Paper short abstract:

This study explores the genealogy of the use of generative AI as an English-language assistance tool by global scholars. Generative AI in this context is a double-edged sword: the diversity of authors who publish in academia may increase, but only by conforming to expected language norms.

Paper long abstract:

Peer review is a technique to quantify differences between what is and is not publishable in the sciences. One increasingly quantified metric is English. Over 90% of indexed journals in the natural sciences, for example, are published in English, and writing which is not assessed as appropriately English will be rejected. As such, generative AI has been celebrated as a boon for inclusion, allowing scientists to instantly "fix" their writing prior to peer review. At the same time, by performing an automated, imperceptibly quantified version of English, these tools reinforce the global hegemony of English in the sciences by masking the language diversity of academic community members and limiting the exposure of readers to writing in different languages and language registers.

This study explores the genealogy of English metrics in the sciences and the use of generative AI by global scholars to resist those metrics. First, we analyze historical peer reviews and rubrics from major scientific publications to explore how reviewers police, negotiate, and uphold English as a publishing requirement. We compare peer review evaluations which critique English use before and after the introduction of ChatGPT in 2023, exploring the implications of widespread use on adjudication practices. Next, we interview scholars to probe motivations and perceptions of the use and users of automated writing tools. The goal of this project is to use this moment of change in historical publishing norms as a lens to de-legitimate, un-hide, and re-question the global quantification infrastructures which determine who participates in global academia.