This week: the economics of Duke, the microplastics bandwagon, Isaac Chotiner vs. Shirley Jackson, Thomas Malthus vs. Thomas Malthus, t-tests = AI, can AI people generate new ideas, Hawaiian architecture, becoming Denzel Clarke, and much more. Including a preview of upcoming posts, and notice of an upcoming public lecture by Jeremy!

Yesterday I summarized a bunch of citation data from Web of Science, suggesting that the total rate of scientific publication peaked in 2021 before declining for a couple of years, and that it still hasn’t recovered to its 2021 peak. I also noted that these data broadly lined up with data on submission and publication rates at various EEB journals published by Meirmans et al. (2025).

In response, several commenters expressed skepticism. You should click that last link and read the whole comment thread (it’s good), but the tl;dr is that commenters suggested various alternative explanations for the data I summarized:

• Maybe something funny happened in 2021 with how Web of Science counts citations. We shouldn’t be too quick to dismiss this possibility. In general, whenever you see a sharp trend break in any time series, one of the first questions you ask yourself should be “I wonder if there was a change in how the data are defined/collected/measured?”
• Maybe number of citations per paper has declined since 2021, so that you can’t do what I did in yesterday’s post and use citation count as an index of publication count. Maybe the total publication rate of ecology papers kept growing on its pre-2021 trajectory (or least not declined at any point), but there was a downward step change in 2021 in the length of the average ecology paper’s lit cited section.
• Maybe there’s been a change since 2021 in how citations are distributed across fields, so that citations to ecology papers have declined since 2021 even though the publication rate of ecology papers (and scientific papers as a whole) hasn’t deviated from its pre-2021 growth trajectory.

Reader Kevin Lafferty was curious about these possibilities, so he pulled a bunch of data from ScholarGPS and emailed them to me (thanks Kevin!). Looking at ScholarGPS data addresses the concern that something funny happened to WoS data in 2021. Looking at both publication and citation data addresses concerns about post-2021 changes in citations per paper and the distribution of citations among scholarly fields. ScholarGPS provides total publication and citation counts for various scholarly fields, and for various scholarly disciplines nested within scholarly fields. “Life Sciences” is one of the 14 fields, and EEB is one of the disciplines nested within the life sciences. Here are the ScholarGPS data for annual total publications, and annual total citations, for all papers in EEB:

Here are the the GPS data for life sciences as a whole (which will include some Covid-19 papers):

Kevin reports that he looked at all 14 fields and they all show the same pattern of declining publications since 2021, although the rate of decline varied among fields. For instance, here’s the field of Public Health (which will include many Covid-19 papers):

And here’s the field of Arts and Humanities, where publication rates apparently have completely collapsed but citation rates have kept growing:

Overall, these data support yesterday’s post. Indeed, if anything, these data suggest that the story’s even starker than suggested by the data in yesterday’s post. Publication rates across all scholarly fields have declined since 2021 and show no signs of recovering (though the rate of decline is slowing). Insofar as citation rates have recovered from a temporary post-2021 dip (or in some fields maybe never declined at all), that seems to be because publications these days have more citations, not fewer.

Still definitely interested in hearing your thoughts on these data. But for the moment, I’m inclined to double down on what I wrote yesterday: science–all of it–peaked in 2021 and still hasn’t recovered.

An ecologist friend* noticed that the annual number of times he’s been cited grew throughout his career up until 2021, when it peaked. It then declined for a couple of years straight. Then it started growing again, but has yet to return to its 2021 peak.

Suspecting that what was true for him was true for approximately every ecologist, my friend downloaded the annual citation counts for several long-established ecology journals. That is, the number of times each year that a given journal’s papers–all of them–were cited. The graph for every ecology journal he checked looked more or less the same as his personal citation count: rising to a peak in 2021, then a decline for at least a couple of years, followed by at best a slight recovery and at worst continued decline:

The consistency of the pattern and its timing across so many different journals suggests a hypothesis to explain it:

I am on sabbatical this semester (hooray!). A top priority for my sabbatical time is to make progress on a variety of manuscripts. In some cases, a manuscript draft exists (with substantial variation in how old that draft is!*), and in others we’ll be starting from scratch. As I think about working on these manuscripts, one thing I keep coming back to is whether generative AI (genAI) would be useful for identifying papers or areas of the literature that I missed. 

