Science Fictions links for November 2024

Alzheimer's: still not cured

Hello. Below you’ll find your monthly injection of cynicism about the world of scientific research. Remember: it’s only by knowing about this stuff that we can do anything about it (but also it’s fun, in a true-crime kinda way). Here we go:

The links

• This would be a great article on bullshit business school “science” even without the twist ending.

• 7-17% of sentences in computer-science peer reviews are estimated to be written by AIs. Going by some of the slapdash reviews I’ve had in the past, I can’t help thinking this is in some ways a good thing…

  • But seriously: at this point letting an AI do the whole review is a dereliction of duty. But given how good they’re getting at spotting errors, doing your review without the “extra eyes” from an AI might soon be a dereliction of duty too.

• The least surprising RCT result ever: after years of fraud accusations and fines for misleading investors, Cassava Sciences finally released the results for its Alzheimer’s drug simufilam and… it doesn’t work at all. Obviously.

  • As ever, you have to wonder whether the field of Alzheimer’s research has a disproportionate level of bad science, or whether it’s just getting disproportionate attention and all fields are like that.

• I enjoyed this metaphor of the scientific literature being like an unsafe lab where you have no ability to label the faulty equipment to protect your colleagues.

• A very annoying aspect of papers getting retracted (for whatver reason) is that the retraction note doesn’t fully explain what went wrong with the article, and uses vague language about “errors” or similar. Here’s a rare instance of an inadequate retraction note getting rewritten.

• It’s still the case that effectively nobody publishes replication studies. “Based on these findings, we argue it would be premature to declare that psychology’s replication crisis is over”.

  • And on that same topic, a useful preprint on what scientists should do when they fail to replicate a result—should they try to replicate it again?

• Room-temperature superconductor fraudster finally fired. No, not that room-temperature superconductor fraudster. The other one. No no, not that one either! The one at Rochester!

• It’s been well over 100 years since the Piltdown Man hoax, so we were in need of another big scandal in palaeontology (or palaeoanthropology I guess). Here it is!

• Jesse is right here: Scientific American’s (now-departed) editor turned it into something not very scientific at all. Glad she’s gone.

  • The response when you point this out is often something like: “Scientific American isn’t a science journal! It’s just a pop magazine! Who cares if it publishes political stuff?”. But a lot of people look up to it, the same way they do to New Scientist here in the UK, which also isn’t a journal. It’s bad if stuff that’s associated with science becomes politicised, full stop.

• Sokal-style testing-the-boundaries hoax paper? The work of someone with psychosis that nevertheless got published in a “serious” journal? Whatever it is, you’ll surely join me in saying “WTF” at “Practice of neurosurgery on Saturn”.

• Coercive citation is a real thing: “As strongly requested by the reviewers, here we cite some references [[35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47]] although they are completely irrelevant to the present work.”

P.S. The Studies Show

You can never predict what’ll be popular: this month people went crazy for our podcast episode on collider bias. So I guess we’ll do more stuff on weird statistical phenomena in future, and to make sure you get them in your inbox you should subscribe to The Studies Show!

P.P.S. “The other place”

Yes I know it’s cringe and all that. But it does seem like a lot of scientists have moved over from Twitter/X to Bluesky now, so to be able to keep tabs on stuff for this newsletter, I opened an account which you can follow if you want: stuartjritchie.bsky.social.

Image credit: Getty