DOs and DON’Ts (and really DON’Ts!)
AI is your tool – an assistant, but not your substitute!
The aim of this document
This document outlines how artificial intelligence (AI) tools may and may not be used in our research group. AI can support your work, but cannot replace scientific thinking, responsibility, or accuracy.
Our lab trains ecologists with scientific integrity, professionalism, and responsibility, for which it is crucial that you understand this policy.
Principles
AI may enhance and speed up your work, but you remain fully responsible for:
- Understanding the science and logics behind what you write or analyse
- AI output will be only as good as your prompts. Good prompts need good logic, facts, evidence, background. It needs to come from you understanding your own field and question at hand first. ChatGPT will not help you out in a debate or defend you thesis.
- Verifying methods and results
- ChatGPT suggested you a transformation or test you never heard of? Great. But DO READ more about it before blindly applying to your data!
- Ensuring accuracy, reproducibility, and traceability
- ChatGPT refers to actual documents, pages, papers? It is your responsibility to check these out without blindly trusting the output. Not to mention that AI hallucinates a lot when it comes to resources. Don’t trust anything that you do not read yourself. This should not be a surprise: you would not blindly trust references cited in other papers either, right?
- Ethical and transparent use of information
- You used it? Say so. Specify the level of use openly (help with code, generating a concept figure, cleaning up text…). It needs to be transparent and honest.
AI can help you think. But it cannot think instead of you. (It will try if you ask, but that’s exactly the problem…)
Allowed (DOs)
AI as a scientific supporting tool:
- Summarising scientific papers (only after you read them)
- Explaining statistical methods, ecological concepts, or lab techniques
- Helping with R when you get stuck with your analyses, figures, cannot find where to change font size, colour, something specific in a really long own code
AI as a tool for improving clarity and structure:
- Finding those three final words to shorten the abstract to 300 words
- Editing an email for a more professional tone
- Improving grammar, flow, or readability of text you wrote. E.g. Grammarly or DeepL Write can correct your typos and grammars. If you feel that two paragraphs are rather disconnected, you can ask for suggestions from ChatGPT.
AI as a tool to speed up routine tasks:
- Generating templates (data sheets, lab notebooks, metadata tables)
- Creating simple figures or diagrams after you define what they must show
- Producing example code that you will adapt and verify line by line
- If you know what you need in Excel, it can help you with the functions to make it work
AI as a tool for brainstorming:
- Listing potential hypotheses
- Double-checking your hypotheses for flaws
- Proposing sampling strategies
- Suggesting statistical approaches that you then check
- Guiding further reading (Elicit can be useful, ChatGPT is less trustworthy)
Important: You must always understand the output before using it!
Not Allowed (DON’Ts)
AI cannot generate scientific content in your name
- No writing of entire Introduction/Methods/Results/Discussion sections or even full paragraphs. (If AI writes substantial parts of your thesis, authorship and academic integrity are compromised. It is no different from asking someone else to write it for you.)
- No copying any AI text blindly into your thesis or manuscript without revision and verification.
AI cannot replace hands-on scientific work
- Absolutely no usage of AI to generate lab data, observations, or measurements.
- No AI-written lab protocols or safety instructions without supervisor approval.
- No blind delegation of your reasoning, decisions, or data checks to an AI.
AI cannot replace your understanding
- If you don’t understand the biological mechanism, ecological theory, or statistical model, AI-generated text must not be used.
- AI cannot justify your choices in Methods or Results. Only you can.
AI cannot “smooth over” problems
- Absolutely no rewriting to hide mistakes or poor-quality work.
Don’t forget that AI outputs are not peer reviewed and AI may hallucinate facts. (To cite Harry Potter and the Chamber of Secrets: “Never trust anything that can think for itself if you can’t see where it keeps its brain”.) You must find and cite real scientific literature. Reading literature is not only important for locating facts but it is a key part of your journey in becoming a scientist who can phrase and formulate scientific arguments.
Required Transparency
When preparing a thesis, manuscript, or report, you must be able to answer the following:
- What did you write yourself?
- Where did you use AI?
- Can you explain every sentence without AI help?
- Can you explain the choice of your analyses without AI help?
- Can you explain the logic of the text or your reasoning without AI help?
If not, the work is not acceptable.
If You Are Unsure
Always ask your supervisor.
Final Message
We want you to think, understand, learn, and grow during your time in the lab.
You are not training to become a master of ChatGPT, but an ecologist.
AI can support you as a tool, but it cannot take your place, and it is not a substitute for scientific knowledge and real expertise.
P.S. An Example of Transparency (As Required Above)
To apply this policy to the current document:
This document has been created with the help of ChatGPT.
- What did I write myself? All core ideas, rules, examples, warnings, quirks, and the overall structure and personal style come from me and my personal experience.
- Where did I use AI? I used ChatGPT to polish phrasing, improve clarity, and correct grammar. I also asked for alternative versions of certain sentences to choose the most precise and professional wording and help with organising my points for a more logical structure. Grammarly pointed out typos while I was typing.
- Can I explain every sentence without AI help? Absolutely, as the core idea of this document is based on my own logic, reasoning, and ethical standing as a supervisor. I have read the Harry Potter series multiple times, which is why this particular reference came to me while writing this document.
- Can I explain the choice of my analyses without AI help? Not applicable here.
- Can I explain the logic of the text or my reasoning without AI help? Absolutely, see above; the core of the document is my work, which I can fully justify. Ask me anytime.
Therefore, this use of AI is acceptable under the lab’s AI policy.
Please note that this document is not a substitute for university and institutional rules and regulations and that those policies and any legal requirements supersede anything in this document.
Lab of Zsófia Horváth 