The University of Hong Kong (HKU) is addressing significant concerns regarding the integrity of references within a doctoral thesis published on the city’s plunging fertility rate, after internal review revealed that several citations were likely fabricated by an artificial intelligence tool.
The paper, titled “Forty years of fertility transition in Hong Kong,” appeared in China Population and Development Studies and examines the sociological factors driving the sustained decline in birth rates over the past four decades, primarily attributing the trend to falling marriage rates. The co-authors included doctoral candidate Yiming Bai, Professor Paul Yip of the Social Work and Social Administration Department, and collaborators from the Statistics and Actuarial Science Department and external mainland institutions.
Reference Errors Emerge During Online Scrutiny
The issue came to light following public online critiques that pointed out numerous non-existent sources within the article’s 61 references. Scrutiny indicated that several Digital Object Identifier (DOI) links failed authentication, and purported citations from peer-reviewed journals could not be located in the specified issues. Alarmingly, one unverifiable reference even listed Professor Yip as a co-author.
Professor Yip, who served as the corresponding author, confirmed that the errors stemmed from the lead author’s unverified use of AI software to assist in compiling the bibliography.
“The first author utilized AI to help generate citations but failed to adequately verify their authenticity,” Yip stated in response to local media. While maintaining that the core research findings and substantive content of the study were valid—noting the manuscript successfully passed two rounds of academic peer review—Yip took responsibility for the lapse in editorial oversight.
Corresponding Author Accepts Responsibility
As the corresponding author, Professor Yip publicly apologized for the error and the potential reputational damage to both HKU and the journal. He stressed that the issue was one of process failure rather than deliberate academic fabrication.
“I regret that I did not adequately gatekeep this crucial step,” Yip acknowledged, clarifying that other co-authors primarily contributed data and expert advice, placing the ultimate oversight responsibility on himself and the student.
The paper’s central conclusion calls for large-scale structural reforms to dismantle barriers impeding marriage and family formation, advocating for more family-friendly social and workplace environments to address Hong Kong’s low fertility problem.
Addressing the Impact of AI in Academia
The incident highlights the growing challenge academics face in managing AI-assisted research tools. Professor Yip confirmed he immediately notified the publisher and is personally overseeing the meticulous verification of every source. A corrected version of the article is slated for publication imminently, accompanied by an official notice from the journal regarding the correction.
Moving forward, Yip is taking disciplinary action requiring the doctoral candidate to complete specialized training on the appropriate and ethical utilization of AI tools in academic research. While disappointed, Yip emphasized that the student’s academic performance had otherwise been strong, framing the citation misstep as a learning opportunity from a first attempt at using these evolving technologies.
This situation serves as an urgent reminder for academic institutions globally: reliance on automated tools, particularly for critical elements like verifiable sources, necessitates rigorous human verification to uphold the fundamental tenets of scholarly integrity. The use of generative tools must be accompanied by explicit institutional guidelines and mandatory training to prevent similar breaches in academic standards.