You'll get that, just what you desire. For what you want is nothing special. So be confident--give yourself over to our agents here.
Aristophanes, The Clouds
I was hoping to post the second installment of my discussion of AI and writing sooner rather than later, as I’m scheduled for cataract surgery this week, but that’s looking increasingly unlikely. So as you read this, I’ve already had surgery on my left eye, with my right eye scheduled for a week or so from now.
For this installment, I want to weave a few things together, starting with John Gallagher’s recent post on the subject. One of the problems with AI-simulated1 writing, he notes, is that these platforms are fully capable of producing “meaningless and boring ideas wrapped up in mechanically correct prose.” Unfortunately, there are a lot of people (my own colleagues in other departments included) who understand writing’s highest virtue to be the absence of mechanical error. I hardly ever hear such colleagues complain, for instance, that their students aren’t thinking deeply enough; instead (and often), they insist that “My students can’t even write a proper sentence,” and proceed to blame this on our writing courses. To be fair, for a long, looong time, responding to student writing was little more than circling errors—the dreaded red pen—so those of us who teach writing aren’t immune from critique, either. We trained generations of students to believe that “good writing” was that which is technically correct. From that perspective, writebots would sound like a godsend and “just what [they] desire.”
Gallagher notes that “What we need, then, is a language about the quality of ideas,” and he spends the rest of his post (which I recommend) offering the term discernment for our consideration. “Discernment,” he explains, “means examining a text and slowly thinking about the meaning of the sentences and the quality of the ideas expressed in those sentences.” It also speaks to our ability to place that text in its social and intellectual context (and understand how the context may change the expectations we have for it). Finally, it requires a classroom practice where teacher and students alike are committed to clear discussions about the qualities of a given piece of writing.
One of the ways that he characterizes this practice is an “openness to criticism,” and that was the passage that suggested its link to what I wanted to write about. So I want to take up the challenge of “a language about the quality of ideas,” and suggest a term that I’ve been thinking about lately: probity.
The Craft of Writing
I need to lead up to a discussion of probity, though, and provide a better sense of what I mean when I use that word. So I want to start first with an old book, David Pye’s 1968 book The Nature and Art of Workmanship. In a very basic sense, Pye’s book is an attempt to articulate the virtues of craftsmanship against a world increasingly given over to the products of mass production; at one point, he contrasts the museum and the department store, for instance. The items in these spaces are the products of two very different motivations, the workmanship of risk in the former case, and workmanship of certainty in the other. I don’t want to dwell too long with this binary—as Pye makes clear almost immediately after introducing it, it’s not a clear-cut opposition but more of a spectrum. They are “approximations” at best. For Pye, craftsmanship is a workmanship of risk:
it means simply workmanship using any kind of technique of apparatus, in which the quality of the result is not predetermined, but depends on the judgment, dexterity and care which the maker exercises as he [sic] works. The essential idea is that the quality of the result is continually at risk during the process of making…(20)
The contrasting term, workmanship of certainty, is typical of mass production, where the goal is to generate indistinguishable copies (often as quickly and cheaply as possible). There are some folks who are critical of this distinction: they argue both that craftspeople develop tools and techniques to increase certainty, and that even in automated manufacturing, there are all sorts of risk vectors. Both things are true, especially the latter, where the demands of insatiable corporate greed almost guarantee a certain degree of risk, whether it’s sitting in an exit row on a Boeing plane or tucking into a sandwich made with Boar’s Head deli meat.
But the distrust of Pye’s terminology sort of misses the point. He’s not arguing for a return to an idyllic world where mass production didn’t exist, nor is he necessarily thinking about risk in everyday terms. My sense is that he’s arguing for a world that isn’t reduced only to what can be imagined by a mass-producing workmanship of certainty.
The danger is not that the workmanship of risk will die out altogether but rather that, from want of theory2, and thence lack of standards, its possibilities will be neglected and inferior forms of it will be taken for granted and accepted (23).
As I put it (drawing on Robin Sloan) last post, “At best, we will get only what we ask for.” As Pye explains, it’s not that the department store isn’t full of products that have been designed and produced with all sorts of ingenuity, it’s that the museum shows us more. There’s nothing wrong with a home that looks like an Ikea showroom, for example, but if that’s the full extent of what we’re capable of imagining for ourselves, then we’ve lost something. (Walter Benjamin, in another essay about “the age of reproduction,” called this “aura.”) Craftsmanship doesn’t guarantee quality—as Pye himself notes, “In many contexts it is an utter waste of time” (23)—but what craftsfolk “risk” to achieve quality is that effort, time, and skill, in the hopes that things will turn out better than they could have planned. It’s the possibility of getting more than we ask for.
