Like many college professors today, Sarah S. Richardson has seen AI-hallucinated citations in work submitted by her students. But she was surprised when an editor recently contacted her about a hallucinated citation involving her own work.
A March 31 email from the editor in chief of Disability Studies Quarterly explained Richardson was one of six scholars falsely cited in a journal article. The editor apologized for the error and alerted Richardson of the paper’s retraction.
“The editor took the extra step of informing all scholars who could have been harmed by the hallucinated citation[s] and did so in an exceptionally thoughtful and caring manner,” said Richardson, the Aramont Professor of the History of Science and Professor of Studies of Women, Gender, and Sexuality at Harvard University. “I suspect that most hallucinated citations are never discovered, let alone redressed by a retraction and outreach to those affected.”
Richardson shared the editor’s email with Retraction Watch because she thought his response was “highly meritorious,” and may help “inform best practice,” she told us.
The retracted paper, by Shawna Sheperd-Murtagh, was published in March and explored how disabled content creators enact “joy, desire, and collective care as resistant practices in the face of ableist and racialized digital infrastructures.” Disability Studies Quarterly is the journal of the Society for Disability Studies.
Two weeks after the article was published, a reader emailed the journal to report being falsely cited in the work, according to a June editorial by the journal’s editor-in-chief Jeffrey A. Brune and his assistant editors.
The journal audited all 117 footnotes in the article and found a total of six citations that appeared to be AI hallucinations.
Brune told us most of the hallucinations included the names of known scholars in the field who were attributed to publications they didn’t write. The journal discovered all the false citations were added after the first peer review, the editorial notes.
“The main step toward accountability was to retract the article, which we did right away, but I also felt it was important to be accountable and transparent with the scholars who were falsely cited,” he told us. “For one, we owed them an apology. In addition, I thought it was important that they should be aware of this because they might come across the false citation at some point.”
Brune’s email to the scholars described what happened and the steps the journal was taking to address it. As a scholar himself, Brune said he “certainly would want to know if a publication had been falsely attributed to me.”
Sheperd-Murtagh told us she understood the journal’s decision to retract the article and said she takes responsibility for the errors.
Her paper resulted from several years of research and writing, followed by an “extensive revision process,” said Sheperd-Murtagh, who was a lecturer at North Carolina State University when she submitted the paper in January of this year and is now director of development for Meals on Wheels Durham. She did not use AI to generate the content, she told us, but “experimented with generative AI” late in the editing process while searching for additional scholarship to “deepen particular areas of the article.”
The AI suggested many scholars who she recognized, and because their names and the subject matter “made sense together,” she assumed the information was legitimate, she said.
“The citation issues were serious, and it was my responsibility to ensure that every source I cited was accurate and verifiable,” she told us. “I should have caught these errors, and I regret that I did not.”
Since discovering the hallucinations, Brune said the journal has implemented a new step to prevent such problems. The journal now requires doi hyperlinks for all citations, when available, so reviewers can easily confirm citations during copyediting. The journal also plans to revise its AI policies this summer and add an AI statement to its instructions for peer reviewers, according to Brune’s editorial.
The Disability Studies Quarterly paper is the latest with hallucinated references we’ve reported on recently. Last month, we wrote about the editors of a higher education journal who coauthored an editorial with fake references. In March we reported on a librarian who discovered 12 of 14 references in a Springer Nature paper didn’t exist.
Richardson said Brune’s email was the first time she’s ever been contacted by a journal regarding a false citation linked to her name and she hopes more journals adopt the practice.
“In my view, hallucinated citations are a prime example of the threat of LLM usage to our expert knowledge systems,” she told us. “The articles can be retracted but they are already online and stay online, and the damage of muddle and disinformation is done if others duplicate those citations or LLMs imbibe them as credible.”
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on X or Bluesky, like us on Facebook, follow us on LinkedIn, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].
Yes, I want to know. I agree that all articles containing hallucinated citations must be retracted because the analysis, deduction(s) and conclusion(s) in those articles cannot be considered proper. Proper means the conclusion(s) is(are) derived systematically and consistently with respect to logical, mathematical AND scientific reasoning as required to establish human-level understanding. Keep retracting all articles containing such citations. Correct-sounding and logically and mathematically sound conclusions from the use of AI but that do not make (any) sense (due to improper or gaps in scientific reasoning as stated above) cannot be allowed to be published. All articles written with AI assistance should be published separately in OA pro-AI journals.