Writing

ChatGPT for Writing Teachers: A Primer

preterite Artificial Intelligence, Pedagogy, Rhetoric, Technology, Writing
or, how to avoid writing like a machine
Background

At this year’s Conference on College Composition and Communication in Chicago, there was a lot of interest in generative large language models (LLMs), or what the popular media more crudely dub AI, or what many today metonymically refer to (like calling photocopies Xeroxes or sneezepaper Kleenex) as ChatGPT. I first played with an earlier version of the LLM, GPT-3, at about the same time I started playing with neural network image generators, but my interest in language and computing dates from the early 1980s and text adventure games and BASIC, to hypertext fiction and proto-chatbots like Eliza, and to LISP and early prose generators like Carnegie Mellon’s gnomic and inscrutable Beak—and also to the arguments I heard John Hayes express in Carnegie Mellon’s cognitive process Intro Psych lectures about how we might try to adjust human neural processes in the same ways we engineer computing processes. That idea is part of what makes ChatGPT and other generative neural networks appealing, even when we know they’re only statistical machines: thinking about how machines do what they do can help humans think about how we do what we do. ChatGPT offers a usefully contrastive approach for reconsidering writing and learning. So it’s worth understanding how it operates. With that desire, and having read devoured lectitaveram everything I could find on the topic, I went to a CCCC presentation and was only mildly and briefly disappointed, given that I was not (as should have been obvious to me from the outset) the target audience.

Here, then, is my attempt at writing an alternate what-if presentation—the one I’d half-imagined (in the way working iteratively with ChatGPT or MidJourney gradually gets one closer to what one didn’t know one was imagining—OK, you see what I’m doing here) I’d learn from in Chicago. And I’ll offer the combination warning and guilty plea up front:

Read more

More from ChatGPT

preterite Artificial Intelligence, Computers, Conference Notes, Writing

Second in what will probably become a series. I recently came back from the Conference on College Composition and Communication (CCCC, or 4Cs) in Chicago, where the organizers put together a panel on ChatGPT that indicated that our institutional memory is better than I’d feared—panelists remembered their Cindy Selfe, though unfortunately not their Doug Hesse. Short version: I was probably the wrong audience for the panel, and I think they did a solid job, though I would have wished for more depth. It was helpful to me in that I made some connections after the Q&A, and the panel also helped me imagine the panel presentation I’d hoped to see, so I’ve been working on a long-read semi-technical ChatGPT explainer with implications for composition instructors that I’ll post here in the next few days. The strongest parts of the panel were those dealing with direct pedagogical applications of ChatGPT. I wonder, though, what Peter Elbow might say about ChatGPT and “closing my eyes as I speak,” since ChatGPT effectively removes one element (the rhetor or writer) from the rhetorical triangle, productively isolating the other two elements (audience and message) for analysis of how they interact. What sorts of rhetorical experiments might we perform that would benefit from reducing the number of variables to analyze by entirely dismissing the possibility of authorship and rhetorical purpose?

Hat tip, by the way, to Clancy Ratliff for proposing the Intellectual Property Caucus resolution on Large Language Model (LLM) AI prose generators like ChatGPT at the CCCC business meeting: seconded by me, and passed by overwhelmingly affirmative vote. The statement: The Intellectual Property Standing Group moves that teachers and administrators work with students to help them understand how to use generative language models (such as ChatGPT) ethically in different contexts, and work with educational institutions to develop guidelines for using generative language models, without resorting to taking a defensive stance.

Read more

GPT-3 Gave Me This Today

preterite Computers, Technology, Writing

“There is something in the telling of our lies that can redeem us, can make us better than we are. We see Abraham Lincoln at Gettysburg battlefield, with his son’s body on a stretcher before him, his hand on the boy’s head, his eyes cast down, the sound of the artillery in the distance like thunder, or like the beating of a great heart, and Lincoln says, This world does not belong to the strong.”

https://beta.openai.com/playground

Metadata and the Research Project

preterite Information, Politics, Research, Technology, Writing

In a widely reported quotation, former director of the NSA and CIA General Michael Hayden said in May 2014 that “We kill people based on metadata.” Metadata is increasingly valuable today: it would also seem that it carries not one but multiple forms of value, some of those forms payable in blood.

