Alcove

Science, culture, complexity

Tag: Mary Norris

  • So you use AI to write…

    You’re probably using AI to write. Both ChatGPT and Google AI Studio prefer to construct their sentences in specific and characteristic ways and anyone who’s been a commissioning editor for at least a few years will find the signs of its influence hard to miss, even if you personally think it’s undetectable.

    You’re probably going to continue using AI to write. There’s nothing I can do about that. In fact, as AI models improve, my ability to prove you used AI to write some text will only decline. And as a commissioning editor focusing articles about science, health, and environment research, I’m also not entirely at liberty to stop working with a writer if I notice AI-written text in their reports. I can only warn them. If I threaten to stop, I’m certain that one day I’ll have to impose the ultimate sanction, and after that I’ll have to look for new writers from a pool that’s already quite small. That’s a bad proposition for me.

    To be sure, I realise I’m in a difficult position. People, especially those without a good grasp of English writing as well as those without the means to attain that grasp, should have the opportunity to be understood as precisely as they wish to be among English-speaking readers. If immaculate writing and grammar also allow their opinions to reach audiences that shunned them before for being too difficult to understand, all the more reason, no?

    This has been something of an issue with the way The New Yorker wields its grammar rulebook, with its commas, hyphens, accents, and umlauts in just the right places. As the author Kyle Paoletta disputed in a 2017 critique in The Baffler:

    For The New Yorker, a copy editor’s responsibility to avoid altering the substance of a writer’s prose is twisted into an utter disinterest in content writ large. … Content must be subordinated—thoughtfully, of course!—to the grammatical superstructure applied to it. Not only does this attitude treat the reader as somewhat dim, it allows the copy editor to establish a position of privilege over the writer. …

    [Famed NYer copy-editor Mary] Norris frets over whether or not some of James Salter’s signature descriptive formulations (a “stunning, wide smile,” a “thin, burgundy dress”) rely on misused commas. When she solicits an explanation, he answers, “I sometimes ignore the rules about commas… Punctuation is for clarity and also emphasis, but I also feel that, if the writing warrants it, punctuation can contribute to the music and rhythm of the sentences.” Norris begrudgingly accepts this defence, but apparently only because a writer of no lesser stature than Salter is making it. Even in defeat, Norris, as the tribune of The New Yorker’s style, is cast as a grammatical arbiter that must be appealed to by even the most legendary writers.

    I shouldn’t stand in judgment of when and how a writer wishes to wield the English language as they understand and have adopted it. But with AI in the picture, that could also mean trusting the writer to a degree that also overlooks whether they’ve used AI. Put another way: if it’s levelling the playing field for you, it’s getting rid of another sign of authenticity for me.

    Perhaps the larger point is that as long as you make sure your own writing improves, we’re on the right track. To be clear: whose English should be improved? I’m saying that’s the contributor’s call. Because if they send a poorly written piece that makes good points, some of what I’m going to be doing to their piece is what AI will be as well (and tools like Grammarly have already been). Second, as AI use becomes more sophisticated, and I become less able to tell original and AI-composed copies apart, I’ll just have to accept at some point that there’s nothing I can do. In that context, all I can say is: “Use it to improve your English at least”.

    Or perhaps, as in the words of Northern Illinois University professor David J. Gunkel, “LLMs may well signal the end of the author, but this isn’t a loss to be lamented. … these machines can be liberating: they free both writers and readers from the authoritarian control and influence of this thing we call the ‘author’. For better or for worse, however, I don’t see how a journalistic publication can adopt this line wholesale.

    * * *

    As a commissioning editor in science journalism specifically, I’ve come across many people who have useful and clever things to say but their English isn’t good enough. And from where I’m sitting, using AI to make their ideas clearer in English only seems like a good thing… That’s why I say it levels the playing field, and when you do that in general, the people who already have a skill because they trained to acquire it lose their advantages while the new entrants lose their disadvantages. The latter is of course unfair.

    In the same vein, while I know AI tools have been enormously useful for data journalists, asking a programmer what they feel about that would likely elicit the same sort of reaction I and many of my peers have vis-à-vis people writing well without having studied and/or trained to do so. That would make the data journalists’ use problematic as well by virtue of depriving a trained programmer of their advantages in the job market, no?

    Short of the world’s Big AI companies coming together and deciding they’re going to insert their products’ outputs with some kind of inalienable watermark, we have to deal with the inevitability of our being unable to tell AI-made versus human-made content apart.

