Let me say something about repetition and patterns. I spent 27 years as a coach and judge of HS Forensics. I've spent countless hours listening to students use 3 point analysis to explain why the US should continue to fund NASA (for example) and I've sat through countless debates on resolutions like, "Resolved: Civil Disobedience is a just response to oppressive government."
By and large, regardless of the topic or the resolution, students followed certain formats. In Debate (Lincoln Douglas debate) it was the Toulmin rhetorical method (Claims, Evidence, Warrants, Impacts) and in Extemporaneous, it was generally 3 point analysis organized around time, location, or hierarchy.
Over time, as a coach, my students became far more fluent and advanced in their rhetorical choices, and their choices were more agile, creative, and nimble.
But they could not have gotten there without first understanding these forms.
And yes, debate is a game with rather predictable patterns at the novice level. But at some point, with enough experience and real-world feedback, they make a huge cognitive leap.
Surely I'm not suggesting that all students should engage in the rigorous and often ridiculous event of HS Forensics, where some of the very best speeches I've ever seen are those that lampoon just how predictable their speeches are.
But I am suggesting that kids need lots and lots of practice and that understanding the importance of form as a scaffold is important...so long as we also understand we need to help them move away from this.
I've spent thousands of dollars and endless hours freewriting and revising work through attendance at Bard College's Institute for Writing and Thinking. I know how I write, and that knowledge is a debt I owe to Peter Elbow, who never taught the way most HS English teacher teach:
"The Teacherless Writing Class
According to Elbow, improving your writing has nothing to do with learning discrete skills or getting advice about what changes you need to make. This stuff doesn’t help. What helps is understanding how other people experience your work. Not just one person, but a few. You need to keep getting it from the same people so they get progressively better at transmitting their experiences while you get better at receiving them."
If what ChatGPT does is, as you and many here and elsewhere are saying, is force out those who profess forms and efficiencies over voice and engaging prose, then count me in. I'll put on my VR headset and lead the way to something more human, something unpredictable, something closer to our own truths in words.
I confess a fondness for the five paragraph theme, but mostly only as a means of writing parodies of five paragraph essays. I don’t really want to teach again (through I did teach an online writing class a year ago), but I do think I’d do a much better job at 47 than I did at 24, in part because I’ve realized that writing is genuinely difficult for a lot of people—and that coming up with forms of writing that make sense as “writing one might actually be called upon to do in one’s actual life” is a better mode of approaching the whole business.
(Some day perhaps someone will decide I am qualified to teach seminars on How to Write a Work Email in addition to my dream job of of teaching people How to Run a Meeting.)
Excellent points, John. I think that the Prussian model of education which is designed to segregate the large majority of students into obedient workers while praising rewarding a subset of leaders has finally met its match in the AI. The AI accomplishes what many of my elementary school teachers wanted us to do, which is to digest the information that they gave us without question and spit it back out in an approved way. The Japanese have a saying that the nail that sticks out furthest gets hit the hardest, and that certainly can apply in American public schools. I agree with you that this system is not worth saving. Mediocrity is easy enough to achieve anyway.
This gets at the problem exactly. Students hate writing because we obsess over the product, the structure, the details of what the writing should turn into, instead of the process of writing itself. We fixate on the outcome of student writing as if 8th graders are going to be able to produce something valuable if they could only figure out how to get those sentence transitions right. And we then lose out on all of the value that comes from encouraging a regular practice of writing as a way of expression, rather than a system for producing an output.
But of courze the output is easier to grade on an AP test, and since the whole point of school is to get good marks on standardized tests I suppose it all makes sense.
This is nice. Concerning GPT3, I knew I had no fear because like others, I don't deny my sentimental desire, or as George Orwell wrote in "Why I Write," for Sheer egoism.
We want to write. We want to communicate. We want to express. And we want to reproduce. As long as we keep these instincts alive - or not deny them because of some hyper notion of rationality, writing remains alive and GPT3 won't be the Lord.
Thank you for sharing your expertise in a really interesting and thoughtful essay.
"The reason the appearance of this tech is so shocking is because it forces us to confront what we value, rather than letting the status quo churn along unexamined" -- right to the marrow; excellent insight.
