Does each forecaster hold one world model?

We ran world modeling on our BTF forecasting questions: related questions are grouped into themes, each agent's forecasts are held against a single view of the world per theme, and forecasts that do not fit that view get revised. Each dot is a question. Color shows the direction of revision, size how far it moved. