About this resource

Generative AI can participate in many parts of a social-science research process: proposing questions, finding candidate sources, drafting analytical code, suggesting robustness checks, challenging interpretations, reorganising manuscripts and assisting reflection on research decisions.

A list of capabilities tells us little about the harder methodological questions:

• which outputs may influence the evidence base;
• which decisions require explicit specification;
• how preferred claims should respond to inconvenient evidence;
• how inferential boundaries survive rewriting;
• what evidence supports later claims about intellectual contribution.

This resource therefore focuses on research functions, governance boundaries and implications.

Selected worked examples are drawn from a PISA research project.

What this resource is for

Use the material to:

• design a governed AI-assisted research workflow;
• distinguish proposal generation from verified evidence;
• document consequential analytical decisions;
• challenge a preferred interpretation rather than merely elaborate it;
• record when evidence narrows or withdraws a claim;
• make contribution claims more specific and auditable;
• separate contemporaneous evidence from retrospective reconstruction;
• state an authorised claim boundary before substantial rewriting;
• audit whether edited prose still maps to current evidence.

What it is not

This resource is not:

• a claim that AI should replace researchers;
• a claim that human participation automatically guarantees quality;
• a recipe for recording every research action;
• a software repository;
• empirical validation of the reason-provenance framework;
• evidence that the Academic Writing Filter improves writing or publication outcomes;
• a substitute for subject-matter, statistical, qualitative, ethical or methodological expertise.

Attribution and desirability are separate questions. A workflow can be transparently attributable without being a division of labour that a university, discipline, journal or researcher should endorse.

Acknowledgements

Thanks are due to Professor Jason Thatcher for his inspirational discussions on AI and research. I also thank seminar participants at The University of Manchester Department of Social Statistics for their feedback on the project.

Licence and reuse

Unless otherwise noted, original text, templates and documentation in this repository are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Appropriate credit, a link to the licence and an indication of changes are required when material is reused or adapted. Third-party and externally linked material retain their own rights.

Full repository rights information is recorded in the LICENSE file.