AI-assisted writing is common because writing is hard

A UC Berkeley math professor, Zvezdelina Stankova, admitted to using AI to edit her August 15th essay that criticized students' preparedness, after the AI detector Pangram flagged the piece as 33% AI-generated. Stankova said the article was the result of 'several hundred person-hours of intensive human work and deliberation, of which about 80 hours are my own,' and that AI was used for editing. The incident highlights the widespread use of AI-assisted writing among professionals and students, raising questions about authorship and trust.

A week ago it came to light that UC Berkeley math professor Zvezdelina Stankova, who went viral in an August 15th essay deploring https://sfstandard.com/opinion/2026/08/15/uc-berkeley-sat-test-blind-admissions-math-scores/ the poor preparedness of some of her students, actually used https://www.dailycal.org/news/campus/uc-berkeley-professor-admits-to-using-ai-in-op-ed-calling-out-students-lack-of/article be1360f7-6e2a-42f6-ae76-f05d8cee74c4.html AI for her essay. The article was flagged by the AI detector Pangram as 33% AI-generated or AI-assisted writing. A similar result was found in a June open letter from STEM faculty advocating for standardized testing. “The faculty team and a number of journalists worked intensively on my draft over a period of three weeks with multiple drafts, meetings,” Stankova said in an email. She added that she used AI to help edit the piece but that the article is the result of “several hundred person-hours of intensive human work and deliberation, of which about 80 hours are my own.” Her argument still stands. I think it does detract a bit from her credibility. Here, she’s taking a moral high ground, when that ground is soft. If you want to use AI, fine, but accusing others of being lazy or unprepared, whilst using AI and having an editorial team, is a bit rich. But it is true that columnists, even for one-off op-ed pieces, have a large staff of uncredited editors–including AI assistance–despite their sole name being on the byline. Conversely, students are expected to to have minimal outside help, and when they avail themselves of such tools or assistance, it’s called being lazy, cheating, or being unprepared. This seems wrong. But that isn’t the point of this post. The question is, “why do so many people, professionals and students alike, depend so much on AI for writing?” It’s well known that AI writing is easy to spot even without Pangram and worse, yet people still use it. It’s interesting, in the comments, to see AI blamed on poor preparedness: Yes, because before AI, everyone was proficient, and then AI came along and ruined everything. Remember “Covid learning loss” from 2021-2023 or so, which was a big deal at the time? Now it’s AI. Before then, video games were blamed. Or TV. It’s always something. There will always be a certain percentage of students who struggle no matter what. It’s possible that this percentage is higher, yet what is often blamed on technology or other factors, merely reflects innate individual differences of ability or other factors such as some students simply having a disinterest in school . Math and computer science have tons of collaboration. It’s not uncommon for a math paper to have more than a single author, whereas writing is more of a solitary endeavor. Or STEM professors will delegate work to graduate students, such as proofreading. When you delegate writing, you’re no longer a writer; it’s something else. Even if the ideas are yours, it’s implicitly assumed when you read something, that the author actually wrote it. AI is not merely a tool in the same way that a hammer is used to build a house. With AI, the authorship itself is uncredited, making it closer to ghostwriting, or worse, plagiarism, whereas the analogue of a hammer for writing would be, say, a pencil, but it’s not like the pencil is doing the writing. This is why AI writing is so controversial, because it breaks the trust between the reader and the author. Indeed, people feel like they are being ripped off when they see AI writing and it’s undisclosed. The most obvious explanation for the proliferation of AI-assisted writing is to save time. It’s simply easier to outsource the writing to AI. Even if AI-assisted writing is obvious and worse, this is a tradeoff that is preferable to no writing . So if given a choice between AI help and less or no writing, the former is a compromise that strikes a balance between the need to communicate and the reader lowering his or her expectations. A second, less explored explanation is that writing itself is hard. Linguists in the 20th century, such as Pinker and Chomsky, argued that language is innate. But this is only speech. Writing and reading involve decoding and encoding printed language, and there is nothing inherently natural about this. This is how you end up with the situation where everyone has verbal fluency, but literary and writing ability are much more unequally distributed. People can spend twenty years reading and writing in school or have doctorates, and still struggle to organize a 1,500-word argument. Of the thousands of people on twitter or readers in general, only a handful do any long-form writing. Of those who do, output tends to be spotty or infrequent e.g. the occasional Malcolm Gladwell op-ed . Examples such as David Brooks, Richard Hanania, Noah Smith, and Freddie deBoer, who write long-form on a weekly, consistent basis, are truly anomalous. Literally maybe only dozens of people. I, for example, wrote about 75% of this essay in a single take sitting down, with the rest the following day. Essay-style writing, as opposed to tweeting, is about taking a chain of inchoate thoughts and turning them into something that has a self-contained structure, which is also hopefully pleasing to the recipient. This is an added level of difficulty and orthogonal to merely the idea-generation process, and is in some ways harder, because there are so many wrong ways to do it, resulting in bad writing even if the ideas are sound. For example, her essay could be broken or distilled into the below ideas: -too many incoming freshmen are unprepared for college-level math at UC Berkeley -a certain percentage lack proficiency of even the basics, such as fractions -this is attributable to making the SAT optional, which has backfired -the SAT should be mandatory for admission These would be perfectly fine as a series of tweets, but not as a standalone essay. That requires: structure, grammar, organization, transitions, style, and some form of narration or storytelling rather than a recitation of facts. And AI still cannot actually do this, without it sounding like AI. Even for this essay, I still had to to run it through AI to find mistakes, which it inevitably did.