    Yes, AI is hurting what it means to be human, at least in terms of our creative expression. But another way to frame all this is that the AI tools we have represent the near-total of human knowledge collected on the internet. Isn’t it wonderful that we can dip into that pool and come away as more skilled, more knowledgeable, and perhaps better persons ourselves?

    Science communication, as it focuses on communicating what scientists have found, might become redundant and low-value, at least among those that can afford these subscriptions (whose prices will also fall and accessibility will increase) because people can feed a model a paper and ask for the output. Hallucinations will always be there but it’s possible they could become less likely. (And it seems to me to be somewhat precarious to found our own futures as journalists on the persistence of hallucinations.) Instead where we can cut ahead, and perhaps stay that way, is by focusing on a journalism of new ideas, arguments, conversations, etc. And there the playing field will also be less about whether you can write better and more about what and how you’ve learnt to think about the world and its people.

    We all agree democratisation is good. Earlier, and to a large degree still, education facilitated and facilitates that to a significant degree. Now the entry of AI models is short-circuiting that route, and it’s presenting some of us with significant disadvantages. Yes, there are many employers who will try to take advantage of this state to bleed their workforce and take the short way out. But it remains that — setting aside the AI models coming up with BS and equivocating on highly politicised issues — the models are also democratising knowledge. Our challenge is to square that fact (which can’t be un-facted in future) with preserving the advantages for those that ‘deserve’ it.

    As if mirroring the question about who gets to decide who should improve their English, who gets to decide who gets to have advantages?

  • The new JNU VC’s statement has bad grammar. So?

    I strongly disagree with some criticism that has emerged on Twitter against the new JNU vice-chancellor Santishree Pandit. The object of criticism is a statement that Pandit has apparently drafted and in which she states, broadly, what she considers to be her mandate. In response, BJP MP Varun Gandhi wrote:

    Here are screenshots of tweets by two other people, both with a not-insubstantial number of followers on Twitter:

    At the outset, while Pandit deserves the criticism that has come her way for her use of abusive language on Twitter against the country’s students and farmers, using the quality of the English language in the statement to deride her is unfair. I have two reasons.

    First, listen to this talk (also embedded below) Pandit delivered in 2015: her diction is much better than her new statement would suggest. It suggests strongly that someone else wrote the statement and that Pandit simply signed off on it, as a formality.

    Second, even if we assumed Pandit wrote the statement, or that the criticism of Varun Gandhi and others were really to be directed at the statement’s real author…

    English is a difficult language to learn and use. Its grammar often has a mind of its own – typically in the form of what linguist Noam Chomsky has called opaque structures: turns of phrase that allow us to deduce nothing about their origins based on their composition itself (“trip the light fantastic” comes to mind; it means, of all things, to dance in a nimble way.)

    There are word and sentence constructions in English into which someone who doesn’t read, write and speak the language regularly is unlikely to ever stumble. As such, the language is part deduction and part memorisation (sort of like biology), and unless someone claims to wish to succeed Mary Norris or Mary Beard, or wishes to draft law, criticising a person’s flawed use of the English language can only amount to a criticism of their lack of access to English-speaking habits, circles, etc., and in turn a criticism of either their inability or their unwillingness to have this access. And this is not a sin. In fact, I suspect that the statement’s author is more fluent in a different language than in English and that if that person had penned their statement in that language, it would have been much less grammatically iffy.

    One may contend that the critics expect better from the vice-chancellor of JNU. What is this ‘better’? At the risk of affirming the consequent, let’s flip the argument such that it becomes: “A vice-chancellor of JNU must be able to string two sentences together in a grammatically correct way.” Why must this requirement be met?

    There is a presumption here, however slight, that the goodness of Pandit’s knowledge of the English language is an indication of her being unfit for the job, or more generally that it could be a proxy for the many, many things that count towards literacy, not to mention her prowess as a teacher and her familiarity with the subject matter. (Before her appointment, Pandit was a professor of political science at the Savitribhai Phule Pune University.)

    If the requirement must be met nonetheless, should we subject all future vice-chancellors to this ‘test’, to have them demonstrate their literacy? Should we also extend these tests to the heads of other important institutions – such as, say, S. Somanath of ISRO or health minister Mansukh Mandaviya? Pandit works at a university where many classes are conducted in English and where English is also an important language of administration – but this is easily true of ISRO and the health ministry as well. Most extant knowledge of space science, engineering studies, medical science and Indian public administration exist in English.