Intransitive
There’s another essay I want to fold in here, one that takes me a bit closer to writing itself, and that’s Roland Barthes’ “To Write: An Intransitive Verb?” Barthes’ essay is about the shifting relationship (again, back in the 60s) between literature and linguistics, and I’m going to skip over most of it. Towards the end of the piece, though, he ruminates on a cultural shift, one that treats “writing” as an intransitive verb rather than transitive. In other words, while we still use the verb in the latter sense—we write memos, letters, emails, novels, poems, screenplays, et al. —that intransitive sense of the verb (for Barthes) says something specific about the practice of writing.
Barthes identifies this as an “important change in mentality,” although it’s subtle enough that most won’t pay it attention. I tried to get at this a bit in my comparison of writing to exercise: we don’t walk into a gym, hop on a treadmill, or attend a spin class thinking about the total weight or distance involved—working our muscles is the point, not any sort of “use” that we put them to. Barthes makes the connection between intransitive verbs and the middle voice (as opposed to active or passive). “In the case of the middle voice,” he explains, “the subject affects himself [sic] in acting; he always remains inside the action, even if an object is involved…it is to leave the writer inside the writing, not as a psychological subject, but as the agent of the action” (142). We have a term for this idea of being “inside the action,” although it’s more often associated with physical exertion (or athletics)—we talk about athletes finding the “zone,” that state of immersion so intense that we lose sense of time.
Part of what falls away when we find that zone is the attachment to a particular outcome, and for me, that’s the connection between Barthes’ intransitive, middle voice and Pye’s workmanship of risk, where the result is not predetermined. This is one of the most difficult pieces of (teaching) writing, the idea that writing isn’t just something we do with/to language, but something that language does through us3. As soon as we arrive at the second or third word in a sentence, our understanding of it (as writers) is already being shaped by what language allows, even if we decide to resist it. Language is technology and tool, certainly, but it’s also the material that writers manipulate, affecting us as we act with and upon it.
This quality, I’ve found, is difficult to characterize, especially with those who don’t particularly enjoy writing. It’s easy to tail off into something that sounds quasi-mystical or meditative, sort of like a “runner’s high” but with writing. Pye and Barthes get me close, but I want to turn to one more text, which is where I take the term probity from, and that’s Jean-François Lyotard’s Peregrinations.
Probity
If you’re looking to find an idea that’s less mystical, it might feel counter-intuitive to turn to the notion of clouds, but they’re a metaphor that runs through Lyotard’s book. Peregrinations is a very short text, the transcript of a brief series of lectures he gave back in the 1980s. Among other things and some pretty dense philosophical work, Lyotard reflects on the relationship between thoughts and ourselves, the practice of thinking:
Thoughts are not the fruits of the earth. They are not registered by areas, except out of human commodity. Thoughts are clouds. The periphery of thoughts is as immeasurable as the fractal lines of Benoit Mandlebrot. Thoughts are pushed and pulled at variable speeds. They are deep, although core and skin are of the same grain. Thoughts never stop changing their location one with the other. When you feel like you have penetrated far into their intimacy in analyzing either their so-called structure or genealogy or even post-structure, it is actually too late or too soon. One cloud casts its shadow on another, the shape of the clouds varies with the angle from which they are approached (5).
We don’t “have” thoughts, but engage with them in particularly tentative ways. “Accordingly,” he explains, “thoughts are not our own. We try to enter into them and to belong to them. What we call the mind is the exertion of thinking thoughts” (6). I’m probably already dipping too deeply into this particular cloud, but I’ll note one more thing, and that’s that Lyotard insists here (“too late or too soon”) on the importance of time. “Time,” he writes, “is what blows a cloud away after we believed it was correctly known and compels thinking to start again on a new inquiry” (7).
Lyotard refuses a negative, skepticist interpretation of this state of affairs. For him, this isn’t a matter of throwing our hands up and despairing of being able to know anything with certainty or permanence. Rather, “it inspires only the principle of the endless pursuit of the task of discussing clouds.” He goes on to explain for this pursuit, the arts and humanities are just as relevant as science and technology. There’s no discipline that has a monopoly on thinking.
It’s in this context that Lyotard introduces probity,
“…an ability to be responsive to slight changes affecting both the shape of the clouds you are trying to explore and the path by which you approach them. Imagine the sky as a desert full of innumerable cumulus clouds slipping by and metamorphosing themselves, and into whose flood your thinking can or rather must fall and make contact with this or that unexpected aspect. Probity is being accessible to the singular request coming from each of the different aspects. It is a sensitivity to singular cases” (8).
In many ways, Lyotard describes probity in a fashion that resembles Barthes’ intransitive, middle voice. Because the clouds shift with time, we are never able to “master” the paths that we trace out among them. At the same time, “we are not entirely determined” by the call to engage in that tracing; there is “space for hesitation, doubt, and criticism.” Insofar as there’s an ethic to thinking for Lyotard, he speaks of it as a commitment to a “je ne sais quoi,” or as Pye (and Bruno Latour4) might characterize it, an acknowledgment (a risk) of uncertainty, a commitment to tracing through and among clouds of thought where the route is not predetermined.
