mamie. I maintain that–outside of mathematics–we don’t know entirely what we think until we have attempted to express our ideas in writing. Writing IS thinking.
True for me. I only wish I could capture what I say to myself while I work alone on mindless jobs. I sure feel smart then.
There are both pros and cons to consider before deciding whether students should use ChatGPT.
On the pro side, ChatGPT could potentially be a useful tool for students who are struggling to understand certain concepts or who need help completing assignments. The chatbot is able to provide information and explanations on a wide range of topics, and it can also assist with research by finding and summarizing relevant information. Additionally, ChatGPT could potentially be a useful communication tool for students who are learning remotely or who are unable to meet in person with their classmates.
However, there are also several potential drawbacks to using ChatGPT as a student. One concern is that relying too heavily on the chatbot could prevent students from developing their own critical thinking skills and problem-solving abilities. It is important for students to learn how to think for themselves and to be able to find and evaluate information on their own, rather than relying on someone or something else to do it for them. Additionally, ChatGPT is not perfect, and it may not always provide accurate or reliable information. It is important for students to be able to critically evaluate the information they receive, rather than blindly accepting everything they are told.
In conclusion, whether students should use ChatGPT or not depends on their individual needs and circumstances. While it can potentially be a useful tool for certain tasks, it is important for students to be aware of its limitations and to use it responsibly. It is ultimately up to each student to decide whether ChatGPT is a helpful tool or a potential hindrance to their learning and development.
HAHAHAAHA! Who else FEELS like this was NOT written by a human being??? And if it was written by a human being, I must say that I’m NOT impressed!!
That was me, using ChatGPT.
I find this stuff very interesting. It’s crude and the language is stilted. Then again, so is most student (and adult) writing. And it surely will continue to improve, probably at an increasing rate. I could imagine this being extremely useful for generating words and ideas when a writer is at a dead end and needs a jump-start.
Thanks Flerp!
The writing of this Chat thing is insipid and lacks any style. If a student handed this to me, I would tell him to go and write it himself with all the chaos, mistakes and horrible grammar he could possibly muster!!!
Exactly my impression when I was about halfway through it, & I wasn’t surprised to learn that was the case! ChatGPT itself has a recognizable “style.” Given the widespread alarm, I’m guess there may be a rash of cases of students falsely accused of using ChatGPT when the writing is their own.
Not impressed with use of the bot here. Why do people assume that a chatbot will improve over time? Its nose would not grow if it were to “tell” a lie. Nothing about ChatGPT will ever grow or change. It’s not a real boy.
The NYC DoE wants to BAN THE BOT because publishing writing which is not your own is impolite and dishonest. It is not model behavior. I certainly hope my students respect me enough not to post for me any writing that is not their own. I always feel disrespected when someone cheats me out of reading authenticity.
Spelling is included.
ChatGPT is really quite literally a “plagiarbot” because it constructs it’s sentences and paragraphs from the large amount of text scraped from the world wide web..
It will be interesting to see if it actually does get better over time.
It js by no means obvious that that will be the case because of tge way it works. It is based on a neural network that is trained on a huge amount of text taken off the world wide web. It is generated automatically and given the shear size of the training data, it’s hard to see how one could hope to change the output to “improve” certain aspects of the writing. It’s not based on logical “rules” that a human can tweak to make it better, as one might do with ones own writing. There is no source code to even look at. It’s just a gigantic neural net of nodes with over a hundred billion nodes that have been automatically weighted by the huge amount of text used to train it. It would make no sense to any human who looked at it. How would one even begin to try to “improve” the output of such a beast?
And if one attempts to “improve” the output by restricting the input training data set to only included”well written ” text, one also restricts the scope of the output.
It might improve, but given how it works, that is certainly not obvious.
I am not alone in grasping the destruction of the SKILL of writing!
I was the NYC cohort for the Harvard research on the Principles of LEARNING!
It’s all about learning
Each seminar for teachers which came to District 2, when I was chosen as cohort, began with this Essential Question: “ What Does LEARNING LOOK like?”
