Alex Durk, Communications Manager at Springer Nature and Stempra committee member, shares behind the scenes thoughts on how Stempra’s AI guidelines might be used.
Generative AI has been adopted for use in a wide variety of professions since the first models were released four years ago — including, of course, science and research communications. However, in many use cases, exactly what constitutes best practice is still unclear, and this is particularly true when it comes to the accurate and careful communication of science on which our members pride themselves.
This challenging new working world has become a day-to-day reality for many over the last year or two. To help our members navigate this and make informed decisions about how to use this new technology, the Stempra committee decided last year that we should provide members with some standardised guidance on using generative AI. Having volunteered to lead on the development of the guide when work started in earnest about a year ago, I’m delighted to finally be able to present it to you.
As discussed at the webinar where we presented the first draft to the members back in January, we tried to abide by several key tenets while drafting the guide.
- The guide must assume no prior knowledge. Although many members are increasingly familiar with the concepts underpinning generative AI, we believed that it was important for our guide to be accessible for all regardless of their existing knowledge. As such, the guide begins with a set of considerations outlining the broad principles and concepts that we believed everyone should understand when making decisions about generative AI use in their work.
- Any guidance must be broadly applicable. We felt that given the wide variety of organisations that our members work for — and the wide variety of budgets available! — it would be inappropriate to make formal recommendations for specific tools or processes. Instead, we focused on producing a set of guidelines that propose minimum best practice standards to work from, with enough flexibility to work across the field.
- The guide can only contain guidance! Again, given the range of organisations represented in our membership body, we felt it inappropriate to specify “do”s or “don’t”s unless we had specific legal reasons for doing so. As such, we have tried to keep the language as permissive and open as possible, allowing our members to instead use the guide to develop or supplement the guidelines that apply within their own organisation.
This guide is not intended as a final “one and done” document — that would be particularly inappropriate given the rapid pace of development in the field. Rather, we hope that this will be a document that can be built upon and refined in future, as has happened with our guidelines on the media promotion of preprints. In the meantime, we hope that the principles underpinning the guide allow it to remain relevant for all who need it.
Finally, some acknowledgements. Many of these principles and concepts have been well-explored in other publicly-available materials, and I would like to take a moment to credit some of those organisations for inspiration — in particular the University of Cambridge, Harvard University, CIPR, and Springer Nature.
I would also like to voice my thanks to the other Stempra members who helped draft this document — my fellow committee members Clare Green and Ed Day, and particularly Beth Baker from PLOS, who first suggested to the committee that we should provide the membership with this kind of guidance.
We hope that you find the guide useful!