Although I prefer5 the term probity and will stand by it, we might also think of it as openness. I tend to understand probity as an overarching quality that includes several of the “openings” that I’ve gathered here. Writing, as I’ve tried to detail, includes being open to the change that happens as one writes. Writing is not the execution of a predetermined script or the transcription of pre-existing thoughts; it’s the practice of thinking through language. It also entails holding oneself open to language, and the ways that our thoughts are shaped by the language that we use to express them. It involves accounting for one’s audiences (both immediate and distant, and often changing), acknowledging that sometimes, the best we can hope for is to bridge the gaps in understanding that come from our differences (and risking failure in those efforts). There’s more to it, of course, but for me, probity captures a great deal of what I understand to be the proper attitude towards writing.
“Writing” in the Cloud
I’m not so impractical as to expect that anyone (myself included) thinks about these things every time they sit down to write. I don’t allow myself to be seized by the glorious chaos of language, for example, if I’m signing an office birthday card, filling out intake forms for yet another medical specialist, or navigating the bureaucracy of higher ed. At the same time, activities like those, and their intrinsic tedium, do not represent any sort of baseline for what writing actually does (and what we as writers do), despite the very best efforts of AI-shills to encourage us to think so.
The outputs of these systems is not writing but its simulation, even if it’s occasionally capable of fooling us into thinking the contrary. A couple of weeks ago, Rob Horning wrote about how they’re not engaged in artificial intelligence but artificial intentionality:
Many of AI’s use cases are predicated not on producing some specific content so much as on generating the illusion that human attention has been paid to something, so that a person can take credit for something they did not do….the machine’s contribution is intrinsically worthless in any task that draws on human connection or presence for its significance, anything that connotes free choice.
If Lyotard uses clouds to stand in metaphorically for writing and thinking, AI companies offer us the inverse “bargain,” transforming writing into a metaphoric husk that disguises what happens in the algorithmic cloud, the generation of statistically likely recombinations of others’ words, what Max Read describes as slop. But as Read notes in his piece for New York magazine, it’s not just the companies (or the platforms) themselves. There are people behind the pollution: “Beneath the strange and alienating flood of machine-generated content slop, behind the nonhuman fable of dead-internet theory, is something resolutely, distinctly human: a thriving, global gray-market economy of spammers and entrepreneurs, searching out and selling get-rich-quick schemes and arbitrage opportunities, supercharged by generative AI.”
Those are the people, the ones who couldn’t give two shits about writing or thinking, who are driving the hype around AI (and sending me emails trying to convince me to flip our students into consumers of their platforms), about whom I try to remind myself, every time I’m tempted to engage in the sort of shortcutting (and slop-piness) that AI promises to me. Writing is thinking, and neither is something we need less of right now.
There’s plenty to more to say, but I’m hitting a busy stretch, so I’ll sign off here. My friend Annette is also visiting campus early next week to talk about AI, so I’ll probably have more to say once they fix up my other eye. More soon.
Maybe someday I’ll get into this, but others have already: there’s a specific sense in which AI “generates” text, but what it produces is more accurately described as a simulation of writing, and we cheapen the value of writing itself (and the “glorious chaos” I cited in my last installment) when we use AI companies’ language to describe what these tools do. So I’m trying to wean myself off of “generative AI” permanently.
This passage resonates for me with Gallagher’s “language about the quality of ideas.”
Nietzsche’s remark that “we are not rid of God because we still have faith in grammar” is a claim about the extent to which we are spoken by language.
I very nearly almost included here a section on Bruno Latour’s articulation of writing, which changed my own attitudes almost 20 years ago. His Reassembling the Social (2005) lays out five “Sources of Uncertainty” that complicate social theory as it’s widely practiced, and he characterizes (good) critical writing in terms of “risk” (see esp. 128-133). It wasn’t until I went back and looked at Pye that the overlap in vocabulary called itself to my attention (Latour doesn’t cite Pye as far as I know).
This is a deep cut, but my first book renamed/recast the rhetorical canons (invention, arrangement, style, memory, and delivery) for our current era of technology, and I used P-words to do so (proairesis, pattern, perspective, persistence, performance). For a good while, I thought about using probity rather than the admittedly more obscure proairesis. I had my reasons, but it was a close thing.
Cool! I didn't expect that post to get read. :-)
Probity is a helpful frame. And I appreciate that you demarcate the kind of thoughtful writing an essay like this entails from the kind of writing that keeps the machines running. That's where I think <a href="https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art">Ted Chiang's recent piece</a> misses the mark. And that's part of what complicates AI and writing for me. The devaluation of writing has already happened, and yet, there's still something different about what AI has done--that simulation of thought, as you say. The other kind of writing--the wheel-greasing kind--never pretended to be thoughtful.
I already loved this, Collin, and then I saw you mention me--thanks for keeping me in great company here with your other references. :)