Copyediting and Philosophy, Part 3: Language, Power, and Copyediting

The Issues in Philosophy Beat is running a three-part mini-series called “Copyediting and Philosophy,” which focuses on issues around copyediting relevant to the philosophy profession: what it is, how to navigate it as an author, and philosophical questions it raises. This post is the third and final installment.

Copyeditors are enforcers of a particular kind, even if they are not the originators of the norms they’re enforcing. Still, we might worry that the practice of copyediting could enforce standard Englishes to the exclusion of other registers and dialects in a pernicious way. Jessica Flanigan, for instance, argues that linguistic pedantry and an overemphasis on conformity with spelling and grammar perpetuates social hierarchies. And a few years ago, an essay by a former copyeditor, Helen Betya Rubinstein, argued that copyediting is a white supremacist project. 

I think it’s important to pay attention to prejudices that attach to different varieties of English, whether these be registers, dialects, or “world” or “global” Englishes. At the same time, I think copyediting can be a way to facilitate communication, especially within academic philosophical writing, which is already subject to certain norms. Working within these norms, editors can work with authors to anticipate points of difficulty with their audience. However, there is a new kind of copyeditor on the scene, artificial intelligence, and there is potential for a new kind of prescriptivism to be built into editing.

Linguistic prescriptivism

Before reflecting on the relationship copyeditors have with norms, it’s worth getting clear about the concept of linguistic prescriptivism. Linguist Geoff Pullum has argued that we shouldn’t confuse “prescriptive” and “normative.” Pullum’s argument is that grammars are normative—in the sense that there are rules that govern grammar. Certain expressions in a language are ungrammatical or incorrect. Linguists investigate the norms of grammar through empirical research into what expressions speakers of a language judge as grammatical, for instance. However, prescriptive grammars (or style guides, etc.) encourage or discourage certain constructions. Sometimes these prescriptions are inaccurate, like zombie rules about ending sentences with prepositions, and sometimes they are based in prejudice against speakers of certain languages. 

But Pullum argues that it’s possible to give cautious, audience-relative, instrumental, and evidence-based prescriptive advice. He cites Garner’s Modern English Usage as an example of such a prescriptive guide. Garner’s advice about which constructions are acceptable depends on evidence about how speakers respond to them. Given certain audiences and certain aims, then, for example, one ought to avoid using “hopefully” as a modal adjunct. Take a sentence like “Hopefully, you are enjoying this blog post.” Not all English speakers judge sentences like that as grammatical; some people think “hopefully” should only be used as a manner adjunct, as in “We should live our lives hopefully, come what may.”

The upshot: linguistic prescriptivism is not simply the judgment that certain expressions are grammatical relative to a language. Prescriptivism involves judgments about how people ought to speak relative to a language. However, if we allow that instrumentalist prescriptivism is acceptable, even if copyediting is a prescriptivist project, it needn’t be intrinsically pernicious. If an academic writer wishes to publish in a venue in which certain norms already apply and a certain audience is already anticipated, a copyeditor can apply prescriptive judgments relative to those aims. They needn’t be universal judgments, beyond the particular manuscript, or come along with moral judgments about the speaker’s use of English. 

Different Englishes and different grammars

One of the challenges raised for copyediting is that it excludes certain linguistic forms, it erases voices, and it values rigid perfectionism and rule-following over humanity. Rubinstein’s essay on copyediting says:

But the argument that texts ought to read “easily” slips too readily into justification for insisting a text working outside dominant Englishes better reflect the English of a dominant-culture reader—the kind of reader who might mirror the majority of those at the helm of the publishing industry, but not the kind of reader who reflects a potential readership (or writership) at large.

Her target isn’t academic writing but presumably the “Big Five” trade publishers and the world of popular fiction. Still, we might wonder whether her claim applies to academic philosophy, ceteris paribus, for a text’s “clarity.”

This raises to salience the global context of academic writing, in which the linguistic forms at issue are not just tied to race, but also to being native and non-native speakers of a language. In light of this, the Barcelona Principles (BP) urges philosophers to be aware of how native and non-native speakers may face disparate judgments about their work.

In the frequently asked questions section about the principles, the authors of the BP answer the question, “Will endorsing the BP come at the cost of publishing lower-quality work?” Their reply points to a challenge in editing and some issues in linguistics, distinguishing between grammatical mistakes and differences in style:

Unless you think that good philosophy requires writing in a native-sounding style, the BP should not change your already existing editorial policies. The BP requires not giving undue weight to style. This does not imply an obligation to publish papers that readers cannot understand because of unduly poor English use.

Back to Pullum again. In a talk for the Modern Language Association, he says this about style:

Style involves skill. It involves making choices between alternatives that the grammar makes available. Those choices can make the difference between using the language brilliantly or using it ploddingly. This is the sort of domain in which the notion of regulative rules makes sense.