An EQ asks something important!
The solution lies in the answer.
I can tell you what learning to write looks like in every single student in my classes.
I can do that because I had the proper tool to evaluate their progress and competency in acquiring the SKILL I was …”teaching “!
My Reader’s Letters program did this for all my students; based on Nancy Atwell’s use of this tool for practicing writing, while talking about what was read — my ‘kids’ learned to write by WRITING!!!!
So, I can predict with some accuracy the result of this ERA of technological TRANSFORMATION of information !
This generation of children will not only be unable to do the critical analysis/thinking that enables excellent writing, they will be unable to use words… as they possess too few.
Did you know the average Elizabethan used 52,000 words?
Our kids don’t read. They watch images, play video games
They don’t talk. Their heads buried in their cellphones as they eat at restaurants, or walk in the streets.
The writing process is what it is— a process.
It begins with ‘listening’ and is followed my ‘speaking’ — and NEXT comes ‘READING ‘— and then ,comes Writing’!
These children do not LISTEN to words, anywhere, and not in school.
In my class, after listening to words, they spoke in groups about their thoughts, and ideas.
Then, we read stories, and we shared their ideas!
Then, they wrote!
Letters to me, and stories and essays.
They won most nationwide writing competitions.
They were third in NYS when 3/4 of NYC failed the first ELA test.
So, what do I think about this technology?
It is the deathknell of writing, and one more nail in the wall being erected to end true literacy!!
Ignorance is the goal of the powerful people who control public education!
An ignorant population gave us Trump and his ilk, and filled our Congress with the likes of MTG, Santos and McCarthy!
This moment in time is unfolding the toll that ignorance takes on a democracy rooted in the WORDS of the founding fathers.
Words Matter. That is the title of the piece I am writing for my blog Speaking As A Teacher!
There was a considerable increase in literacy over during the Elizabethan Era. By then end of it, about 30 percent of adults in England could read at least minimally. Only about 30 percent. That even the average passive vocabulary reached 52,000 words seems highly doubtful.
Shakespeare used a little over 31,000 distinct words in his plays (active written vocab) and might have known another 30,000 (passive spoken and written vocabulary). And he was exceptional, a great genius.
It’s doubtful that the average total individual vocab at that time was only slightly smaller than was Shakespeare’s. Most people were illiterate, and the vocabularies of illiterate people aren’t that large, usually (though there are exceptions–ancient Homeric singers, for example, had lare vocabularies).
I suspect that the 52,000 figure comes from people misremembering some estimate of Shakespeare’s vocabulary.
I know of no reliable study that suggests that average vocabulary is in decline in the US. I do know that average IQ has increased about 1 percent per decade for over a century now.
Thanks, Bob. I was about to raise the same question about the average size of the Elizabethan vocabulary. Not to argue the basic point that we are seeing a shrinkage of vocabulary in modern times. I think this is true of our language, largely due to the effect of the compartmentalization of life in general. We know jargon about finance, mechanics, or plumbing, but we probably do not branch out as much as some generations have.
A well-placed argument might change my perception of the matter. I speak from personal experience, which can be simplistic.
The kids today are gleaners of very, very broad fields.
Recommended, Roy:
Mortimer, Ian. The Time Traveler’s Guide to Elizabethan England. Penquin, 2012.
Also great:
Mortimer, Ian. The Time Traveler’s Guide to Medieval England. Viking, 2012.
cx: Penguin
Estimates of vocabulary size must be taken with a large grain of salt. Most are utter bs. Estimating vocabulary size is quite difficult. First, what counts as a vocabulary item? Base or root words? Each inflected (sing, sang, sung) or derivative (imply, implication) form? Each meaning (electrical potential, human potential)? Second, does the “study” involve gathering a large representative corpus from each participant? Or does it involve a few items that some nonlinguistic thinks of as hard words? What data was compiled, and how was it evaluated? Third, are we talking active vocabulary, passive vocabulary, or both?
Bottom line: most of what people say on this subject is unscientific white noise.
cx: What data were compiled?
cx: some nonlinguist
cx: Individual base words, roots words, and function words?
