The Boston Globe published an article about a familiar subject: Cambridge’s zoning policies. While I can’t say for sure how the piece was written, it contains a few AI-isms we have all come to look out for: “Unit size matters, too” (the short tagline to introduce a larger thought; the use of “matters” to link it to the thesis with no explanation); sneakily, the “not X, but Y” construction that large language models love, as in “Cambridge must also count what is lost. Replacing a two-family house with two new, more expensive homes preserves the unit count but reduces affordability.” This was once a concise and clever means of making a point. Now it just kind of makes you look like you don’t know how to write.
The ideas in the opinion piece are probably genuine, and they were reviewed by the Globe’s editorial department (which likely cut down on some of the awkward wording). If I had to guess, the person who wrote it ran it through ChatGPT or the like after putting her basic thoughts on paper.
Normally, the Independent’s investigative team would leave it here. There is a well-known name attached to the piece, though: Suzanne Preston Blier, president of Cambridge Citywide Coalition and professor of fine arts and African and African American studies. (Blier says the article was not written with AI, but that like many researchers, she uses artificial intelligence to search and analyze large datasets.)
I encourage you to (attempt to) read the CCC’s website to see what it has to say. The main feature is the sprawling webpage labeled “Research posts” or “blogs” that, when open on my computer, makes the scrollbar appear so small that it is hard to use. Certainly it’s great to support civic discourse, run a blog and conduct research. It takes work; I am expected to write a weekly article, and it’s not always easy to keep up. The problem is that it feels obvious Blier isn’t writing this herself; AI is.
A piece drawing connections between various Cambridge political groups is the best example. Clocking in at 6,000 words and 57 fully formatted citations, this amount of text and research would have taken multiple days for a human to write and organize. While writing long pieces of research is not in itself proof of anything, the prolific output of novellas such as these on the site makes it seem improbable to me that a human wrote them. When I wrote the Wikipedia article on rivers, it took 6,000 words and 44 sources. I certainly wasn’t doing it multiple times per week.
Without the benefit of the Globe’s word count limit or editors, the use of tropes common in AI multiply on the coalition website. In a single paragraph:
“The argument here is not that […]. Nor does […]. The issue is accumulation: how many […] before reasonable questions arise about institutional independence, conflicts of interest and public disclosure? The question is not […], but […].”
Here we have two uses of the “not X, but Y” construction, and the now-grating “rule of three” argument style. Given the article’s tone and writing, the website’s context and the generated images within, I’m drawn to think the piece has more AI than the analysis Blier acknowledges in saying she uses “the most advanced AI (ChatGPT).”
The social contract of writing used to include a promise that the writer put in more work than the reader. Now that someone can, with a few clicks of a button, generate thousands of annotated words they might not not read or even understand, this social contract is broken. Notice I haven’t dug into the substance of Blier’s arguments? That’s because the entire signaling around this body of work is that it lacks substance and is designed to waste my time.
If we know writers can’t or won’t formulate their own arguments and use their own words, how are we meant to engage seriously with their arguments? Say, for example, that a writer who uses AI responds to this piece. Even if they didn’t use AI, it’s fair to assume they would – and I would have to account for the possibility I was not arguing with a human.
Suzanne Blier is smart. I think many people are generally smart and good; that’s why I live here. I am less interested in framing our current culture war as a quest to overcome idiocracy – I see it as the gap between people who have a basic level of conscientiousness and those who don’t. I have no idea how to convince someone to care about these things. It’s interesting, if not scary, that you can teach at Harvard or Dartmouth or the University of California, Berkeley, have a massive body of academic work and lose your ability to write.
When someone is deep into AI, it is almost impossible to explain to them why it’s not the right tool for every situation and decision. It does not occur that there are some instances where it should be avoided – and despite being warned of this and aware it makes them look out of touch and clumsy, they expect to be taken seriously.
I don’t have to take anyone seriously. If I have any advice to give, maybe just save AI for your work email or your blog. Leave it out of the paper of record.