I want to stress that I am NOT talking about asking GenAI to write an introduction for me. Instead, I started thinking about this more because, at the same time that I was struggling to feel confident that I’d found all the papers that were relevant to a study (based on variable terminology for the phenomenon), I heard from a colleague who had fed a model like ChatGPT a whole grant proposal and asked it to critique what he’d written (and who was happy with the result). 

I don’t see myself ever using it as the primary way of identifying literature (for reasons that include but aren’t limited to its tendency to hallucinate), but I do wonder if I could be a useful supplement. My thinking is that genAI might (emphasis on might) be able to, for example, remind me that a paper that I have filed in my brain as being about topic X also has data on topic Y, or help me find papers that don’t use a particular keyword (or set of keywords) but that is relevant. The standard approaches of reading a paper and seeing what it cites and what cites it can catch some of these things, but certainly not all cases.

I have fairly limited experience so far with generative AI – my main use so far has been to ask the UMich version of ChatGPT for help when I get stuck coding. I didn’t know a ton about how genAI and large language models work, and found the overview in a draft of Stephen Heard’s forthcoming third edition of A Scientist’s Guide to Writing very helpful.** In addition to explaining what the models do, he clearly and thoughtfully goes through different reasons not to use them. While you wait for that revised version to appear, go read some of his blog posts on genAI, including this one and this one. Jeremy pointed me to this opinionated guide to using AI, which I also found useful. 

All of which brings me to the goal of this post. I’m really curious whether other folks are using genAI for this specific purpose – that is, as a supplement to other ways of finding relevant literature. For those who are doing so, I’m especially curious about:

• how they are using it,
• what specific tools they are using, and 
• what they view as the pros and cons. 

I’m also curious about whether they disclose this use for journals that ask about AI use. There are definitely many people who have a kneejerk ‘genAI is lazy/bad’ reaction, and it could be hard to convey the nuance of how something was used. 

So, let’s try a quick poll here, and I also would love if people added more in the comments about how they are using genAI and specific resources they are using (e.g., I recently learned about Research Rabbit from another colleague). (Note: when I preview the post, the actual question at the top of the poll is pretty subtle relative to the answer options, so I’ll repeat it here. The poll asks: Which of the following is most accurate regarding your use of genAI tools to identify relevant literature?)

* I’ve come to realize that the version of R mentioned in the methods ends up being a sadly accurate indicator of just how long a particular manuscript has been languishing.
** Relatedly, Bethann Garramon Merkle and Steve have a new book out on Teaching and Mentoring Writers in the Sciences. My copy is on order from my local bookstore, and Steve has shared that it has a chapter that also goes into genAI from a mentoring writing perspective.

This week: the latest thinking about AI, evaluating open science, how to write a better literature review, against Wright’s Law, LinkedIn vs. blogging, tell me again what wildfires are, tell me again what poles are, the banality of evil enshittification, learning to become a real-life Spiderman, and more. Probably should’ve saved some links for next week, honestly. Oh, and there’s a preview of coming attractions!

Years ago, I compiled data on the teaching experience of ecologists hired into TT faculty positions in N. America during the 2017-18 job season. Over the holidays, I decided to redo that exercise for the 2024-25 job season.

If you’re new around here, see here and here for background on how and why I compile data on the N. American ecology faculty job market.