In using the term “white noise,” I am alluding to Don DeLillo’s wonderful novel with that name (White Noise), recently turned into a delightfully funny movie. The move is great. The book, better. One of the lovely pieces of invention used in this novel is that the characters are constantly inserting into their casual conversation facts that are simply wrong but that they do not recognize as such. They and we swim in a sea of misinformation and disinformation.
As with other things, it’s not the size of the vocabulary but what you do with it that counts.
And I have known people who can say much with just one word (albeit conjugalated as a noun, verb, adjective, adverb and other parts of speech)
Perspicacious, ineluctable, indefeasible, SomeDAM!
An jist grate!
I think it was the linguist Reginald Quirk who edited a dictionary with over a hundred distinct meanings for the much-feared f word (which is almost as scary as drag queens are). And then there is the s word, which is also so very expressive to so many ends. Both noble Anglo-Saxon inheritances!
Size is overrated
And often overstated
And less is really more
When digging to the core
I think it was the linguist Reginald Quirk who edited a dictionary with over a hundred distinct meanings for the much-feared f word ”
Yes, that’s pretty quirky.
When I read that, I searched to see if you had just made up the the name Reginald Quirk.
His parents must have laughed all the way home from the hospital after coming up with that name Reggie Quirk
Why are his days on poetry spent?
Because he’s quite frankly neologoplacent.
Knee AH lo go PLA sent
Reginald, of the Quigsby Quirks, nephew to Lord Bottom of Frogg
Buried in the Quigsby Quirk Kirk
Pretty quirky and pretty twerpy
Twerky
You are interrupting me during my work on the Rhonda Santis 2024 campaign song, SomeDAM!
Saw a drag queen, and my heart stood still.
Uh do run Ron Ron, uh do run Ron
Thought I’d better write another hate bill.
Uh do run Ron Ron, uh do run Ron.
Yes, his voice was shrill.
Yes, he wrote that bill.
At first, of course, he was just Trump’s shill.
Uh do run Ron Ron, uh do run Ron.
That’s a good one!
RQ is a deeply learned fellow.
One of the things I like about poetry is that you increase your vocabulary without learning any more actual words — by just making up new words.
I’m totally with you, Susan, about the value of listening to increase in vocabulary and in grammatical fluency (the number of syntactic forms recognized, understood, incorporated into the child’s internal grammar, and used). The value of speaking and listening exercises needs to be more widely understood.
While ChaptGPT does pose a valid concern, it’s not the disaster for human writing many alarmists believe. As stated elsewhere in these Comments, its writing tends to be generic, based on study of all available text throughout the internet. Once you get beyond the impressively sophisticated sentence structure, it’s not going to pass for the writing of a student with whose style the teacher is familiar, perhaps from a couple of in-class writing assignments.
It also makes random glaring errors that are sure giveaways:
“Do I have to wear seatbelts when I’m carrying my car?”
“Yes, you have to wear seatbelts when you’re carrying your car.”
I’m sure there’ll be a bit of a shakeup & some adjustment required in teaching procedures, but I don’t think it’s “the end of human writing.”
Or, to put it another way:
It is understandable that ChatGPT and other language generation models may raise concerns about the potential impact on education, particularly in the teaching of writing. However, it is important to recognize that these models should not be viewed as a replacement for traditional writing instruction, but rather as a tool that can supplement and enhance the learning experience.
One of the main arguments against the use of ChatGPT in education is the potential for the model to produce biased or inaccurate language. It is important to keep in mind that ChatGPT is trained on a vast amount of internet data, which may include biased or inappropriate content. It is the responsibility of educators to critically evaluate the output of the model and teach students how to do the same.
Overall, ChatGPT and other language generation models have the potential to be valuable resources for students and teachers alike, but they should be used with caution and in conjunction with traditional writing instruction. By teaching students how to responsibly and critically evaluate the output of these models, we can harness their power to enhance the learning experience and better prepare students for the future.
