Generative AI is now involved in enough academic and analytical work that saying nothing about it is itself a position. This page says where I use it, where I do not, and what you can assume about anything published under my name.
Where I use it
- Code. I use AI assistants when writing R and Python: debugging a function, working out why something fails and converting an approach from one programming language to another. I read word by word and run everything before it goes anywhere near an analysis.
- Editing. Tightening prose I have already written, and catching the places where a sentence says less than I meant it to.
- Learning. Getting an explanation of an unfamiliar method, then checking it against a textbook or paper rather than trusting it.
Where I do not
- Numbers. No figure, table, coefficient or p-value on this site or in my research was produced by asking a language model for it.
- Citations. I do not let a model supply references. Fabricated citations are the single most common way AI-assisted academic writing goes wrong, and every reference I give is one I have opened.
- Anything presented as my own judgement. Opinions, recommendations to students and clients, and assessments of other people’s work are mine.
Research
For my thesis to published research article, everywhere when modelling, preparing data and interpreting findings are solely done by me. AI assistance was involved in writing, debugging code and grammar correction, not in generating results or conclusions.
This website
Although the initial versions of this website (my first one was in 2019, which I took down due to maintainance issue and later created this one with new theme) was done solely by me watching hundreds of hours of tutorials, over time, I took advantage of AI tools, specifically Claude to modify the templates, layouts and configuration behind this site. It took several sessions, with me directing and reviewing the work. The posts, the course material and the statements of opinion are mine.
In the classroom
I teach statistics, and I would rather my students learn to use these tools well than pretend they will not use them. What I ask of them:
- Use AI to understand something, not to avoid understanding it. If you cannot explain your own code line by line or explain statistical findings, you do not yet have an answer.
- Say so when you have used it, and say what for. Disclosure is not a penalty.
- Never submit numbers, citations or claims you have not verified yourself.
Assessment in my courses is designed around this: if a tool can produce the whole answer, the question was not worth asking.
Training on this site
Option A, permissive: Everything here is published under CC BY 4.0. You may train on it under the same terms as any other reuse, with attribution.
Option B, restrictive: I would rather my writing and teaching material were not used to train commercial models. That preference is not enforceable by a licence, and I know it, but it is a preference I would like on the record.
Option C, declining to answer: I have not settled on a view here yet.
Last reviewed September 2026. If something on this site looks wrong, machine made or badly reasoned, tell me; I would rather hear it.