Utilizing and Citing AI in Your Research
Over the past several months, we’ve collected feedback on the beta Papers AI Assistant. We are thrilled at both the enthusiasm and meaningful use cases our customers have shared with us along the way. We’ve recently encountered a question about how researchers should use and cite the Assistant and other AI tools in their research projects and articles. This is a key question when it comes to artificial intelligence and a rapidly evolving topic throughout academia and across research spheres. Currently there isn’t a set in stone ‘right or wrong’ answer. For now, the following best practices can help shape practices for you, your class, lab or department to leverage AI with integrity and care.
Best Practice Tip #1: Check Journal Policies
There is a broad spectrum of AI sentiment in the scientific journal community. Some journals are embracing how AI can advance scientific research, and some are steadfast that they don’t want AI tools used in any of their articles. Either way, if you are considering submitting a manuscript, it is important that you check journal policies or contact the editor for clarification on what they deem acceptable for AI tool usage.
Best Practice Tip #2: Document It
Transparency is key to responsible and reproducible research and this is a major consideration when utilizing AI tools. If you want to use AI tools in your research, we recommend documenting all the details of your usage, including date and time used, what tools you used (including the version), and the data input and output. Any usage of AI should be clearly described in the body of the article along with the documentation.
Best Practice Tip #3: Understand It
Often, AI tools are mystery boxes and researchers rarely understand the algorithms that determine how their input data is being used or transformed. AI tools will frequently give different answers when provided with the same prompt and thus follow up studies will never be able to exactly reproduce results despite identical inputs. Perhaps this is appropriate in the context of your research methodology, but it needs to be carefully considered before you decide to use AI tools in your research.
Best Practice Tip #4: Don’t list AI as an Author
There is also the question: if an AI tool is used for a significant amount of the data generation or analysis, should it be treated as a co-author? This anthropomorphization is faulty because co-authors are responsible for upholding the integrity of the research and an AI tool is incapable both of agreeing to the authorship contract and identifying follow-up issues with the research.
Best Practice Tip #5: Check out Writefull
Many journals encourage the usage of AI for assisting with grammar and AI tools can be incredibly helpful in writing clear and coherent text. Writefull, part of the Digital Science family, uses AI software to help authors first paraphrase text they wish to reference and then assist in writing and copy editing research manuscripts. For more information including a free account, visit writefull.com.
To summarize, best practices regarding the use of AI tools in 2024 include:
- Check Journal Policies: If you are submitting a manuscript for publication, check the journal policies or contact the editor for clarification.
- Document It: Be very clear and transparent regarding the exact tools you used, dates, and input/output datasets.
- Understand It: Do not ever use any output, including text, from an AI tool in your research that you do not understand fully and can’t be reproduced.
- Don’t List AI as an Author: AI tools are incapable of the full range of responsibilities of authorship.
- Check out Writefull: Grammar AI tools, like Writefull, are great for helping write clear and coherent text.