Artificial Intelligence at GDD

AI tools may improve quality, efficiency, and reproducibility of scientific research when confidentiality is protected, outputs are critically verified, and responsibility remains with the researcher. The landscape is changing daily with AI and the following pages reflect our attempt to provide guidance and a framework in which we can all work together with AI in a positive way.

This page contains the principles that apply whenever GDD work involves generative AI. The topic pages provide longer practical guidance where it is needed, but recall that institutional, funder, journal, partner, ethics, and legal requirements also apply; in practice, follow the strictest applicable requirement.

Our group's policy around AI is a living document. We will continue to refine these pages as we discuss further as a group and AI applications evolve.

Key principles

Protect people and data

No identifiable or sensitive data enters a generative AI system. Give AI tools access only to the files, directories, repositories, and services they need. If you are unsure whether material is safe to use, stop and ask before sharing it.

See AI and Research Data for the detailed rules, including minimal data extracts, generalisation, small-cell checks, dummy data, and what to do if data is shared accidentally.

Keep humans accountable

Researchers remain responsible for methods, code, claims, citations, interpretations, and final outputs. AI output is not evidence. Verify it independently, and review AI-generated analysis code before it contributes to scientific work. AI systems cannot be authors or accountable investigators.

See Coding & Analysis for practical expectations.

Preserve thinking and learning

AI should support thinking and skill development, not replace them. For GDD scientific writing, the researcher writes the first draft. Trainees should be open with supervisors about substantive AI use and follow any applicable course or programme rules.

See Scientific Writing for the longer guidance.

Be transparent

Record and disclose meaningful AI use when it affects scientific content, analysis, interpretation, decisions, or wording beyond routine spelling and grammar assistance. Follow the current rules of the relevant journal, funder, institution, and partner.

See AI Use & Disclosure for what to record and adaptable disclosure wording.

Treat higher-risk uses separately and with greater consideration

This handbook does not authorise deployment of AI in field settings, clinical tools, or participant-facing research. Those uses require separate ethical, scientific, security, and regulatory review.