Chapter 2 Culture and conduct
by Jess Grembi
2.1 Lab culture
We are committed to a lab culture that fosters creativity, integrity, enthusiasm, inclusivity, and rigor. As an interdisciplinary team, our differences are our strength. We aim to work with each other in a collaborative, supportive, inclusive, and open manner, and our lab space is free from discrimination and harassment.
We aspire to be a lab where everyone feels motivated to share their thoughts and ideas in a respectful and constructive way. Each of us sees things differently and comes to the table with different expertise. Even new lab members who are still learning about our research can offer valuable critical evaluation of our work and are exceptional resources to gain feedback on what is not clear in how we present our research.
We work together to share our knowledge and seek assistance when needed. This lab manual is a resource for sharing such information with each other, in particular how to get your analyses running smoothly. Please let Dr. Grembi know if you have ideas about new content to add!
2.2 Protecting human subjects
All lab members must complete CITI Biomedical Human Subjects Research (IRB) Course Stage 1 - Basic Course training and share their certificate with Jess. She will add team members to relevant Institutional Review Board protocols prior to their start date to ensure they have permission to work with identifiable datasets.
One of the most relevant aspects of protecting human subjects in our work is maintaining confidentiality. For students supporting our data science efforts, in practice this means:
- Be sure to understand and comply with project-specific policies about where data can be saved, particularly if the data include personal identifiers.
- Do not share data with anyone without permission, including to other members of the group, who might not be on the same IRB protocol as you (check with Jess first).
Remember, data that looks like it does not contain identifiers to you might still be classified as data that requires special protection by our IRB or under HIPAA, so always proceed with caution and ask for help if you have any concerns about how to maintain study participant confidentiality.
2.4 Responsible use of AI tools
This guidance draws heavily from a version shared by Dr. Kim Meier (Assistant Professor at the University of Houston)
I encourage lab members to think of large language models (LLMs) and other AI tools the way we think about calculators or Wikipedia: they can be incredibly useful, but only if you already have an understanding of what you’re doing. They’re best used as assistants – not as authorities or substitutes for careful thinking. This guide outlines how I expect AI tools to be used in our lab context. This guide was written with the help of ChatGPT.
2.4.1 Philosophy
AI tools can help us work more efficiently, learn more quickly, and get past sticking points – whether that’s debugging a script, brainstorming how to visualize data, or wading through a dense research article. I see them as part of the broader toolkit we use to support our work, like textbooks, search engines, or discussion with labmates. But just like those other tools, the usefulness of AI depends on the thoughtfulness of the person using it. It can help you think, but it can’t do the thinking for you. Think of AI tools like a lab whiteboard: a place to sketch ideas, not publish results. It will not take responsibility for errors, misinterpretations, or ethical lapses. That’s still on you.
2.4.2 Helpful and encouraged uses
You’re welcome to use AI tools to assist with tasks like:
- Googling stuff! Even excluding Gemini, Google search uses AI
- Getting help understanding a research paper or background topic
- Brainstorming or refining ideas (e.g., for a figure, a method, or a statistical approach)
- Getting feedback on a piece of writing you drafted
- Asking for explanations of code behavior or R functions
- Writing or revising your own code, as long as you understand what the code is doing
- Generating visualizations or suggesting ways to summarize data
It’s fine to use AI to “talk things out.” Just don’t stop there – make sure you apply your own judgment and review anything it suggests carefully.
2.4.3 Use with caution: known limitations
AI tools often sound confident even when they’re wrong. They can fabricate citations, introduce subtle bugs in code, or suggest statistical approaches that don’t fit your design. Some specific risks include:
- Code suggestions might run but be logically incorrect
- Statistical advice might not match your data or assumptions
- Summaries of research papers can miss key details or invent conclusions
While AI tools are improving — for example, ChatGPT now offers a DeepResearch feature designed to reduce citation hallucinations and summarise literature — these are still limited to open-access articles only, and errors still happen. Always verify citations at the source. To use DeepResearch, you have to ask ChatGPT to use it as it won’t search carefully by default. For example, you would need your prompt to be something like: “Using deep research, give me a paper that describes X”
If you don’t already understand the concept or method, you probably won’t be able to tell if the AI got it wrong. That’s a sign you should pause and ask a labmate (or me) for help.
2.4.4 Boundaries and responsibilities
- No uploading data: Never paste data, participant responses, or sensitive information into AI tools – even in small samples. That includes data from active studies, even if it looks anonymous. If you’re getting help with code, describe the structure of your data instead (e.g. “I have a 64×1000×120 matrix called motionData representing channels x timepoints (in ms) x trials”, or “I have a data frame in R where each row represents a participant and the columns include group, age, and threshold”). If you’re not sure how to describe your data without showing it, ask a labmate or bring it to our next meeting.
- No identifiable content: Avoid including names or sharing personal details with these tools, including information about yourself - whether you’re getting help with writing a reference letter, drafting an email, or talking through a lab-related issue. Even if it seems harmless, AI tools are not private spaces.
- You’re responsible for vetting the output: If you don’t understand the code, math, or language an AI tool gives you, don’t just copy and paste it into your project. Run it by someone else, or step back and learn the concept first. That’s how you grow as a researcher. AI is here to help you work, not do your work for you.
- No AI-written research content: AI should never be used to generate text for a research paper, abstract, or manuscript section. You can use it to help organize your thoughts or improve the clarity of something you wrote – but the core ideas, language, and citations must be your own. Even if you’re using AI to help you phrase something better, avoid copy-pasting text that it wrote for you. You’ll learn a lot more (and avoid problems) by rewriting in your own words.
- This includes embedded AI tools: This policy applies not just to standalone tools like ChatGPT, but also to built-in features like Cursor, GitHub Copilot, Microsoft Editor, or Google Docs suggestions. Use the same judgment no matter where the AI shows up.