AI revolutionises academic writing, prompting debate over quality and bias

In a groundbreaking shift for the academic world, AI now contributes to at least 10% of research papers, soaring to 20% in computer science, according to The Economist. This transformation is driven by advancements in large language models (LLMs), as highlighted in a University of Tübingen study comparing recent papers with those from the pre-ChatGPT era. The research shows a notable change in word usage, with terms like ‘delivers,’ ‘potential,’ ‘intricate,’ and ‘crucial’ becoming more common, while ‘important’ declines in use.

Researchers are leveraging LLMs for editing, translating, simplifying coding, streamlining administrative tasks, and accelerating manuscript drafting. However, this integration raises concerns. LLMs may reinforce existing viewpoints and frequently cite prominent articles, potentially leading to an inflation of publications and a dilution of research quality. This risks perpetuating bias and narrowing academic diversity.

As the academic community grapples with these changes, scientific journals seek solutions to address the challenges as the sophistication of AI increases. Trying to detect and prevent the use of AI is increasingly futile. Other approaches to uphold the quality of research are discussed, including investment into a more solid peer-reviewing process, insisting on replicating experiments, and hiring academics based on the quality of their work instead of quantity, promoted by public obsession.

Recognizing the inevitability of AI’s role in academic writing, Diplo introduced the KaiZen publishing approach. This innovative approach combines just-in-time updates facilitated by AI with reflective writing crafted by humans, aiming to harmonize the strengths of AI and human intellect in producing scholarly work.

As AI continues to revolutionize academic writing, the landscape of research and publication is poised for further evolution, prompting ongoing debates and the search for balanced solutions.