Information Scientist Jeffrey Pomerantz, in his book Metadata (Cambridge, MA: The MIT Press, 2015), argues that until recently, the term “metadata” has typically been used to refer to “[d]ata that was created deliberately; data exhaust, on the contrary, is produced incidentally as a result of doing other things” (126, emphasis mine). That’s an interesting term, “data exhaust,” as perhaps an analogue to the pollution associated with the economic production and consumption of the industrial age. And of course corporations and governments are finding new things to do with this so-called data exhaust (like kill people, for example, or just to chart the social networks of potential insurgents like Paul Revere, as Kieran Healy charmingly demonstrates, or even to advertise Target products to covertly pregnant teenagers until their parents find out, as the anecdote popular a while back noted). It’s got cash value, click-through value, and my Digital Technology and Culture (DTC) students last semester put together some really terrific projects examining the use of cookies and Web advertising and geolocation for ubiquitous monitoring and monetizing.

But that idea of useful information as by-product keeps coming back to me: I wonder if someone has ever tried to copyright the spreading informational ripples they leave in their wakes as they travel through their digital lives, since those ripples would seem to be information in fixed form (they’re recorded and tracked, certainly) created by individual human activity, if not intention. There’s a whole apparatus there that we interact with: as Pomerantz notes, “[i]n the modern era of ubiquitous computing, metadata has become infrastructural, like the electrical grid or the highway system. These pieces of modern infrastructure are indispensible but are also only the tip of the iceberg: when you flick on a lightswitch, for example, you are the end user of a large set of technologies and policies. Individually, these technologies and policies may be minor, and may seem trivial. . . but in the aggregate, they have far-reaching cultural and economic implications. And it’s the same with metadata” (3). So the research paper has as its infrastructure things like the credit hour and plagiarism policies and the Library of Congress Classification system, which composition instructors certainly address as at once central to the research project and also incidental, because the thing many of us want to focus is the agent and the intentional action; the student and the research. Read more

The Syllabus as Ossuary

preterite Composition Pedagogy, Pedagogy, Writing

The common and ongoing complaint is that first-year composition (FYC) is a repository of dead forms. In composition’s associated disciplines in English studies, critical examinations of writing and reading technologies ossify into periodized media studies, and in first-year composition, radical experimentations in how college students continue to learn to write well become the formeldahyde frog in the wax-backed metal tray from Biology 101, its belly razored open and skin peeled back so that students might safely identify the intestines, kidneys, heart, and probe around inside, perhaps a little grossed-out by the process, but able to name its components and mark them on a final quiz.

The formeldahyde frog masquerades as object of inquiry, even inasmuch as everyone knows that the annual and ongoing mass death of millions of appropriately-sized frogs serves only the purposes of a school exercise that will be swiftly forgotten. The research essay in its current commonly accepted form is the frog with its belly-flaps pinned back, poked around upon in JSTOR and ProQuest and the Library of Congress subject and keyword headings like well-preserved amphibious digestive and evacuative systems investigated by the earnest and industrious student, indicating little more to that student than this is where food goes in and this is where poop comes out.

To shift metaphors: the research essay assignment is pedagogy as archaeology. In the information age, I am largely in agreement with the common and ongoing complaint about first-year composition pedagogy and dead forms, especially as that complaint indicts the research essay. As much as anyone else, I am guilty of teaching the dead form, the corpse of the beloved, knowing all too familiarly the workings of the forms of library research I insist to myself that students must know. Even if I frame the research assignment as “inquiry” or “documented argument,” even if I congratulate myself on helping students to see that writing research means something beyond the assemblage of regurgitated stale quotations about innovative environmental applications for hemp and cannabis ash or the burial habits of ancient Egyptians, I am still simply trying to animate a cadaver or vivify a golem, making the body of my own knowledge do what I want, and inflicting that upon the students in my class.

Yes, but: Doesn’t it operate as an introductory form? Doesn’t it do work that helps prepare students for other more sophisticated tasks? Doesn’t it help alert students to modes beyond Google of navigating our rapidly-expanding tombs of information?