ChatGPT-generated response to, “Respond in 3 paragraphs to an article warning of the dangers ChatGPT poses to teaching writing in public schools.”
Note that the response doesn’t mention “public schools,” the focus of the question, once.
Bear in mind that it’s going to get a lot better. Right now it’s essentially an open beta, open for free public use for research that will improve the model.
There was a 20-30 year period where automated voice transcription was absolutely terrible, and a lot of people were convinced that it would never work accurately. About a decade ago it broke through to mainstream use and it’s more impressive every time I try it. Microsoft Tesms has a live transcription function that is close to flawless in my experience.
There are some writers that are using apps like this one to churn out novels. I don’t remember the name of the one that I read about, but he’s published more than 450 Science Fiction/Fantasy novels on Amazon under his name.
After one of the AI written novels is completed, he turns that rough draft over to an intern who cleans it up and then this author publishes them in his name, taking full credit for them. The intern gets no credit and probably earns a few hundred dollars for the job.
Many of those novels have thousands of reviews and run more than 500 to 1,000 pages. I watched a YouTube vido of this do-called author promoting his work, and he looked like he wasn’t 40 yet. He also reminded me of a used car salesman.
Does anyone here know someone, anyone, that could write hundreds of 500 to 1,000 page novels by themselves and successfully publish them all in a couple of decades? I don’t. It took me more than a decade to finish one historical fiction novel running 640 pages.
I’ve read that Joyce Carol Oats, who started writing when she was 12, works long days on her novels, often 12 hours a day without end.
“Joyce Carol Oates is an American writer. Oates published her first book in 1963, and has since published 58 novels, a number of plays and novellas, and many volumes of short stories, poetry, and non-fiction.”
He also reminded me of a used car salesman.
Was his name Elon Musk, by chance?
Now that Tesla’s new car sales are on the toilet, Elon might have to become a used car salesman on weekends to pay Twitter schools.
Bills.
The DOE neglected to consider that virtually every person in every city school is carrying a phone that does not require the DOE network. Perish forbid they should hire someone capable of considering the blatantly obvious.
Ya think !
That doesn’t mean it’s stupid to restrict access to the web site on school devices or over school networks. It obviously won’t eliminate the problem but making it marginally more difficult to access isn’t a bad idea.
What’s stupid is that a minor thing like this is presented like it’s a significant news story.
The biggest threat from this has nothing to do with student writing.
It has to do with pollution of the human sphere of communication with an exponential explosion of bot generated garbage, much of it false.
There will soon be so much of this bot generated stuff that the genuine human communications will get completely lost in a sea of botshit.
ChatGPT is just the first of the botshit artists.a be once these things get really cooking, there will be no stopping them.
Of course, it probably won’t affect social media like Twitter and Facebook at all, since they are already effectively filled with botshit produced by Musk and other humans.
But it will have an impact in areas of human communications that are actually still meaningful.
botshit
perfect, someDAM!
Well said, SDP. I imagine humans resigning and allowing the world and their lives to be run by bots.
There is another big problem with all this “generative AI like ChatGPT and it’s art analog DALLE 2.
Since it trains on data from the internet — much of it cooyrighted — when it generates it’s output, it may be infringing cooyrighted wlrks.
These AIs are cleverly designed so they purposely do not simply exactly copy from the training data, but what they do do is identify patterns and reproduce those patterns, albeit with tweaks.
You can ask that DALLE produce a painting in the style of a partivular artist , for example.
Such derivative works would be considered cooyright infringement were a human to produce them.
Most people and even some artists are not aware that you can not legally do a painting of someone else’s artwork (painting of photo) without the express permission of the artist.
Companies like OpenAI plan to charge for use of their bots and they are allowing people to use them to sell what they produce.
This is stealing in my book.
These people are profiting on the backs of all the artists and authors who produced the works on which these AIs were trained without giving the original creators anything for it.
It turns out MIcrosoft, GitHub and OpenAI are already being sued for violating cooyrighted computer code with a code generating bot that trains on code on the internet.
https://www.theverge.com/2022/11/8/23446821/microsoft-openai-github-copilot-class-action-lawsuit-ai-copyright-violation-training-data
It will be interesting to see where this goes and I hope more artists of all types get involved because the folks at OoenAI and elsewhere should not be allowed to get away with profiting off of others’ creative works.