    Attacking Pandit’s grammar could in effect set up a requirement for vice-chancellorship that could easily say nothing at all about the appointee’s competency. Pandit’s statement expresses itself without confusion and, given the context, its real author likely didn’t have and/or enlist the assistance of other people who could fix the grammatical mistakes (it’s entirely possible they didn’t check for grammatical mistakes and/or that they believed that they wouldn’t matter to her intended readers).

    In fact, I appreciate that the statement dispenses with appearances and seems to come straight from Pandit’s or another author’s desk without having visited a PR unit in between. If such a PR team had helped draft the statement and we all hadn’t discovered today that her English isn’t perfect, we wouldn’t have lost any bit of the information we actually need to scrutinise her Twitter comments and her vice-chancellorship – as much as we haven’t gained anything today by knowing that Pandit is okay with using “would” instead of “will”.

    Ultimately, the only criticism that makes sense here, assuming someone else did draft Pandit’s statement, and if that person’s job is to draft statements, then a) they should do better, and b) Pandit should read public statements before signing them. If she doesn’t, that signals another kind of problem.

    Featured image: Santishree Pandit delivering her 2015 talk. Source: YouTube.

  • The unclosed clause and other things about commas

    The Baffler carried a fantastic critique of The New Yorker‘s use of commas by Kyle Paoletta on August 23. Excerpt:

    The magazine’s paper subscription slips have long carried a tagline: “The best writing, anywhere.” It follows that the source of the best writing, anywhere, must also be the finest available authority on grammar, usage, and punctuation. But regular readers know that The New Yorker’s signature is not standard usage, but its opposite. Nowhere else will you find an accent aigu on “élite” or a diaeresis on “reëmerge.” And the commas—goodness, the commas! These peculiarities are as intrinsic to the magazine’s brand as the foppish Eustace Tilley, and, in the digital age, brand determines content. But the rise of the magazine’s copy desk has done more for The New Yorker than simply generate clicks. It has bolstered the reputation of the magazine as a peerless institution, a class above the Vanity Fairs and Economists of the world, even if the reporting and prose in those publications is on par with (if not often better than) what fills the pages of The New Yorker.

    Paoletta’s piece was all the more enjoyable because it touched on all the little notes about commas that most people usually miss. In one example, he discusses the purpose of commas, split as they are between subordination and rhythm. The former is called so because it “subordinates” content to the grammatical superstructure applied to it. Case in point: a pair of commas is used to demarcate a dependent clause – whether or not it affects the rhythm of the sentence. On the other hand, the rhythmic purpose denotes the use of commas and periods for “varying amounts of breath”. Of course, Paoletta doesn’t take kindly to the subordination position.

    Not only does this attitude treat the reader as somewhat dim, it allows the copy editor to establish a position of privilege over the writer. Later in the same excerpt, [Mary] Norris frets over whether or not some of James Salter’s signature descriptive formulations (a “stunning, wide smile,” a “thin, burgundy dress”) rely on misused commas. When she solicits an explanation, he answers, “I sometimes ignore the rules about commas… Punctuation is for clarity and also emphasis, but I also feel that, if the writing warrants it, punctuation can contribute to the music and rhythm of the sentences.” Norris begrudgingly accepts this defense, but apparently only because a writer of no lesser stature than Salter is making it.

    I’m firmly on the subordination side of things: more than indicating pause, commas are scaffolding for grammar, and thus essential to conveying various gradations of meaning. Using a comma to enforce a pause, or invoke an emphasis, is also meaningless because pauses must originate not out of the writer’s sense of anticipation and surprise but out of the clever arrangement of words and sentences, out of the use of commas to suppress some senses and enhance others. It is not the explicit purpose of written communication to also dictate how it should be performed.

    Along the same vein, I’m aware that using the diaeresis in words like ‘reemerge’ is also a form of control expressed over the performance of language, and one capable of assuming political overtones in some contexts. For example, English is India’s official language, the one used for all official documentation and almost all purposes of identification. However, English is also the tongue of colonialists. As a result, its speakers in India are those who (a) have been able to afford education in a good school, (b) have enjoyed a social standing that, in the pre-Independence period, brought them favours from the British, (c) by virtue of pronouncing some words this way or that, have had access to British or American societies, or combinations of some or all of them. So beating upon the reader that this precisely is how a word ought to be pronounced could easily be The New Yorker using a colonial cudgel over the heads of “no speak English” ‘natives’.