"It’s not every week that someone with my particular employment profile and expertise has something they’re knowledgable about become a hot topic of national discussion"
Hmm. Sounds just like what an AI chatbot *would* say. (LOL)
In my senior AP English class, once we sat for the exam in early April, fake curriculum was over and the real teaching began. Groups had to create a 3 minute movie based on any novel. Our film may have been heavily influenced by Cruel Intentions (17 year olds have notoriously sketchy taste). I will remember that assignment for the rest of my life.
Also, which question on the quiz did you miss??? 🙃
John, I've been interested to hear your take on this and it doesn't disappoint. It reminded me of two great (not-five-paragraph) essays worth sharing. Your thoughts about what needs to change in college writing pedagogy made me think of this provocative piece by John Michael Colón about what needs to change in college reading pedagogy (to save the humanities in the same way you'd like to save college writing): https://thepointmag.com/letter/on-the-end-of-the-canon-wars/?mc_cid=0f46429ec7&mc_eid=21b801de99. Colón says: "The value of the humanities is, upon exposure to real humanistic practice, self-evident"--that when students read and engage with great books, when they wrestle with them, debate and discuss them, they learn from them. Then they value them without needing to be told "why." Reading and writing are two sides of the same coin. We can't have one without the other, and both need to be valued if either is to be saved.
The second essay is my all-time favorite piece on NLP-generated writing by one my favorite essayists (and writing teacher) Meghan O'Gieblyn, cleverly titled "Babel": https://www.nplusonemag.com/issue-40/essays/babel-4/#fn27-13678). The essay was published in 2021 (a lifetime ago in AI years), but it feels uncannily prescient given the current buzz around ChatGPT. O'Gieblyn questions the nature of machine writing and human writing, and subtly challenges their differences. Of GPT-3, she says: "There was something prismatic in its voice, an uncanny chorus of intertextuality, the haunting of writing by other writing. The internet was driven from its earliest days by the promise of universal authorship. Hypertext and collaborative software were going to revive the myth and the folktale, narratives created not by a single creative genius but by the collective effort of any oral storytelling culture. It is tempting at times to see this technology as the realization of that dream, a repository of the collective wisdom and knowledge we’ve accumulated as a species. All of humanity speaking in a single voice."
I'm no O'Gieblyn, but I've done some writing and research in this area. (I did an independent study called "Can a Bot Read? What Happens When the Digital Becomes Literate" with two wonderful, well-respected Information Science professors.) And I'd gently push back on the claim that ChatGPT is a "bullshitter" who doesn't "read" or "write." I think that's true only in the most literal sense. As O'Gieblyn points out, algorithms like GPT-3 and its cousin ChatGPT, are trained on massive data sets of human language. So while it's not explicitly programmed with grammar, it absorbs our speech patterns, which contain grammatical structures. We humans have effectively taught it to "speak," and we've done so by giving up our data to corporations who then use that data to train these machines. The machines learn by trawling or "reading" that data and looking for patterns, a process that I can't help but compare to close reading (only the machines are infinitely better close readers than humans are. They see things our brains can't even register). So, while these systems don't "read" and "write" using the exact same processes as humans, there are some similarities and the effect is often the same. I worry that when we dismiss ChatGPT and its ilk as "bullshitters" or "toys," we ignore their real power over us. I'm no futurist either but I suspect these technologies are going to have a much bigger impact on society than most people realize. (Remember when the iPhone first came out and everyone thought it was just a Walkman you could call your friends on?) Anyway, I'm happy people in reading and writing communities are talking about these technologies, even if it's just the start of the conversation. Thanks for writing about it!
You're correct in every point, but ... For most students, all of the immediate incentives are to get the grade, or if they've got a somewhat longer view, to learn to write the sort of things schools will ask them to write in the future. As someone commented about Harvard students, they're "incredibly good at figuring out how to do exactly what they need to do to get the grade. They're incredibly strategic. And I think that's really true of students everywhere." I want to agree with the concept of students wanting to learn how to write to express themselves, but the fraction of students who want to express themselves creatively in writing is fairly small, and I suspect the need for that skill in the "real world" outside school is significantly smaller than the supply of people who can do it now under the current blighted system.