It could. I wrote about this challenge — about the essay as database, the database as essay — in 2007, but I’ve been thinking it about it since 1998, when I was working on a Microsoft Access database during my day job and taking an evening research methods seminar with another young graduate student named Becca, who had a complex journalistic research project she was undertaking and was looking for a way to manage it as part of her class project, and I suggested building a database. I don’t know if she took my suggestion, but that woman was Rebecca Skloot, whose research project became The Immortal Life of Henrietta Lacks. Part of what’s so impressive to me about Rebecca’s book is that it attends deeply to research as an evolving process: she talks very carefully about how she’s doing it. I’d like to see more of what Becca does in the first-year composition research project assignment.

My FYC students begin their annotated bibliography essay tomorrow, their second essay assignment, as a lead-in to their third, which is ostensibly the research paper assignment. I love the perspective I heard from a colleague yesterday, who posed the annotated bibliography as edited collection, complete with introduction and conclusion: yes, I said, that’s it. That’s the production of new knowledge, focused enough to be interesting, acknowledging its antecedents, edgy enough to push the boundaries. I’ve been reading a lot about information these past few years, and the idea I keep returning to is that information is the work and process of building itself, and as the asset itself that gets exchanged, aggregated, built upon. Information, and the work of research, is labor become capital.

Security Dreams

preterite Family, Teaching, Writing

I probably shouldn’t do class-related readings on the NSA and information security right before going to bed. The past two nights have been a combination of first-week anxieties and stuff related to Bruce Schneier’s Data and Goliath and Frank Pasquale’s The Black Box Society, the latter of which is one of the books I assigned for Digital Technology and Culture (DTC) 356, Electronic Research and the Rhetoric of Information, plus some weird house- and family-related stuff.

In Monday night’s dream, I’m wandering in an abandoned, crumbling neighborhood in late afternoon, the facades of houses caved in, burned and abandoned cars lining the streets, a smoky haze in the sky like what’s been visiting Pullman, dimming the sun. I go into one of the houses and it’s filled with irregular but oddly assembled debris: Read more

Reading and Error

preterite Writing

I’m in the final work of editing down an article to potentially publishable brevity, and there’s a moment where I realize there’s an absent connection, a place where it switches tracks and goes into a weird place. It’s an article that deals in part with reading and self-awareness while reading, and I’d like to leave that moment of confusion in  there.

It makes me think of student error in composition papers, and how composition instructors read student papers. When do we notice that things get weird, and why? And what does it signal to us?

Imagine you’re in a small seminar-sized professional development meeting for first-year composition instructors, and you’re talking about how to read student papers, and the topic of error comes up. What if, instead of having the instrumental conversation, somebody at the meeting handed out a photocopy of the Charles Kinbote Foreword to John Shade’s Pale Fire, as represented in the Nabokov novel of the same name? What if you took turns reading it out loud, together, for the half hour or forty minutes or so that it took, and you agreed to use a pencil and mark only the moments where things got weird for you as a reader?

Do you characterize those moments as places that are problematic, or places that are interesting?

Oh, but some will say, composition students don’t know what they’re doing like Nabokov did. My response: what about Kinbote? And isn’t part of the fascination looking at what Kinbote does? And might that not help us think about how we might be more fascinated by students?

I want to leave in the place where it switches tracks because it makes a moment of difficulty right where I’m talking about making moments of difficulty. That’s a big leap to make, though: to ask readers to say, “Wait, what?” and still ask them to go back to it.

Seeking Feedback on an Algorithmic Poem

preterite Writing

I’m working on a presentation and would welcome some help. I wrote a poem, and am well aware that it’s a bad poem in any number of ways. I’m OK with that.

Here’s the help I would like: please look for a single line that interests you. It can be a line that’s terrible, a line that you like, a line that does something you find engaging or stupid or funny or terrible or exciting in whatever way.

Find all the stanzas in which that line occurs. In comments, enumerate the stanzas in which the line occurs. (For example: I, II, IV.)

Don’t tell me what the line is: I’ll actually tell you what the line is in response to your comment. (Yes, in this way, I’m asking you to help me perform an online parlor trick, with a poem.)

I’ll tell you more about this once I figure out if it works, but here’s the short version: poems can be computers. Help me out? Poem follows.

