preLighting with a little help from AI

Written by: Reinier Prosée, preLights Community Manager

Note: this post got a little help from Microsoft Copilot to keep things clear and to-the-point.

Writing is perhaps more an art than a science, something I realised while co-moderating a creative workshop on science communication earlier this year. This is not to say that effective writing isn’t a crucial skill in any scientist’s toolkit. At preLights, we provide a platform that allows early-career researchers to find their own voice (and style) through the practice of writing. Not just writing, but also through reading and editing other people’s work.

Helping each other, as a community of peers, explore different ways to communicate science is key to what we do. As such, the use and potential of large language models—such as ChatGPT, Gemini and Claude AI—in communicating science is something that needs to be explored. In fact, this is already being done: preLight posts have been published that use AI and, importantly, declare its use. Recent examples include posts prepared by a Peer Review class at the Federal University of Rio de Janeiro in Brazil. Here, the students used AI tools to, for example, “assist in language editing and improve clarity during the revision of this manuscript”; a good use of AI and an excellent example of how AI can be used as a tool to make a platform like preLights more inclusive to non-native English speakers.

Besides using AI tools to ensure that texts are free of grammar and syntax errors, it can also help the creative process of writing. In this context, it is sometimes more difficult to distinguish appropriate from inappropriate practices. One wonderful example of using AI to enhance creativity is the article series ‘The Kahneman Chronicles’ prepared by Sameer Thukral, one of our preLighters, for our sister community site the Node. Sameer used Claude AI to (in his own words) “act as an ideation and discussion partner, to hone and refine writing ideas. Free from the bounds of language, I could focus on finding just the right content, flow and ideas, rather than getting lost in expression. In my opinion, my job changed from that of content writer to an editor, director and moderator.”

Another way in which AI tools can influence creative outputs is its ability to create images either for illustrative or aesthetic purposes. Though we welcome and applaud the use of visual imagery in the context of preLights, we have decided against the use of Generative AI or AI-assisted tools to create or alter images. The reason is simple: we often cannot be sure where (elements from) images are taken from and would like to avoid any concerns around copyright.

Whatever is next in the rapidly changing world of AI, for preLights the emphasis remains on voicing the opinions of early-career researcher on the latest developments in biology. So how do we make sure that our posts retain their ‘personality’? Since AI tools excel at providing summaries—BioRxiv started a pilot in 2023 providing AI-generated summaries for preprints posted on their server—we now tend to emphasise the reason for highlighting a preprint on preLights, why others should care, rather than providing a full overview of the preprint findings. We believe this will ensure that preLights remains relevant, championing the voice of early-career researchers.

As part of our commitment to responsible use of AI on preLights, we will be updating our Acceptable Use Policy (AUP) to include specific guidance on the use of large language models. This is a rapidly evolving area, and we anticipate that policies will need to adapt over time. We welcome feedback from preLighters to help shape these updates and ensure our approach remains thoughtful, transparent, and aligned with the values of our community.