A Scrumptiously Quick Retraction
A dog's breakfast struck from the scientific record.
I’ve been busy with my new magazine project, a project that has totally blown me away. The support has been crazy.
If you haven’t read about that yet, then you should go here:
That hasn’t left a lot of time for scrounging around in the research literature. But here’s a scrumptious update about a scrumptious error.
[God it feels good to type scrumptious. Just a real good hand feel. We really underrate hand feel on the keyboard I think. What feels better? Typing I hate or typing I love? I think the latter. Sorry distracting.]
Anatomy of a Retraction
Overnight, Springer Nature’s Scientific Reports retracted the paper titled “LungGANDetectAI: a GAN-augmented and attention-guided deep learning framework for accurate and explainable lung cancer detection”.
The retraction came on St Patrick’s Day. Lucky.
The retraction note states:
Following publication, concerns were raised regarding irregularities in the Article’s content. These include nonstandard phrases, non‑scientific terminology, errors in the data presented in Figure 15, and inaccuracies within the cited references. The Authors did not provide any explanation for the raised concerns, nor the underlying data for the figure in question.
S. Sudeshna and B. Umamaheswara Rao disagree with this retraction.
This paper came to my attention after being flagged by the “Problematic Paper Screener” (PPS) for a couple of tortured phrases, but what the screener didn’t pick up (only because it’s not tuned for this, that’s no knock on PPS) is the word scrumptiously.
Buried in the text:
“The results show that both the test and training samples achieve a scrumptiously high accuracy of 95%.”
I flagged this phrase, and a host of other errors in the paper that I documented here (a true dog’s brekky), with the Springer Nature integrity team back in February. Within a day, it had placed an Editor’s Note on the paper.
It’s strange to see both authors disagree with the retraction, though they did not provide explanations or underlying data. That feels like a pretty solid tell to me that perhaps the underlying data is nonsense. And I think it’s a pretty solid stance to retract this, within a month of it being flagged, when that underlying data isn’t produced.
Of course, the fact this made it past peer review is concerning, but maybe we’re seeing Scientific Reports enter its Folklore era1. As Rafal Marszalek, the chief editor told no breakthroughs last month:
“While we already have a range of checks in place, Scientific Reports introduced an additional workflow in January 2026 to help identify potentially problematic submissions more effectively.”
Looking at the Problematic Paper Screener, there seems to have been just a single addition from Scientific Reports since I looked in February. So, are things changing? Has this additional workflow enabled better detection of problematic submissions?
[Shakes Magic 8 ball]
Ask again later.
What am I taking away from this?
PPS is a tool science journalists should familiarize themselves with, particularly if they are interested in reporting on integrity breaches and the publishing process. It’s been invaluable in scoping shonky papers, but it also provides a kind of education on what you should be looking for when reading papers. Use it.
We need more science journalists who want to do that!
I think we’d all prefer these errors be caught during the peer review process given they’re so obviously unscientific. My gut feel is that Scientific Reports and other journals of its ilk should be a little more transparent about what processes are in place. This may enable submitters to get around those processes more easily — it is an arms race — but the way papers are treated could also work in the opposite direction, where legitimate science gets flagged as problematic.
Scrumptious is a delightful word.
See you next week!
Some scribbles
— The Kickstarter campaign for CONTINUE? was fully funded in seven hours. This is unreal and I cannot believe it. While we have some challenges to overcome in terms of shipping, war, all that crazy stuff, I am so buoyed by the response. People want real things. They want to know that what their consuming isn’t AI garbage. They want to… (sorry) Continue.
— My good friend Sam Nichols told me he would subedit my newsletter for free because I sometimes completely forget to finish my
— Hi Sam!
— I’m also working on a big investigation behind the scenes which may make this newsletter a little less frequent. There’s only so much I can do while still paying rent, which seems to be getting harder and harder!!!!!!!!!!!!!
A Taylor Swift reference. Folklore is her best album. Thank you and good night.