• Public

  • Public
  • Network
  • Groups
  • Featured
  • Popular
  • People

Adjust the format of papers to improve description by AI The chatbot ChatGPT and other tools based on large language models (LLMs) can make scientific research more efficient, but they can also introduce mistakes when they describe scientific work (Nature 622, 234–236; 2023). I suggest that small changes to the format of scientific papers could improve the training of LLMs. Describing the research method rather than citing a reference to it, for example, would enable the LLM to summarize methods accurately. Writing in the third person would identify individuals more clearly. This would require a significant shift in conventional scientific writing style, but LLMs do not recognize who ‘we’ are. The use of in-line citations by authors would improve LLM training by reducing the chance of a disconnect between a statement and its supporting reference. These should use just the DOI, without the journal title, volume and page numbers. The DOI makes it easier to verify that a reference is genuine (T. Day Prof. Geogr. https://doi.org/k5f6; 2023). Providing papers with an extended summary written in the third person and in a journalistic style could also improve AI-produced literature reviews. This summary would be more effective than a conventional abstract at highlighting the significance of the work. It would also enhance readability for both machines and humans, helping to boost societal awareness of scientific research.