petersuber
@petersuber@fediscience.org
I work for #OpenAccess to research. Formerly directed the Harvard Office for Scholarly Communication. Now semi-retired but still working for the cause. Also an emeritus professor of #philosophy and non-practicing #lawyer. Posting on open research, academic culture, US politics, democracy. Dad of daughters. For #OA news, also follow the Open Access Tracking Project (@oatp). I aim for comprehensive coverage there, not here.
fediscience.org
When #AI tools hallucinate citations, their made-up author names are not random. "We show that #LLMs do not merely default to high-probability individual names when generating fictional experts: they produce correlated character ensembles, pairs and trios [of co-authors] whose co-occurrence rates far exceed chance and are consistent across independent generations."
https://arxiv.org/abs/2606.02184
They fool people and tools:
"We identify 1,655 ghost-authored records claiming nonexistent journals with fabricated publication dates…Server-side #DataCite timestamps prove deliberate backdating…These carry real #DOIs…making them harvestable by any scholarly aggregator that ingests DOI metadata."
#Citations #Hallucinations #ScholComm