Also this week: changing gender balance of US undergrads in math-intensive majors, brown bears vs. insurance fraud, LLMs vs. electricity demand, LLMs vs. signalling theory, against writing advice, the reverse population bomb, and more.
Writing in Science, Cimpian and King use a giant near-census dataset to show that the gender balance of physics, engineering, and computer science majors has steadily improved at US colleges and universities over the 21st century–but mostly at highly selective institutions, where entering students have the highest SAT math scores. Both enrollment and retention of women in these majors seems to vary with mean SAT math score of entering students. The improvements in gender balance also are strongest among white and Asian students than students of other races. The causes of these patterns aren’t clear, though the authors note that “social belonging” interventions intended to improve enrollment and retention of women in math-intensive majors seem to have the largest positive impact at the most selective institutions. I’m far from an expert here, but these results make sense to me. My possibly-oversimplified thought is that, if your institution has (i) a lot of money, and (ii) many more well-prepared undergraduate applicants than it is prepared to accept, then it surely ought to be able to increase the diversity of its undergraduate physics/engineering/math/compsci majors on any dimension, if it sets its mind to it. Whereas if your institution lacks (i) and/or (ii), it’s going to be swimming upstream if it wants to diversify its physics, math, engineering, and compsci majors.
Some local news from my neck of the woods: the latest provincial government review of higher education funding in Alberta will ignore the elephant in the room.
How much current electricity demand is due to LLMs, and how might it increase over the next five years? Hannah Ritchie crunches the numbers–which don’t support alarmism or moral panic over LLMs. I also was interested in the historical context regarding past alarmism about computer use and electricity demand. Finally, I appreciated that the piece addresses obvious questions/pushback head-on. Just a very good piece all round.
Honest signalling theory and LLMs. Scroll down to the “AI IR” subsection. It’s about investor relations, but the connections to honest signalling theory will be obvious to all our behavioral ecologist and evolutionary ecologist readers.
I don’t really follow online discussion or commentary on human population growth and demography, so I don’t really know what others (particularly environmentalists) think about this.
All writing advice is a lie. This is about fiction writing advice. I’m going to need a ruling from Steven Heard on whether it generalizes to scientific writing. HAVE YOU BEEN LYING STEVEN? 😉
How William Hazlitt predicted the modern “discourse” novel.
The Village Voice in the 1960s/70s as proto-group blog.
Marxist philosopher Liam Bright on whether we should continue to read and honor philosophers who did bad things.
Athene Donald in praise of technicians.
Data mining Easter egg from 1986. It’s a lot easier to spot than it was back in 1986, because now we have R and 4K monitors. 🙂
See, if you’d paid attention in your zoology courses, you’d have known to rent a black bear costume as part of your California insurance fraud scheme, not a brown bear costume. 🙂 Now I want one of our zoologist readers to chime in with a detailed, step-by-step guide for how to commit insurance fraud by dressing as a bear. I want zoology nerd protips on choice of costume, mimicking bear behavior, etc. Come on readers, don’t let me down! 🙂 In semi-seriousness, I find stories like this a useful reminder of the limits of blaming fraud (scientific fraud, insurance fraud, whatever) solely on “incentives.” Such as, say, incentives to “publish or perish.” I mean, sure, by all means think about systemic incentives and how they might be improved so as to discourage fraud. Incentives definitely can matter sometimes, for some people. Just make sure you’re also thinking about the fact that there are people in the world who will try to commit fraud by dressing as bears. If you’re the sort of person who would commit fraud by dressing as a bear, you’re not the sort of person who will respond to incentives, at least not in the way that most people would. Now, what Jonathan Pruitt and Danielle Dixson did wasn’t quite the scientific fraud equivalent of dressing as bears, but it was in the same neighborhood. Stories like this one also provide a useful comeback to any fraudster (scientific or otherwise) who defends themselves by saying something like “Why would I do that? That would make no sense, you’d have to be either dumb or crazy to do that.” To which: yes, you would. 🙂