What I have been working on lately. I update this page about every four months. For the week-by-week version, see my LinkedIn newsletter.
August 2026
IFRC Everyone Counts. I am lead data analyst on the Everyone Counts Report 2026, the Federation-wide report on what Red Cross and Red Crescent National Societies do and who they reach. My contract runs to the end of August. Most of my time has gone on the analysis behind the chapters, written as scripts so that every figure in the report can be traced back to the data it came from.
Causal Map at EES 2026. Two of our submissions were accepted for the 16th EES Biennial Conference in Lille, 26 to 30 October, on the theme “Evaluation for Vibrant Democracies”.
- Non-mosquitoes do not cause non-mosquito bites, an individual paper with Fiona Remnant, in the strand on responsiveness.
- Interrogating evaluation: a live experiment and conversation, with a little help from AI, with Silva Ferretti, in the strand on new methods.
Causal Map as a chat. We are opening up an optional way to use the app. You tell MapCat, our assistant inside Causal Map, what you want in words. It loads the sources, sets up the coding, filters the map. The buttons and panels are all still there if you would rather click.
Causal Map for Zotero. Since June there is a Zotero plugin: right-click a PDF in your library and get a causal map of what that paper says causes what. It is free up to a monthly page allowance. It is meant for the literature review case rather than for coding fieldwork.
A third Qualia app. QualiaInterviews interviews one respondent at a time. We have been building a group version, where many people answer into one growing network of questions and the AI adds follow-ups and statements that draw several answers together. Early on it looks for the range of opinion, later for where people agree and where they still do not. It is in testing rather than open to the public.
Causal Map Highlights. I now write a second weekly LinkedIn newsletter, Causal Map Highlights, which takes one story out of a real dataset each week rather than arguing about method.