Research Integrity Indicators in the Age of Artificial Intelligence

Generative artificial intelligence (AI) and large language models significantly change how disciplines and communities analyze and report research. Leveraging these new tools, such as ChatGPT or Bard, authors can easily generate text and analyses for research articles. As a result, we have already witnessed several instances in which generative AI was used to write a paper or manuscript which unknowingly contained fake citations or false information. The scholarly community needs new indicators to signal, assess, and evaluate manuscripts and research quality to fortify public trust in research. This paper proposes a set of indicators for research integrity that encompasses the much-needed transparency for generative AI. We have then used AI to train algorithms to detect these indicators and applied them to 33 million full-text research publications. We can now see the key indicators as metrics to understand where various fields of research are in communicating and signalling trust.