Read more

The Forensic Imagination and the Commodification of Process

preterite Composition Pedagogy, Composition Theory, Computers, Economics, Writing

In his discussion of William Gibson’s Agrippa, Kirschenbaum notes that “while the title _Agrippa’s_ immediate referent is to a brand of photograph albums, it also hearkens back to Renaissance mage Heinrich Cornelius Agrippa von Nettesheim” (230), and while the reference is appropriate, it apparently overlooks what I believe is a much more relevant reference: Marcus Vipsanius Agrippa and his son Agrippa Postumus, so named because he was born after the death of his father. Agrippa the senior was elaborately memorialized by Augustus, while Agrippa Postumus was executed following the death of Augustus, and his step-father Tiberius became emperor. These figures seem much more closely connected to the subjects of Gibson’s ephemeral poem in their representation of father-and-son relationships and in their relation to memorialization.

But perhaps such contestation is part of the point Kirschenbaum makes: a reading is always only ever a reading, informed as much by the reader’s material and social and historical contexts that she brings to the reading as by the forensically unique allographic textual artifact itself. On page 185, Kirschenbaum uses a screenshot of multiple windows running different electronic versions of Michael Joyce’s _Afternoon_ to demonstrate how digital texts are not purely virtual, and so shows us what revision means, in its re-use, re-reading, and re-attending to a text from a position located within and conscious of a particular material context. Revision is always situated in a kairotic moment. In Kirschenbaum’s words, “formal materiality. . . serves to fetishize via the computational distance (or torque, or simply effort) necessary to. . . access certain objects in certain ways. In my own case, the first time I successfully opened a first edition of _Afternoon_, I was exquisitely self-conscious of something very much like bibliophilia, precisely because I had to couple the file itself with the right Macintosh operating system and the right version of Storyspace, thereby imposing a formal regimen on the binary object that was _Afternoon_, which then led it to execute, consume system resources, and ultimately present itself for my inspection and manipulation. This kind of access and recovery will, I suspect, ultimately prove more enduring th[a]n a collector or connoisseur’s sensibility, which seeks to acquire and possess” (186). If formal materiality is effort or work, Kirschenbaum’s example also demonstrates that it can be pleasure, as well. It’s both the process and the kairotic/phenomenological moment of the experience of a text that remediates it and reforms/performs/deforms it within a specific material context, to and from which there are specific material and textual inputs and outputs that negotiate between different levels of textual, social, and technological systems. In other words, the process Kirschenbaum describes is economic: value and labor are circulating, and in texts just as in computers, “[v]ersioning. . . exposes the cumulative labor that attends a piece of software” (202). The process is an instance and an example of the economic aggregation problem, by which we cannot measure all the inputs and outputs of any economic activity.

This is what happens, then, “whenever process collapses into product” (Kirschenbaum 253): the forensic imagination takes the meaning of a text as its material form and that form then takes on secondary meaning and value in its aestheticization and commodification. Such a move is also performed by the corpus of composition pedagogy (in its theorized condition) does.

Courseblogging Machine and Meaning

preterite Computers, Writing

After a summer of upheaval, I’m starting to get settled into the new gig. I’m excited about both courses I’m teaching, and I’m keeping a courseblog with my students for one of them, an undergraduate elective (DTC 356) titled “Electronic Research and the Rhetoric of Information.” It’s interesting: I get to look again at material and concepts I’ve become pretty familiar with in the past 10 years or so, stuff I have some ostensible expertise on and that I’ve been thinking about for a while and that I know other scholars in the field have considerable familiarity with, and yet this is the first chance I’ve had to teach a course like this — and so my courseblogging feels like a weird mix of old material, new insights, and responses to re-framings I hadn’t considered before. That’s a good thing, and I’ll post now (and continue to cross-post) some of my entries for the course, as a way to continue getting settled into the routine of the new gig. Plus I’ve got about eight billion thoughts about the big thing I’m working on that I want to share, and there are only tiny corners of it here, but that’s OK: there’s time.