The only sector of literature that I know anything about is science fiction, and some famous S.F. writer noted that the number of people who made a living writing S.F. at any one time "could fit in a van".
There's an irony, in that writing prompts to get ChatGPT to produce adequate versions of writing assignments is itself a matter of writing, and fairly creative writing at that, in that it's not a known process (yet). In a way, it's like a lot of the better-grade writing people need to do in life, not an expression of "what they have to say" but requiring considerable skill to write a text that will induce someone/something to do some particular act the writer wants.
Fascinating that the best takes I've seen on the ChatGPT are all from people who have also taught writing, either at a high school or at a college level.
I think it's because those of us that taught and assessed writing in those academic contexts had to develop a very precise BS detector. To help somebody improve their essay, you need to quickly diagnose where they are writing things they don't understand.
ChatGPT very much sounds like a B essay cobbled together the night before.
Thanks for this article. It hits on so many valuable and interesting points that get at the heart of why so many students, rather than seeing education as a tool to open up a wide world of opportunities, see it as a guided tour through a prison.
Thank you. Somewhat reminds me of Harry Chapin's 1978 song "Flowers are Red". Since high school English has become reading propaganda (name your poison) and encouraging ChapGPT, yes, let it fall into the abyss of uselessness. Then also, work on its renewal as a useful art, not a robotic tool.
Everyone has been focused on the students using ChatGPT to cheat, but that's like worrying that the invention of cars will startle the horses. This will transform the classroom experience.
I've been using ChatGPT to learn Japanese through conversation. I can definitely see teachers incorporating this technology into their classes. The model can suggest vocabulary words, discuss the cultural context of those words, create a quiz, accurately grade my answers, provide a dialogue written in Japanese, and give feedback as I translate each line of the dialogue.
Currently, I do have to watch for the BS and search for errors, but I foresee a (near) future where AI models like this are an essential tool for teaching and learning.
Let me say something about repetition and patterns. I spent 27 years as a coach and judge of HS Forensics. I've spent countless hours listening to students use 3 point analysis to explain why the US should continue to fund NASA (for example) and I've sat through countless debates on resolutions like, "Resolved: Civil Disobedience is a just response to oppressive government."
By and large, regardless of the topic or the resolution, students followed certain formats. In Debate (Lincoln Douglas debate) it was the Toulmin rhetorical method (Claims, Evidence, Warrants, Impacts) and in Extemporaneous, it was generally 3 point analysis organized around time, location, or hierarchy.
Over time, as a coach, my students became far more fluent and advanced in their rhetorical choices, and their choices were more agile, creative, and nimble.
But they could not have gotten there without first understanding these forms.
And yes, debate is a game with rather predictable patterns at the novice level. But at some point, with enough experience and real-world feedback, they make a huge cognitive leap.
Surely I'm not suggesting that all students should engage in the rigorous and often ridiculous event of HS Forensics, where some of the very best speeches I've ever seen are those that lampoon just how predictable their speeches are.
But I am suggesting that kids need lots and lots of practice and that understanding the importance of form as a scaffold is important...so long as we also understand we need to help them move away from this.
I've spent thousands of dollars and endless hours freewriting and revising work through attendance at Bard College's Institute for Writing and Thinking. I know how I write, and that knowledge is a debt I owe to Peter Elbow, who never taught the way most HS English teacher teach:
"The Teacherless Writing Class
According to Elbow, improving your writing has nothing to do with learning discrete skills or getting advice about what changes you need to make. This stuff doesn’t help. What helps is understanding how other people experience your work. Not just one person, but a few. You need to keep getting it from the same people so they get progressively better at transmitting their experiences while you get better at receiving them."
If what ChatGPT does is, as you and many here and elsewhere are saying, is force out those who profess forms and efficiencies over voice and engaging prose, then count me in. I'll put on my VR headset and lead the way to something more human, something unpredictable, something closer to our own truths in words.
