I recently came across a comical recommendation in teacher-land that faculty (and students) should be regularly using tools like Grammarly and other checkers in their writing, tools which may or may not incorporate different large language model components. As you might expect, my knee-jerk revulsion was quickly followed by an imperative (with special urgency, because of the rate at which such tools are surging in casual adoption by naïve users) to taxonomically classify this unholy policy "fly" in the language ointment. Is the commodification of prose so essential to institutional aims that a school is willing to forego the disciplines of writing, along with the reflection, humility, and consideration in editing the written word?
The first objection that lodged in my throat like a stuck chicken bone was the conceit that usage of such tools should be, not just an indicator of, but a core standard for excellence students and instructors alike. Beyond the fact that it imbues a commercial product with a quality attributable to human production, it also adopts the lie that every writer needs to have editorial nannying to meet success. What twists the knife even more, as a teacher, is the embedded fallacy that this will facilitate better learning and communication. As someone who often composes text in the most minimal environment possible, my clarity of thought would be compromised as I am not only distracted by interface elements from the main task of transposing ideas into letters, words, and sentences, but even more encumbered by my anticipation of how my prose may be nipped, tucked, augmented, or elided for failing to exceed some black box threshold of approval by a proprietary tool.
One argument that I've heard is that LLMs enable the production of better text with greater speed. The profligacy of words they enable is undeniable, and their quality of basic constructions is sound...literally, it is statistically ensured. This line of thinking fails to articulate that the world does not always need volume; it's already a noisy place, after all. The world wants volume because it emulates a proof of work that may be interpreted as good stewardship of ideas, or is suggestive of creativity. Originality is where large language statistical models fall down, both in interest and in factuality. They can be programmed to resist the most pedestrian of verbal tropes, but it is ultimately the lack of facility at combining novel expression with insightful revelation that limits the trustworthiness of machine authorship. AI companies at large are trying to persuade investors that they can overcome these limitations if only enough training data is ingested, and if only billions more in venture capital is invested.
And then there are the intellectual property implications. Not only is there the real possibility that the large language model in use has been trained on data not licensed appropriately, or without the clear consent of users, but there is the ownership elephant in this augmented "reality" room. Federal Copyright in the US vests automatically to "To promote the Progress of Science and useful Arts"[1] when written works (and other media) are placed into fixed form...by a human author. Works that are the output of machines are ineligible for the protections of federal copyright. But what is the dividing line? How much language model voids the protections of this construct? Whatever your opinion on the state of copyright in the US, is that not a slippery slope?
I'll address what I consider the two most grievous of conceits behind the Grammarly Imperative: loss of thoughtfulness, and standardization of style. Writing is thinking...as I've had many wise teachers expound. Ergo, a tool that promises to speed writing and increase its volume for output may be doubly counterproductive if an aim of the writing is to engender deeper, clearer thinking in the author themselves. Grammar may be more or less rule-bound, but any experienced English speaker knows that exceptions abound. Truly, it is only the grimmest of grammar gruffs who would fail to acknowledge that mastery of language thrives when rules are carefully bent, and selectively broken.
- https://constitution.congress.gov/constitution/ Article 1, Section 8↩︎