AI hallucinations infiltrate academic papers, threatening research integrity

In a recent interview, writer and academic John Duncan highlights the alarming rise of AI hallucinations in academic papers and articles. He attributes this phenomenon to the exploitation of academics, particularly in the UK, who face poor working conditions and low pay, compelling them to rely on large language models (LLMs) for expediency. The resulting errors—ranging from fabricated citations to nonsensical conclusions—pose a direct threat to the integrity of scientific literature and the cumulative production of knowledge. Duncan warns that if unchecked, this could severely impair our ability to respond to future public health crises, as critical research data becomes increasingly unreliable.

The conversation extends to the Pope's recent Encyclical on the AI industry, which Duncan argues is less revolutionary than widely reported. He contends that the document fails to address the structural inequities that drive AI misuse, and he dismisses any notion of 'ethical AI' as a superficial label that distracts from the need for systemic reform. The episode underscores the urgency of safeguarding academic rigor against the seductive but error-prone shortcuts offered by generative AI, especially in contexts where human expertise is already undervalued.