Back in 2017-18, I was surprised at just how common it was for new hires at R1 unis to have much more teaching experience than I did back when I was hired. And I was surprised to learn that literally everyone hired at a bachelor’s college had experience as an instructor of record, co-instructor, or equivalent experience. I was curious to see if those patterns still hold today, and if they hold in a larger dataset. Back in 2017-18, I only compiled data on the teaching experience of new hires at R1 unis and bachelor’s colleges, figuring that this would bracket the range of possibilities. This time around, I compiled information on the teaching experience of every ecologist hired into a TT faculty position in N. America in 2024-25.

I hope you had a happy holiday. I certainly did, in no small part because my book is finally out! (woohoo!) (update: link fixed) The Ecology of Ecologists: Harnessing Diverse Approaches for a Stronger Science is available from your favorite bookseller. So now, rather than waiting for my book, I’m waiting for people to read my book, and for some of them to hopefully tell me and others what they think of it.

To help that process along, have I got a deal for you! I just made a bookplate that would look great in your copy of The Ecology of Ecologists (or on your laptop):

I will sign a bookplate and send it to you for free if you’re one of the first 50 people to do one of the following:

• Write a review of The Ecology of Ecologists on Amazon, Goodreads, StoryGraph, or LibraryThing.
• Post something (anything!) about The Ecology of Ecologists on social media (Bluesky, Twitter, Instagram, Facebook, Mastodon, LinkedIn, Reddit, etc.).
• Blog about The Ecology of Ecologists
• Start or join a reading group on The Ecology of Ecologists
• Teach or join a graduate seminar on The Ecology of Ecologists
• Comment on this blog to tell the world what you thought of the book
• Email me to tell me what you thought of it (jefox@ucalgary.ca)
• Tell me in person what you thought of it.*

To claim your signed bookplate, just email me. Include a link or screenshot or something so I can see what you did–not because I don’t trust you, but just because I like seeing what people are saying about my book. 🙂 And don’t forget to tell me where to send your bookplate.

*Yes, I know this last option is impractical for most of you. Although if you travel to Calgary specifically to tell me what you thought of my book, you’ll be an instant legend, at least in my mind. 🙂

Happy New Year! This week: no paper is that good, pots vs. kettles and meta-science vs. science, the best nature photography of 2025, fighting the last war pandemic, Wyoming vs. escalators, Canadian measurement units, and more.

My son and I have a holiday tradition: every year I read Terry Pratchett’s Hogfather out loud to him. In case you didn’t know, Hogfather is a fantasy novel that parodies Christmas (while also articulating its “true meaning” better than anything I’ve ever read). Anyway, the passage below will bring a smile to the face of academics everywhere, I think. The no-nonsense Archancellor of Unseen University, Mustrum Ridcully, has just asked a faculty member (the nameless Senior Wrangler) why the university hangs up mistletoe every holiday:

“Well, er…it’s…well, it’s…it’s symbolic, Archchancellor.”

“Ah?”

The Senior Wrangler felt that something more was expected. He groped around in the dusty attics of his education.

“Of…the leaves, d’y’see…they’re symbolic of…of green, d’y’see, whereas the berries, in fact, yes, the berries symbolize…symbolize white. Yes. White and green. Very…symbolic.”

He waited. He was not, unfortunately, disappointed.

“What of?”

The Senior Wrangler coughed. “I’m not sure there has to be an of,” he said.

“Ah? So,” said the Archchancellor, thoughtfully, “it could be said that the white and green symbolize a small parasitic plant?”

“Yes, indeed,” said the Senior Wrangler.

“So mistletoe, in fact, symbolizes mistletoe?”

“Exactly, Archchancellor,” said the Senior Wrangler, who was now just hanging on.

“Funny thing, that,” said Ridcully, in the same thoughtful tone of voice. “That statement is either so deep it would take a lifetime to fully comprehend every particle of its meaning, or it is a load of absolute tosh. Which is it, I wonder?”

“It could be both,” said the Senior Wrangler desperately.

“And that comment,” said Ridcully, “is either very perceptive or very trite.”

“It might be bo–“

“Don’t push it, Senior Wrangler.”

Have a lovely holiday, but don’t push it. 🙂