Artificial Intelligence (AI) is everywhere. In the last few years there has been an explosion in the availability of AI tools that can generate text and images, answer questions, summarise information, and more. Naturally, scientists and scholars, with their enquiring minds, are at the forefront of both work on developing AI tools and also investigating how they can make use of them as part of their research.
Springer Nature (SN) has clear policies on what constitutes acceptable usage of AI tools as part of creating a manuscript for publication. These policies are as follows:
- SN does not attribute authorship to AI.
- SN does not allow the inclusion of generative AI images in our publications.
- SN asks peer reviewers not to upload manuscripts into generative AI tools.
There is also more detailed information available on the policy pages of individual publishing groups, for example BMC, Nature and Springer. However, this covers the publishing part of the process and does not consider the use of AI tools as part of the research itself. What is acceptable and how should that be managed? Consider the following conundrum:
Which of the following possible uses of AI tools are acceptable as part of research and should be declared in a submitted manuscript?
A: use of narrow AI (AI systems designed to perform specific, well-defined tasks) to classify cell phenotypes and improve efficiency
B: use of convolutional neural networks (CNNs) for object recognition tasks
C: use of Computer vision (CV) and visual AI as part of research on autonomous driving
D: use of an image-generating tool to create an illustrative image of a key concept in your research, but not data display
E: Use of ChatGPT to generate an introduction for a manuscript, but not the results and discussion
F: Use of an AI tool to improve the readability and grammar of a drafted text
[Image description: a laptop with a hand controlling the mouse pad. Various graphs and visuals appear in the air, linked to an AI symbol]
The correct answers are A, B and C!
Researchers can already use narrow AI, CNNs, and CV/visual AI as part of their work, and this should simply be declared as part of the methodology. However, according to Springer Nature policies, it is not acceptable to use generative AI tools to create images for inclusion in manuscripts or the text of the manuscript itself. This is true even if only part of the manuscript has been generated, or if the image is only illustrative of concepts. Answer F is not correct because use of tools to improve the readability or grammar of a manuscript does not need to be declared at all. However, researchers may choose to use AI as part of their methodology and should then declare that use as part of that section of the submission.
Considering when and how to use AI is an important decision. Researchers will need to take into account not only the reliability and accuracy of the tools, but also the policies of their institutions and the ethical aspects of their research. To help researchers navigate this complex set of considerations, UKRIO has released guidance for researchers – “Embracing AI with integrity: A practical guide for researchers”. This general guide can be used by anyone and includes not only the aspects already mentioned, but also such issues as the impact on creativity and critical thinking and the research integrity considerations. This guide may be useful for anyone wrestling with decisions in this evolving arena.