I confess a fondness for the five paragraph theme, but mostly only as a means of writing parodies of five paragraph essays. I don’t really want to teach again (through I did teach an online writing class a year ago), but I do think I’d do a much better job at 47 than I did at 24, in part because I’ve realized that writing is genuinely difficult for a lot of people—and that coming up with forms of writing that make sense as “writing one might actually be called upon to do in one’s actual life” is a better mode of approaching the whole business.
(Some day perhaps someone will decide I am qualified to teach seminars on How to Write a Work Email in addition to my dream job of of teaching people How to Run a Meeting.)
Excellent points, John. I think that the Prussian model of education which is designed to segregate the large majority of students into obedient workers while praising rewarding a subset of leaders has finally met its match in the AI. The AI accomplishes what many of my elementary school teachers wanted us to do, which is to digest the information that they gave us without question and spit it back out in an approved way. The Japanese have a saying that the nail that sticks out furthest gets hit the hardest, and that certainly can apply in American public schools. I agree with you that this system is not worth saving. Mediocrity is easy enough to achieve anyway.
This gets at the problem exactly. Students hate writing because we obsess over the product, the structure, the details of what the writing should turn into, instead of the process of writing itself. We fixate on the outcome of student writing as if 8th graders are going to be able to produce something valuable if they could only figure out how to get those sentence transitions right. And we then lose out on all of the value that comes from encouraging a regular practice of writing as a way of expression, rather than a system for producing an output.
But of courze the output is easier to grade on an AP test, and since the whole point of school is to get good marks on standardized tests I suppose it all makes sense.
This is nice. Concerning GPT3, I knew I had no fear because like others, I don't deny my sentimental desire, or as George Orwell wrote in "Why I Write," for Sheer egoism.
We want to write. We want to communicate. We want to express. And we want to reproduce. As long as we keep these instincts alive - or not deny them because of some hyper notion of rationality, writing remains alive and GPT3 won't be the Lord.
John, thank you for breaking down AI like this — and for your kind advocacy for Ancestor Trouble!
Thank you for sharing your expertise in a really interesting and thoughtful essay.
"The reason the appearance of this tech is so shocking is because it forces us to confront what we value, rather than letting the status quo churn along unexamined" -- right to the marrow; excellent insight.
"It’s not every week that someone with my particular employment profile and expertise has something they’re knowledgable about become a hot topic of national discussion"
Hmm. Sounds just like what an AI chatbot *would* say. (LOL)
In my senior AP English class, once we sat for the exam in early April, fake curriculum was over and the real teaching began. Groups had to create a 3 minute movie based on any novel. Our film may have been heavily influenced by Cruel Intentions (17 year olds have notoriously sketchy taste). I will remember that assignment for the rest of my life.
Also, which question on the quiz did you miss??? 🙃
John, I've been interested to hear your take on this and it doesn't disappoint. It reminded me of two great (not-five-paragraph) essays worth sharing. Your thoughts about what needs to change in college writing pedagogy made me think of this provocative piece by John Michael Colón about what needs to change in college reading pedagogy (to save the humanities in the same way you'd like to save college writing): https://thepointmag.com/letter/on-the-end-of-the-canon-wars/?mc_cid=0f46429ec7&mc_eid=21b801de99. Colón says: "The value of the humanities is, upon exposure to real humanistic practice, self-evident"--that when students read and engage with great books, when they wrestle with them, debate and discuss them, they learn from them. Then they value them without needing to be told "why." Reading and writing are two sides of the same coin. We can't have one without the other, and both need to be valued if either is to be saved.
The second essay is my all-time favorite piece on NLP-generated writing by one my favorite essayists (and writing teacher) Meghan O'Gieblyn, cleverly titled "Babel": https://www.nplusonemag.com/issue-40/essays/babel-4/#fn27-13678). The essay was published in 2021 (a lifetime ago in AI years), but it feels uncannily prescient given the current buzz around ChatGPT. O'Gieblyn questions the nature of machine writing and human writing, and subtly challenges their differences. Of GPT-3, she says: "There was something prismatic in its voice, an uncanny chorus of intertextuality, the haunting of writing by other writing. The internet was driven from its earliest days by the promise of universal authorship. Hypertext and collaborative software were going to revive the myth and the folktale, narratives created not by a single creative genius but by the collective effort of any oral storytelling culture. It is tempting at times to see this technology as the realization of that dream, a repository of the collective wisdom and knowledge we’ve accumulated as a species. All of humanity speaking in a single voice."
