Cochrane announces new policy on AI generated content
Cochrane has introduced a new policy addressing the use of artificial intelligence (AI) in generating written content for its publications. We have been carefully monitoring the use of Artificial Intelligence in the publishing sector and developed this new policy in response to the rapid emergence and adoption of AI writing tools.
It's important to clarify that this policy exclusively covers AI writing tools and does not impact the already established use of approved AI tools for certain tasks within Cochrane's review processes. This is something we are continuing to monitor and review.
Use of AI for writing is permitted, but we need to be transparent
Our policy has deliberately been designed in line with the guidance from the Committee on Publication Ethics (COPE) and the AI policy of our publisher, Wiley. Like them, we are open to the benefits that AI may well have on writing systematic reviews, so our policy permits the use of AI-generated content under certain conditions:
AI tools cannot be credited as authors
Authors bear full responsibility for the article's accuracy and validity
A transparent and detailed description of the AI tools used, and content generated is mandatory
The policy aims to encourage transparency and the use of validated AI tools for writing rather than endorsing specific tools. Given how quickly things are moving in AI, we realise that a flexible approach is necessary, and we will be sure to evaluate manuscripts that disclose the use of AI writing tools to inform future iterations of the policy. Authors are entrusted with the task of selecting their tools wisely and are ultimately responsible for the ensuring the manuscript they submit upholds Cochrane standards of quality and integrity.
“We are deliberately adopting a cautious approach. Our primary focus will always be to ensure the trustworthiness of our evidence and the integrity of our research”
- Ella Flemyng, Cochrane’s Head of Editorial Policy and Research Integrity
Acceptable use of AI in review conduct
Cochrane acknowledges the vast potential of AI for research development and publication, while recognising the unique challenges it presents. Any AI tools incorporated into our review processes must uphold the trust that users place in our evidence, so validating these tools is crucial.
One successful instance of AI developed by Cochrane and already in use within our processes is the Screen4Me service. It combines volunteer citizen scientists and machine learning to reduce the number of references authors need to assess during screening by excluding those that are not randomised controlled trials (see the published evaluation). Covidence now also uses these machine learning parameters, and predicts this will automatically remove nearly half of the references before manual screening. Please see Cochrane’s author guidelines for full details about Screen4Me and the other types of support available to authors.
Another example is the AI-powered quality assessments of computer-assisted translation within our translation management system Phrase TMS, which continues to develop AI functionalities to enhance machine translations.
What next?
The future in this area is potentially very exciting and Cochrane will continue to explore acceptable applications of AI tools while balancing that with upholding research integrity and transparency standards.
Ella Flemyng, Cochrane’s Head of Editorial Policy and Research Integrity explains: “At Cochrane, we’re excited at the potential of using AI to enhance the efficiency and accuracy of our review processes. However, we are also deliberately adopting a cautious approach. Our primary focus will always be to ensure the trustworthiness of our evidence and the integrity of our research, so we will look very closely at new AI tools before their integration and will maintain complete transparency in introducing them.”
We continue to be open to collaborating with others in the sector to establish validation standards to ensure that any AI tools used do not compromise people’s confidence in Cochrane evidence. Opportunities for collaboration also come in the form of re-using Cochrane data from published Cochrane Reviews to inform AI developments. The Cochrane Library data request page has full details about terms and conditions for both commercial and non-commercial re-use.
For more details on Cochrane's process for proposing new tools and methods, including AI tools, please refer to the Cochrane Methods website.