Responsible AI for Researchers: A Practical Use Framework
Use generative AI for productivity while protecting research integrity, confidentiality, verification and disclosure.
Use AI for tasks it can support safely
Generative AI can help brainstorm search terms, explain unfamiliar concepts, restructure notes, generate coding templates or improve workflow efficiency. It can also make confident mistakes. The researcher remains responsible for checking facts, calculations, references and interpretations.
Do not treat generated references as verified evidence
AI systems may invent citations or mix details from different sources. Search scholarly databases directly and verify every source against the original publication. A reference is not trustworthy merely because it looks correctly formatted.
Protect confidential and personal data
Do not paste identifiable participant data, confidential client files, unpublished sensitive results or restricted organisational information into a tool unless its data-handling terms and your ethics or institutional rules permit it. Anonymisation and approved platforms may be necessary.
Keep the intellectual contribution visible
AI can assist with drafting or language refinement, but research questions, methodological decisions, analysis and scholarly interpretation should remain accountable to the researcher. Follow the university, journal, employer or funder rules that apply to AI assistance.
Document and disclose use when required
Where policies require disclosure, record which tools were used and for what purpose. A simple internal log can capture prompts, outputs used, verification performed and changes made by the researcher. This is particularly useful when AI contributes to a research workflow rather than only spelling or grammar correction.
Use prompts as structured instructions, not magic commands
A good prompt defines the task, context, constraints and desired format. Iteration can improve usefulness, but no prompt removes the need for verification. The strongest workflow combines AI assistance with source checking, domain knowledge and human judgement.