    That said, debating the purpose of commas from the PoV of The New Yorker is one thing. Holding the same debate from the PoV of most English-language newspapers and magazines in the Indian mainstream media is quite another. The comma, in this sphere, is given to establishing rhythm for an overwhelming majority of writers and copy-editors, even though what we’re taught in school is only the use of commas for – as Paoletta put it – subordination. A common mistake that arises out of this position is that, more often than you’d like, clauses are not closed. Here’s an example from The Wire:

    Gertrud Scholtz-Klink, described by Hitler as the “perfect Nazi woman” was held in check by male colleagues when she proposed that female members be awarded similar titles to the males.

    There ought to be a comma after woman” and before was but there isn’t. This comma would be the terminal counterpart to the one that flagged off the start of the dependent clause (described by Hitler as…). Without it, what we have are two dependent clauses demarcated by one comma and no independent clauses – which there ought to be considering we’re looking at what happens to be a full and complete sentence.

    The second most common type of comma-related mistake goes something like the following sentence (picked up from the terms of service of Authory.com):

    You are responsible for the content, that you make available via Authory.

    What the fuck is that comma even doing there? Does the author really think we ought to pause between “content” and “that”? While Salter it would seem used the comma to construct a healthy sense of rhythm, Authory – and hundreds of writers around the world – mortgage punctuation to build the syntactic versions of dubstep. This issue also highlights the danger in letting commas denote rhythm alone: rhythm is subjective, and ordering the words in sentences using subjective rules cannot ever make for a consistent reading experience. On the other hand, using commas as a matter of an objective ruleset would help achieve what Paoletta writes is overarching purpose of style:

    [Style], unlike usage, has no widely agreed upon correct answers. It is useful only insofar as it enforces consistency. Style makes unimportant decisions so that writers don’t have to—about whether to spell the element “sulfur” or “sulphur,” or if it’s best to italicize the names of films or put them in quotes. It is not meant to be noticed: it is meant to remove the possibility of an inconsistency distracting the reader from experiencing the text as the writer intends.

    Here again, of course, I’m not about to let many Indian copy-editors and writers off the hook. Paoletta cites Norris’s defence of the following paragraph as an example of style enforcement gone overboard:

    Strait prefers to give his audience as few distractions as possible: he likes to play on a stage in the center of the arena floor, with four microphones arranged like compass points; every two songs, he moves, counterclockwise, to the next microphone, so that people in each quadrant of the crowd can feel as if he were singing just to them.

    Compare this aberration to nothing short of the outright misshapenness that was an oped penned by Gopalkrishna Gandhi for The Hindu in May 2014. Excerpt:

    In invoking unity and stability, you have regularly turned to the name and stature of Sardar Vallabhbhai Patel. The Sardar, as you would know, chaired the Constituent Assembly’s Committee on Minorities. If the Constitution of India gives crucial guarantees — educational, cultural and religious — to India’s minorities, Sardar Patel has to be thanked, as do other members of that committee, in particular Rajkumari Amrit Kaur, the Christian daughter of Sikh Kapurthala. Adopt, in toto, Mr. Modi, not adapt or modify, dilute or tinker with, the vision of the Constitution on the minorities. You may like to read what the indomitable Sardar said in that committee. Why is there, in so many, so much fear, that they dare not voice their fears?

    A criticism of the oped along these lines that appeared on the pages of this blog elicited a cocky, but well-meaning, repartee from Gandhi:

    Absolutely delighted and want to tell him that I find his comment as refreshing as a shower in lavender for it cures me almost if not fully of my old old habit of taking myself too seriously and writing as if I am meant to change the world and also that I will be very watchful about not enforcing any pauses through commas and under no circumstances on pain of ostracism for that worst of all effects namely dramatic effect and will assiduously follow the near zero comma if not a zero comma rule and that I would greatly value a meet up and a chat discussing pernicious punctuation and other evils.

    It is for very similar reasons that I can’t wait for my copy of Solar Bones to be delivered.

    Featured image: An extratropical cyclone over the US midwest, shaped like a comma. Credit: gsfc/Flickr, CC BY 2.0.