Are we all system architects now? (I hope not!)

Brief remarks on using AI in political research without losing what drew us away from more stable jobs

Allison Koh

Centre for Artificial Intelligence in Government at

EPSS — June 2026

As background, this is a flexible format “EPSS” panel, rather than a standard paper panel. So there are no papers and no discussants. The idea is to open up a broader discussion about AI in political science (in politics and in research about politics) that could be of broad interest to members. So contributions could range from short presentations of new tools, to presentation or discussions of new workflows, to broader remarks – or presentation or research– on impact for politics, for research, or for teaching. A simple format would be time for each person to present some thoughts, with or without slides, and then open up to a general format moderated by Ken.

I’d suggest that we all get five slides each, max, for your opening remarks, then we have a roundtable discussion and Q&A. The slides are just for a visual presentation or demonstration of what we are talking about, not a tutorial, limited to no bullet points.

Hi! I’m Allison

I use a range of computational methods to study the geopolitics of communication technologies.
I also integrate open-source coding agents into workflows for analyzing data on human rights violations.
I’ve been very into coding and open-source online communities since ca. 2007

My stack

Task	Tool(s)
Lit review/annotated bibs	Retrieval-augmented tools (e.g. Perplexity, Elicit)
Coding & analysis (command line)	Claude Code + open source alternatives
Revisions/framing/phrasing	Formalizer by goblin.tools + ollama

Hot take?

As coding agents become integral to computational workflows in the social sciences, we risk conflating getting more done with doing fulfilling work.

What to automate? A Rorschach-esque test

Food for thought

In what areas might AI use be less efficient?
How to use AI within coauthor teams when optimal AI use is highly personalized?
In what ways might AI use accelerate burnout?
How to stand out when AI standardizes formatting?
How to figure out what’s optimal for me?

Thank you!

a.w.koh@bham.ac.uk

https://allisonkoh.github.io/

🦋 @allisonwkoh

/allisonkoh/