You can withdraw a discredited paper, but the long tail of citations lingers on and on, cautions Russ Swan.
I’ve said it before, and I'll say it again: scientific fact is nothing more than this year’s best guess.
Ask your proverbial person in the street about a heart disease and diet, and I’ll bet you a litre of olive oil to a gram of lard they will cheerfully tell you all about the Mediterranean diet. Eat like a Spaniard, with lots of fresh fruit and veg, oily fish, and lashings of both olive oil and red wine, and you'll outlive your pension. Science, innit.
It feels like this common knowledge has been part of the collective understanding for decades, but in fact it’s both recent and wrong. The seminal paper on the topic, ‘Primary Prevention of Cardiovascular Disease with a Mediterranean Diet’ (Estruch et al), was published just ten years ago in the prestigious New England Journal of Medicine. It changed eating habits in the global north and spawned a boom in Mediterranean cookbooks.
It also has the distinction of being the most highly cited retracted paper in history, with 3,000 citations – a third of those after it was shown to spout nonsense.
That, however, is nothing compared to the second and third on the list of influential retracted papers. A 1998 effort by Andrew Wakefield – he of bogus claims about MMR and autism – concerning developmental disorders in children, was cited almost 50% more times after retraction than before.
Even more startling is a 2005 paper claiming, falsely, that a protein called visfatin is an insulin substitute. Two hundred and thirty-two citations before retraction, no fewer than 1,232 after.
Clearly the simple act of retracting a paper does not stop the propagation of the untruths it may contain. It might even make it worse. Citations breed citations, in a way any decent virologist will understand.
A 1998 effort by Andrew Wakefield – he of bogus claims about MMR and autism – concerning developmental disorders in children, was cited almost 50% more times after retraction than before
I think there is a real danger here with the sudden surge in so-called artificial intelligence. Most of the things that are called AI today are really not intelligent but are merely very polished chatbots. They are built from the ground up to deliver plausiblesounding passages of text, and this has fooled many inexpert commentators into believing that there really is a ghost in the machine.
Services like ChatGPT are something like Google on steroids. We’ve already seen that the bots will invent papers that simply don’t exist, and credit these to genuine researchers, which must be pretty close to slanderous behaviour.
I’m sure it will also cite genuine papers, but which have been retracted, to prove whatever specious point it is trying to make.
Before long it will cite its own artificially generated science, which cites retracted science, in a maelstrom of malevolent misinformation. We’re doomed, I tell you.
Meanwhile, you can while away a pleasant period checking out the league tables of retracted work and the most prolific authors of it at retractionwatch.com. This is an excellent resource that deserves to be better known. And cited.
Among the gems to be uncovered, there is the observation that every one of the 31 authors with the most retractions is male. They each have dozens; some over 100. You could almost admire their sheer nerve in trying to get away with it.
In fact, it’s probably a good idea to doublecheck any published work that your current project relies on before you go too far down the rabbit hole. That’s something the bots are unlikely to ever do.