So in thinking about recent applications of the Labor Theory of Value to the so-called information economy, one of the questions I posed to the students in DTC 356 was: how much of a role does effort play in how we interact with digital technologies? (Cross-posting begins here; longtime followers of this blog will notice the change in intended audience in relative degrees of explicitness.) In one DTC356 blog post, a student wrote,

When I think of a world without the social media and technology we have now, I imagine a world that was connected in only a few ways instead of a million ways (twitter, facebook, blogging, etc.) to communicate with each other. Could you imagine having to listen intently to clicks or beats? Technology would not have ever advanced as far as it has today if it weren’t for these signals, tones, and phrases that began centuries ago.

The point about “having to listen intently” is important, because of the ways digital technologies seem to make communicating information so easy. Brown and Duguid talk about “the conduit metaphor” and how “[b]asic ideas of sending and receiving make digitization, for example, seem easy. You distill the information out of book or articles and leave the paper residue behind” (184). The problem is, though, that there are other important aspects of the act of communication that we often ignore: as Brown and Duguid go on to point out, “[i]t’s not pure information alone, but the way the information was produced that supports interpretation” (185). This is what Lessig is getting at in his discussions of the borrowings of Steamboat Bill, Jr. and doujinshi, and what we were getting at in our discussion of cover songs and Girl Talk: so much of information is context. You don’t fully appreciate a cover version of a song unless you’ve heard the original (think about the 33,000+ covers of Gershwin’s “Summertime”), and part of the reason that Disney movies resonate so much (as Lessig suggests) is that they’re built on stories that our culture knows really, really well; stories that resonate with us. (Why so many Batman and Spider-Man movies, right?) So there’s this ideal that we have of some sort of pure, easily transmitted information — just a few 1s and 0s to decode, and if you know about logarithms and exponents, you can derive meanings from tables of numbers that others might not be able to see — but that ideal isn’t actually the way things work.

Information transmission isn’t, in fact, efficient. That’s the point of the story about talking drums (“allocate extra bits for disambiguation and error correction” [Gleick 25]) and the story about Clytemnestra receiving word of the fall of Troy 400 miles away in Mycenae: “To transmit this one bit required immense planning, labor, watchfulness, and firewood” (Gleick 16-17). Transmitting information is expensive, in terms of labor and in terms of capital — and in an information economy, context is kind of like capital. (Actually, in terms of the factors of production described by the old political economists like Adam Smith, David Ricardo, and Karl Marx, context is probably more similar to land than to capital.) One student asks, “Could you imagine having to listen intently to clicks or beats?” and of course that’s what we all do, all the time.

That’s also what computers do with 1s and 0s (true and false, high and low, fire or no fire). Computers use logic gates with transistors designed to let current through in certain ways and control other transistors, so that combinations of transistors with combinations of current going on or off through them according to how they’re designed to work in conjunction with each other — to signal AND, OR, or NOT, as well as more complex combinations like NAND, NOR, XOR, and XNOR — build up, store, and manipulate more complex numbers out of simple 1s and 0s. And because information-as-capital builds upon itself, computers have been able to get increasingly complex while their prices have dropped. Context builds on itself, and technology is a part of context. As Lessig points out, there was once a “distinction that the law no longer takes care to draw — the distinction between republishing someone’s work on the one hand and building upon or transforming that work on the other. . . Before the technologies of the Internet, . . . [t]he technologies of publishing were expensive; that meant the vast majority of publishing was commercial. Commercial entities could bear the burden of the law. . . It was just one more expense of doing business” (19). Now, though, because our technological context has become increasingly complex and avaiable to all, we’re all increasinly bearing that “burden of the law” and having to figure out how to revise our own social, legal, and political contexts to account for that increased complexity. Doing so requires not only attention to the alphabet and syntax and orthography and grammar of these 1s and 0s but also to the rhetoric: in moving from the high and low tones of the drums and the morse code of the battleship’s signal lamp to the mashup video of “Oh No” (if there had been a clip of Michelle Obama dancing to “Teach me how to Dougie” in that video, would it have been in the public domain?), we need to think about a rhetoric of remix wherein inventio is the current and the initiating spark, dispositio is the linking of gate to gate, elocutio is the purposing of the gates themselves whether NAND or NOR, memoria is the storage of what those gates arrange to produce, and pronuntatio is the moment of its transmission: the interface between machine and meaning.