It never ceases to amaze me just how completely lacking in ethics some of these techniques is are.
Techies
Not incidentally, in the case of the code generating bot, The folks who have brought the say the bot DOEs just copy unchanged snippets of code and use them in the generated code — without any compensation or even attribution to the original author, of course.
If
The AI Bots
They’re mimicking the works of art
And modifying stories
Reconstructing part by part
Accepting all the glories
The bots will get the Nobel Prize
And bots will get the Fields
But credit will be based on lies
And training data steals
Even the mere download of cooyrighted images by these art bots is a violation of copyright.
“Uploading or downloading works protected by copyright without the authority of the copyright owner is an infringement of the copyright owner’s exclusive rights of reproduction and/or distribution.”
Copyright and Digital Files (FAQ) – U.S. Copyright Office
I hope artists ban together to shut down illegal operations like OpenAI
Not just art bots but any bot downloading images Toma database to do its training.
You can not even store a cooyrighted images on a computer without tge express permission of the copyright owner.
And a work becomes cooyrighted as soon as it is created. The owner does not have to file a formal application to the Copyright office.
Any bot downloading ANY cooyrighted work (images or text) is in violation.
Finally, it’s going to be really hard for Microsoft, OpenAI and others to claim they are doing it for an “educational purpose “when they are making money off their bots.
What they are engaging in is a criminal operation.
If it were not so serious, it would be hilarious that at least one of the AI photo generating bots that had illegally trained on photos from Shutterstock was actually including the Shutterstock watermark in its generated photos.
Busted!!
Some of the computer “scientists” who produce this stuff ain’t the sharpest spoons in the drawer.
And the ones who are bright enough to ensure that their bots cont recreate patterns like watermarks are unethical — even more unethical than the ones who don’t make any effort to keep the watermarks out.
There is actually another significant issue with these AI generative bots.
A significant fraction of the data (especially images) on the internet are pornographic.
So unless the computer “scientists” producing these bots are specifically screening out such material from the training data or screening tge output , such bots will be like ticking time bomb just waiting for tge first person to put in the “right” prompts to produce all sorts of artificially generated porn, including undoubtedly kiddie porn generated from 12 year old classmates images.
Companies like Open AI are going to quite literally be “open season” on all sorts of lawsuits unless they can somehow ensure that none of this stuff gets through.
Good luck with that is all I can say.
All it takes is one time and OpenAI will be shut down quicker than you can say “pornbot”
Ticking time bot” — a bit that was trained on pornographic images and is now waiting for the prompt that generates kiddie porn of the President’s daughter
Maybe they should rename it OoenlegsAI because that is precisely what it is bound to become
OpenlegsAI
And you just have to wonder how many of the developers of This stuff have ALREADY used it to generate porn.
Also, since those hypothetical child images don’t show an actual person, would they even be illegal? Do the laws apply only to real persons, or cover artistic representations of persons?
Some of this art AI will produce paintings of people from images of them, so my guess is that you could have a face of a real person put on a generated pornographic image..
I could be wrong about that but if it can be
done, you can be sure that someone (or millions of someone’s) will do it..
These bots are going to be used for all sorts of nefarious and criminal purposes.
Count on it.
There are no controls on any of this stuff and the problems are in the cusp of compounding exponentially.
At that point, tge only way the government will be able to control any of it is to shut it all down completely.
And OpenAIs “nonprofit” status is simply a joke that will not help tgem in Court, particularly not when they are backed by Microsoft.
There is actually precedent for compketely shutting down operations that are even facilitating illegal activities: Napster file sharing
What about . . .
Google Translator
Photomath
Brainerly
CheggStudy
Socratic
Fact Monster
Study Geek
Quizlet
SpakNotes
StudyPool
99Papers
EssayBot
EssayTyper
. . . and the list is endless if you Google free homework help.