AI tools can improve equity in science

Berdejo-Espinola, V. & Amano, T. (2023) Science, 379: 991.

After Science updated its editorial policy to ban the use of text generated by AI tools in scientific papers, we wrote a letter to the Editor raising our concern about the decision made. We argue that AI tools like ChatGPT and DeepL might help alleviate current linguistic disparities in academia and thus improve equity in science.

Non-native English speakers face significant challenges when writing scientific papers in English, including rejection and requests for revision due to their English writing. Indeed, human English translation or editing services are costly and time-consuming, creating a multifaceted disadvantage for non-native English speakers. However, AI tools are cost-effective tools that can proofread English text with high accuracy, offering a unique opportunity for all non-native English speakers to have their research edited and proofread quickly and at a low monetary cost. This opportunity might be even more beneficial for those in low-income countries who cannot afford human editing services.

AI tools are evolving at a fast pace, showing great improvements in their accuracy, transparency, and legitimacy. A free tool is now available to distinguish between AI- or human-written text; thus, we suggest that journals allow authors to use AI tools for proofreading manuscripts before submission and request to declare its use (as Nature’s new policy does) and submit the original, pre-AI-edited version as well as the AI-edited version of the manuscript.

AI tools like ChatGPT are exciting technologies that can revolutionise our efforts to overcome linguistic disparities in disadvantaged communities.

View full publication, here.

We are also pleased to see that, in response to our letter, Science now recognises potentially acceptable uses of AI tools for writing papers and may consider adjusting its policies in future.