Editorial Position of the American Thoracic Society Journal Family on the Evolving Role of Artificial Intelligence in Scientific Research and Review

Abstract

In recent years, significant advances in large language models (LLMs) and generative artificial intelligence (AI) have caused a rapid integration of AI tools into research (1).For example, researchers are increasingly able to leverage these tools to support literature reviews, data management and analysis, data synthesis, and the editing of scientific writing (2-4).On the editorial side, many medical journals now use AI tools for detection of plagiarism and image manipulation, identification of expert peer reviewers, and other applications to support the publication process.However, with the rapid integration of AI tools, clear challenges have arisen that both authors and editors must navigate-together, we hope.If not, these challenges potentially undermine the goals of academic publishing, which we posit is to curate and disseminate new knowledge toward the pursuit of truth.For example, LLMs may base their output on outdated or lower-quality information or generate content that appears to be of high quality but is factually inaccurate.Furthermore, bias in algorithm output and lack of transparency ("black box") in how LLMs operate raise concerns about the appropriate uses of such tools (5).The foundation of the scientific publication community, peer review, relies on human judgment and expertise, and therefore unfettered adoption of these AI tools runs the risk of undermining the quality and the community's trust in the integrity of the publishing process.As the scale and speed of these tools expand, their benefits to the academic and publication communities will grow.Still, it is imperative that we balance excitement with vigilance for their potential limitations and known flaws.

Ähnliche Arbeiten