The use of artificial intelligence (AI), in particular Large Language Models (LLMs) is changing the world. There are lots of wonderful attributes, but also some important pitfalls that require scientists to be direct and clear whenever it is used to maintain high quality and reproducible science.
The following rules and guidelines surrounding the use of AI in our research were developed by the lab. Tools are named as practical suggestions, not endorsements; equivalent alternatives are fine, and any tool that stores lab data should follow UC Davis data-security guidance.
Protect Sensitive Information: Do not use GenAI tools with sensitive systems and data. Do not provide any text or other inputs to large language models or generative AI that might contain sensitive data, including identified or de-identified student information, other’s publications/grants that are not public or owned by you, personnel data, personnel conduct information, patient data, confidential financial data, confidential research information, or otherwise sensitive information.
Appropriate uses
- Code writing and debugging (e.g., with GitHub Copilot).
- Improving plots in R or Python.
- Editing an existing draft’s spelling, grammar, logical flow, and sentence structure (see Writing papers). This can include abstracts, papers, and your own grants/fellowships. You must always have explicit permission if you are not the owner.
- Brainstorming ideas (as long as you do not limit this to only AI brainstorming).
Inappropriate uses
- Initial abstract or manuscript drafts.
- Creating generated images for publication.
- Uploading any data produced or collected in our lab or elsewhere without express written permission (see Data management).
- Fabrication or falsification of data.
- Sharing any personnel or student information.
- Presenting AI content as your own. All usages must be acknowledged. Note that AI should never be included as an author.
- Using AI to replace your learning or problem solving. You do yourself a disservice and future tasks may become more challenging.
Don’t take AI output at face value; verify it and do your due diligence before you rely on it.
AI golden rule: use AI with others as you would want them to use AI with you. Source: Stanford Report
When you do use AI
Acknowledge your use of AI
If a generative AI tool (e.g., ChatGPT) is used, your research publications (thesis, conference papers, and journal papers) must acknowledge how it was used, even if no generative AI content was incorporated in the work. This acknowledgement should mention:
- Which AI tool (including the version #) was used
- How the AI tool was used
Be aware of AI limitations
- Bias and stereotypes. Generative AI is trained on large amounts of text and data collected from across the internet, so it is possible for responses to include prejudice and bias. Researchers and developers are trying to mitigate these issues, but humans are biased so systems trained on our outputs will be biased as well. Be aware that these tools cannot recognize bias, and may perpetuate them.
- Fake citations. Generative AI can produce fake information (hallucinations), including fake references and citations, which can lead to academic misconduct and plagiarism. Because its outputs draw on real source material, AI text can also reproduce that material closely enough to count as plagiarism, even unintentionally.
- Limited calculation ability. Generative AI is not designed to perform calculations, thus its responses may be incorrect and inconsistent. You must always compare quantitative results with tools designed for calculations.
Follow the UC Davis Code of Academic Conduct
- Review and verify. Necessary for any AI-generated content that you publish or share with others. Always verify the accuracy of AI-generated output.
- Consider copyright issues: AI-generated content can contain copyrighted or plagiarized material. You are responsible for any AI-generated content that you publish or make available to others.
Never copy and paste AI-generated language. Even if acknowledged, the risk that you unknowingly plagiarize a colleague is huge.
References: UC Davis & Generative AI guidance · UC Davis student guide to GenAI
Spot something to improve? Anyone in the lab can edit this guide. Edit this page suggest a change how edits work
Last updated