This page contains resources on using Generative AI (GenAI) for research. My work on GenAI in teaching and assessment is on the Teaching tab. Here I will share useful resources and the papers that I have created using supervised GenAI. At the point of launch there is an initial project on the ability to use supervised GenAI to build academic papers from data releases. I also share a writing filter that I have co-created which aims to further facilitate the generation of academic text.
All work is done with OpenAI ChatGPT using the most recent model available at the time. A core distinction is thus made between the work described here and the use of agentic systems to automate the flow. The emphasis throughout is on human supervision of the process through the "chat". Resources are developmental and all comments and feedback are welcome.
This section features work specifically on the flow from data release to working paper draft. Discussion is also made of alternative sharing platforms and writing styles.
This section will be based upon the Department of Social Statistics Seminar delivered at the University of Manchester on 13th October 2026.
A .md file which brings the learning process of academic writing in ChatGPT into a compressed implementable form. Latest v10_2026.
In all of my work with GenAI, the consistent challenge is ensuring that the writing which is produced is of the academic standard needed to communicate with the relevant disciplinary experts. Drawing on my own PhD training with Cochrane (2005), I developed a .md file which can be shared with ChatGPT to help the conversation understand the writing style that is required. The filter, and its development is documented in a working paper.
October 2026 release (10_2026)
The Academic Writing Filter (AWF) is a reusable framework for the development and revision of academic manuscripts. It treats manuscript revision as a constrained process in which improvements to presentation must preserve the evidence, authorised claims and inferential boundaries of the research while also respecting disciplinary and outlet-specific requirements.
The filter provides a structured set of rules and checks covering contribution, manuscript architecture, evidential warrant, disciplinary terminology, literature use, claim strength, robustness, limitations, paragraph construction, exhibits and reproducibility. It is intended for academic journal articles, working papers intended for scholarly circulation, and substantial manuscript revisions.
The Academic Writing Filter can be used:
as a manuscript-development and final-review checklist;
as a shared editing protocol for authors, co-authors and supervisors;
alongside discipline- or outlet-specific writing guidance; and
as an instruction file for AI-assisted manuscript revision.
Use of the filter does not imply use of artificial intelligence. Where it is supplied to an AI system, authors remain responsible for the accuracy of the manuscript and for complying with relevant institutional, publisher and journal requirements.
The filter takes John H. Cochrane's Writing Tips for Ph.D. Students as an important architectural starting point, while developing an independent operational framework for manuscript governance. It should not be interpreted as having been authored, approved or endorsed by John H. Cochrane.
Academic Writing Filter — October 2026 release
Download the Markdown file below:
The Markdown file is the canonical public version of the Academic Writing Filter for the October 2026 release.
A formal citation and associated working paper will be added once the permanent repository record and publication metadata have been finalised.
Author: Simon Rudkin
Affiliation: Department of Social Statistics, University of Manchester
Release: October 2026 (10_2026)
A web resource for social scientists interested in the ways AI use can be incorporated within the research flow. There are resources for flexible adoption, links to workshops (to follow) and illustrated examples. Please use this link Social Sciences