AI statement

Why a researcher is still irreplaceable

A trained mind, not a stack of answers. A PhD is not mainly about producing a thesis. It is years of training your judgement: noticing what is odd in a dataset, asking a question nobody has asked, and knowing when an answer is too neat to be true. AI tools can produce text and code quickly, but they cannot tell you which question is worth a decade of your life. That judgement is built slowly, through your own mistakes, and it stays with you whatever you do next.

Data starts on the ground. A model is only as good as the data it learns from, and in our work that data comes from the physical world. AI cannot sample trees in a forest, extract DNA at a remote field station, or work through the permits and agreements, such as those under the Nagoya Protocol, needed to move genetic material between countries. Someone has to do that work carefully and honestly, or even the most powerful model has nothing true to learn from.

Trust is built between people. Research runs on trust. A landowner lets us sample their trees, a community shares its knowledge of a forest, a policymaker acts on our advice, and a student confides that they are struggling. That trust is earned through emotional and social intelligence: listening, empathy, keeping promises and showing up. No tool can earn it on your behalf.

Someone must be responsible. When a paper is wrong, a tree is planted in the wrong place, or a conservation decision fails, a person must answer for it. AI tools cannot be authors, cannot be held to account, and do not care about the forests we study. We do.

How we use AI in our lab

These principles apply to everyone in the group, me included.

1. You are the author. You may use AI tools to help you think, code, translate or polish your writing, but every word, figure and line of code you share is yours. If you can’t explain it or defend it, it doesn’t go out.
2. Be open about it. Tell your co-authors and me how you used AI in a piece of work, and follow the rules of the journal, funder or university. Using AI is not shameful; hiding it is.
3. Check everything. Verify every fact, reference and result an AI tool gives you against the original source. An invented citation in a paper or thesis is a serious matter, however it got there.
4. Protect data and people. Do not put unpublished data, personal information, interview material or the exact locations of threatened trees into tools that may store or reuse them. Some of the species we work with are targets for illegal logging, and their locations must stay safe.
5. Respect confidentiality. Never upload manuscripts or grant proposals you have been asked to review, or other people’s unpublished work.
6. Learn first, then speed up. In your early stages, do the hard parts yourself: write your first analysis scripts, draft your first chapters, read the papers in full. Once you understand the work, AI can save you time; before that, it can stop you learning.
7. Use it thoughtfully. AI tools have environmental and social costs. Use them where they genuinely help, not by default.

This field is changing fast, so this statement will change too. If you are unsure whether a use is fine, ask me: there is no penalty for asking. Only for not asking.