ChatGPT: author, acknowledgement, method, or tool?

Like everybody else, I’m fascinating by the fast-moving world of ChatGPT and its kin. Today: what do these tools mean for authorship, in a world where scientists are already using them to help them write their papers?

Tools like ChatGPT are often referred to as “artificial intelligence”, but that’s a terrible label for them – it misleads people, who are then surprised when ChatGPT output is dumb, or wrong, or just plain fabricated.*

Advertisement

Is there an easier way to teach scientific writing?

Each winter semester for the last 5 years or so I’ve taught a course for upper-year and grad students in scientific writing. My course has three components to it: (1) a series of (mini) lectures; (2) an accompanying series of small-group workshops; and (3) a series of assignments, via which each student submits, piece by piece, a first draft and then (following comments from me) a revised draft of a scientific paper.

I’ve just wrapped up this year’s version, which (rather sadly) will be my last, as I’m retiring at the end of this year. Every time I teach the course, I come to the same realization: teaching scientific writing the way I’ve been doing it is hard.

Why you should write “In this study, we…” – and then delete it

Each year, when I teach my Scientific Writing Course, I find my students committing many of the same writing sins (which is only fair, as I commit many of them too). But from year to year there always seems to be a different standout. This year it’s “in this study”. Once, I would have counted “in this study” as one of my writing pet peeves; but I think of it a bit differently now. It’s still a pet peeve – but an editing one, not a writing one. I think that’s a useful distinction, so let me explain.

The case against “in this study” is simple.

SOURs: Strong Opinions Unmoored from Rationale

Scientists really value evidence. Or at least, that’s what we all tell each other: we test hypotheses by confronting them with data, and our view of how the world works reflects the results of all these hypothesis tests. We get very, very upset when charlatans push irrational nonsense like intelligent design, ivermectin treatment for COVID, or the supposed dangers of vaccines. After all, in each case we have a well-founded rationale for declaring “nonsense”: there’s a mountain of evidence that all life on Earth has evolved through natural selection, that ivermectin is useless or worse against COVID, and that vaccines are safe and effective.* If we have a belief about the world, we ask if there’s data to support that belief; and if there isn’t, we change it. Right?

Well, surprisingly often, no.

“A Gentleman in Moscow” and scientific writing

In a career, how many extraordinary papers might a scientist write?

I got thinking about this, believe it or not, after noticing a copy of Amor Towles’ A Gentleman in Moscow in a neighbourhood Little Free Library. That novel is astonishing (in fact, it’s not at all clear why you’re reading this post instead of A Gentleman in Moscow; but thanks.) Since reading it, I’ve read Towles’ other novels (Rules of Civility and The Lincoln Highway); and while both were fine books well worth my time, neither grabbed me the way A Gentleman in Moscow did.*

Towles isn’t the only author to show this pattern.

Brevity in scientific writing is a good thing – until it isn’t

Nearly all of us need to work at making our scientific writing more concise. I’m definitely part of “nearly all of us” – my usual first draft needs to be trimmed down by 30%, producing a second draft that needs to lose another 30%. In my experience, making a manuscript shorter nearly always makes it clearer and better, in part by forcing it to become less “science-y” (forcing it into active voice, requiring the jettisoning of long fancy words in favour of short simple ones, and so on). So my attention was caught the other day by a reference to the recent publication of the “second shortest” philosophy paper. That paper consists only of its title: “Can a good philosophical contribution be made just by asking a question?”*

Unfortunately, the answer to the paper’s title is pretty clearly “no”.**

My love affair with the footnote

If you’ve been reading Scientist Sees Squirrel for a while, or if you’ve read The Scientist’s Guide to Writing, you’ll have noticed my tendency to footnote. Footnotes drive some folks up the wall; others love them. You can tell that I’m in the latter camp – but why?

I guess there are two ways to answer that question: historically and functionally.

Historically: I’ve always loved weird digressions and unexpected connections – things that might not quite be germane to the main point, but are interesting or surprising or funny.

My most heterodox scientific-writing lecture (or, how I annoy my colleagues in a good cause)

I’m well into my Scientific Writing course now, and I’ve just given the lecture that consistently annoys my faculty colleagues the most (well, it annoys many of them). It’s the one on writing the Methods section, and it’s heterodox in two rather different ways. This lecture stands out a bit – I don’t think my approach to IMRaD structure, or the content of the Discussion, or outlining, or writer’s block is all that different from the approach anyone else might take. But the Methods is different.

I said what I teach about the Methods is heterodox in two different ways.

ChatGPT did not write this post

Like everyone else, I’ve been watching the rise of “generative AI” with both interest and trepidation. (“Generative AI” is software that creates “new” text (ChatGPT) or images (DALL-E) from a user prompt – I’ll explain the quotes on “new” later.) Now, I know only a smattering about how generative AI works, so don’t expect technical insights here. But I’ve noticed an interesting gap between what I think these systems are doing and how people are reacting to them.

My interest in generative AI, especially text generators, is easily explained and probably obvious. Since I was in high school I’ve watched software get very slowly better at imitating the kind of writing humans do with great effort, and the kind of conversational interaction that humans do without a second thought.* The latest round is, superficially, really impressive: it can chatter pleasantly about nothing much, write a poem, program in R,** write an essay about Canadian history, explain linkage disequilibrium, and more. Or at least, it often looks like it can.

It’s not lazy to do the easy writing first

I’ve used jigsaw puzzles as a metaphor for writing before, and today I’m returning to that surprisingly fertile ground. If you’re not into jigsaws, don’t worry: as an alternative framing, I can offer Things My Mother Told Me That Were Not True. (Also ground I’ve ploughed before.)

When I was growing up, my family put a lot of stock in what’s sometimes described as a Protestant work ethic. My mother in particular was pretty strong on this one point: if you have an easy job and a hard job ahead of you, don’t be lazy: do the hard job first. To return to the jigsaw puzzle metaphor: start with the sky, not with the deck chairs. Well, there are surely tasks for which that advice makes sense, but I’m here to tell you that neither jigsaw puzzles nor writing are (usually) among them.

When I write a new paper (or anything else), I ease myself in.