(“Regulative” in his quote is the same as “prescriptive.”) There are several distinctions at play here. There is native-sounding style and non-native-sounding style. There is good and bad English use. And there is writing that readers can and cannot understand. How do we disentangle these opposing pairs? We might think that an editor should just target bad English use that leads to misunderstanding, leaving whichever style was used in place. However, the question is which English use are we considering? If certain constructions are grammatical in a non-native English dialect but ungrammatical in standard American English, should they be corrected? It’s often a point of contention whether constructions are errors or emerging features of a World English.

Now we are back to the challenge of writing for the standard English dialect of the reader. The Barcelona Principles, like so many others challenging negative responses to non-standard linguistic dialects or World Englishes, are rightly trying to change biased responses to language. However, we are not in an ideal world. Depending on the World English being used, the reception may be poor. Singlish, for instance, a form of Singaporean English that incorporates elements of British English, Malay, Chinese, and Tamil, is often viewed as a colloquial and substandard form of English. A perfectly written Singlish academic paper would likely not pass muster at an academic journal.

And so, we could think of copyeditors not as enforcers who exclude voices, but as assistants who enable voices to be heard that would otherwise have been excluded. A Singlish speaker who is not comfortable writing in standard British or American English, despite English being one of their “native languages,” or a non-native Spanish speaker could both make use of copyeditors to help render their voices into “standard” Englishes. 

Further, the editors I have met and whose works I’ve read are deeply concerned with language and power. For instance, in a jab at Bryan Garner’s appeal to “careful usage,” copyeditor Erin Brenner emphasizes that good editors consider their audience, not the pronouncements of pedants. (For linguist Mark Liberman’s take on Garner, the author of Garner’s Modern English Usage, see this Language Log post.) Alex Kapitan at The Radical Editor talks explicitly about being “guided by care, not correctness,” which they take to mean focusing on how editing for and in particular contexts enables real communication. Their blog post talks a lot about the singular third-person pronouns “they” and “them”—and in 2026, three years after their post, I can report that my copyediting course spent a lot of time talking about pronoun use and gender identity, person-first and identity-first language, and so on. Copyeditors today, at least many of us, are reflecting on linguistic norms and power.

Standardizing and AI editing

Still, we might worry that human editors are supporting an unjust system even while helping individuals survive within it. And won’t they have their own biases they inject into the editing process? Perhaps artificial intelligence (AI) might be a better approach to editing. After all, if correctness conditions for languages are a matter of things like statistical prevalence, a large-language model (LLM) trained on a vast corpus might be a better, more objective editor than a human being, with all of their idiosyncrasies and prejudices.

In fact, a recent preprint investigating AI editing suggests that LLMs are not just faster versions of human editing. Instead, rather than replicating human biases, they are editing in ways no human editor ever would. This suggests a worry: if non-native English speakers are concerned with standardizing their English, they may use AI editors for this purpose. However, these LLM copyeditors may not preserve their voices—or even their arguments. The article is worth reading, as it investigates AI editing (and refereeing) in several ways. The two investigations relevant to this post focus on counterfactual editing and scientific peer review.

The investigation considering counterfactual edits fed human-authored essays and expert feedback to a number of LLMs (Claude Haiku 4.5, Gemini 2.5 Flash, and gpt-5-min). The research corpus included the essays revised after authors took expert feedback into account, but these essays were not given to the LLMs. The researchers then compared the LLM-revised essays and the human-revised essays. Here’s what they found:

LLMs replace a much larger fraction of the original writing than humans do when revising their own work. This substitution of words contributes to the loss of individual voice, style, and meaning, as the unique lexical fingerprint of each writer is overwritten by the given model’s preferred vocabulary.

They also found that the semantic shifts in revision due to LLM use are “substantially semantically different from that produced by humans.” And they found LLMs would even change the conclusion of the essay, “even when LLMs are tasked to perform minimal edits to the essay.”

Further, this same article argues that AI journal referees are sensitive to different criteria than human referees. So, a potential feedback loop looms: LLM copyeditors revise in non-human ways, leading to undesired and unexpected changes. These AI-revised manuscripts are “judged” by AI referees who judge in ways human judges would not. The manuscripts that pass this process of AI revision and AI judgment are then used for corpus training, reinforcing both editing and judgment.

This means we may find a new set of prescriptivist fantasies being built for us. Instead of looking back to Latin and hallucinating a linguistic basis for banning prepositions at the end of sentences, we may find ourselves, without reflection at all, incorporating AI speech patterns and taking them as the basis for “good” English. And non-native English speakers are in a further bind, since there is some evidence that AI-detectors are biased against work that they’ve written.

So, it seems, handing over human responsibility for editing to the machines does not magically, algorithmically, solve our problems. In fact, it may introduce new ones. Human judgment may be flawed, and the structures we have built may be unjust, but there is no guarantee that AI will build us a better linguistic world.