Welcome to “Vibe Science”
Anecdotes mount in a worrisome way, and open data is possibly slopsquatting
Matt Spick, who studies data analytics at the University of Surrey, was quoted in a recent Retraction Watch story talking about studies from public databases being churned out by AI systems — or researchers abetted by AI systems — at a massive scale:
It will be hard to force retractions for a lot of these, and then it becomes a whole philosophical thing for meta-scientists. Should we allow all science to be published, even if it’s meaningless?
Many of these studies share figures despite coming from different authors and institutions (a common reuse tactic LLMs invoke to be efficient) while doing more different analyses than a smart human would to fake sophistication.
The explosion of these dubious papers happened so fast that many were caught off-guard — and it’s not just research papers. In 2024, an orthopaedic journal editor wrote a not unusual appeal to readers to “keep the conversation going” by submitting more letters to the editor (LTEs). Just 18 months later, the same journal had to announce new scrutiny and requirements owing to a relative explosion of LLM-generated or -flavored LTEs.
AI and LLMs are influencing the scientific claims space in multiple and somewhat insidious ways. We’re hearing stories of reviewer decisions overruled without explanation by an AI-driven editorial system, of authors using AI brazenly and challenging editors to do something about it because it will be just “you said, me said” at the end of the day, and reviewers being castigated for using AI to detect the use of AI because it’s OK for authors and editors to use it apparently but not reviewers.
It is no longer “the medium is the message.” There is more and more often middleware sitting in the medium, messing with the message.
Welcome to the era of “vibe science.”