Mapping Medieval Population
An ML approach to recovering previously unobservable medieval population density from indirect historical signals.
Status: In progress (pipeline/data work; co-author actively working)
Medieval population data is fragmentary: direct counts don’t exist, and indirect signals — tax records, church registers, place-name density — are scattered and inconsistent. This project builds a machine learning pipeline to synthesize these signals into spatially continuous estimates of medieval population density.
The approach recovers measurements that were always implicit in the historical record but never extracted at scale. It connects to the broader agenda of using ML as a measurement tool to extend the empirical frontier of economic history into periods where data is thin.
Links: Slides — Counting houses: Mapping late medieval populations using the DOMUS model
With Chiara Zanardello.