I'm no O'Gieblyn, but I've done some writing and research in this area. (I did an independent study called "Can a Bot Read? What Happens When the Digital Becomes Literate" with two wonderful, well-respected Information Science professors.) And I'd gently push back on the claim that ChatGPT is a "bullshitter" who doesn't "read" or "write." I think that's true only in the most literal sense. As O'Gieblyn points out, algorithms like GPT-3 and its cousin ChatGPT, are trained on massive data sets of human language. So while it's not explicitly programmed with grammar, it absorbs our speech patterns, which contain grammatical structures. We humans have effectively taught it to "speak," and we've done so by giving up our data to corporations who then use that data to train these machines. The machines learn by trawling or "reading" that data and looking for patterns, a process that I can't help but compare to close reading (only the machines are infinitely better close readers than humans are. They see things our brains can't even register). So, while these systems don't "read" and "write" using the exact same processes as humans, there are some similarities and the effect is often the same. I worry that when we dismiss ChatGPT and its ilk as "bullshitters" or "toys," we ignore their real power over us. I'm no futurist either but I suspect these technologies are going to have a much bigger impact on society than most people realize. (Remember when the iPhone first came out and everyone thought it was just a Walkman you could call your friends on?) Anyway, I'm happy people in reading and writing communities are talking about these technologies, even if it's just the start of the conversation. Thanks for writing about it!
You're correct in every point, but ... For most students, all of the immediate incentives are to get the grade, or if they've got a somewhat longer view, to learn to write the sort of things schools will ask them to write in the future. As someone commented about Harvard students, they're "incredibly good at figuring out how to do exactly what they need to do to get the grade. They're incredibly strategic. And I think that's really true of students everywhere." I want to agree with the concept of students wanting to learn how to write to express themselves, but the fraction of students who want to express themselves creatively in writing is fairly small, and I suspect the need for that skill in the "real world" outside school is significantly smaller than the supply of people who can do it now under the current blighted system.
The only sector of literature that I know anything about is science fiction, and some famous S.F. writer noted that the number of people who made a living writing S.F. at any one time "could fit in a van".
There's an irony, in that writing prompts to get ChatGPT to produce adequate versions of writing assignments is itself a matter of writing, and fairly creative writing at that, in that it's not a known process (yet). In a way, it's like a lot of the better-grade writing people need to do in life, not an expression of "what they have to say" but requiring considerable skill to write a text that will induce someone/something to do some particular act the writer wants.
Fascinating that the best takes I've seen on the ChatGPT are all from people who have also taught writing, either at a high school or at a college level.
I think it's because those of us that taught and assessed writing in those academic contexts had to develop a very precise BS detector. To help somebody improve their essay, you need to quickly diagnose where they are writing things they don't understand.
ChatGPT very much sounds like a B essay cobbled together the night before.
Thanks for this article. It hits on so many valuable and interesting points that get at the heart of why so many students, rather than seeing education as a tool to open up a wide world of opportunities, see it as a guided tour through a prison.
Thank you. Somewhat reminds me of Harry Chapin's 1978 song "Flowers are Red". Since high school English has become reading propaganda (name your poison) and encouraging ChapGPT, yes, let it fall into the abyss of uselessness. Then also, work on its renewal as a useful art, not a robotic tool.
Everyone has been focused on the students using ChatGPT to cheat, but that's like worrying that the invention of cars will startle the horses. This will transform the classroom experience.
I've been using ChatGPT to learn Japanese through conversation. I can definitely see teachers incorporating this technology into their classes. The model can suggest vocabulary words, discuss the cultural context of those words, create a quiz, accurately grade my answers, provide a dialogue written in Japanese, and give feedback as I translate each line of the dialogue.
Currently, I do have to watch for the BS and search for errors, but I foresee a (near) future where AI models like this are an essential tool for teaching and learning.
Your original tweet thread knocked me out and this post is a wonderful followup. Really enjoying WHY THEY CAN'T WRITE. Thank you!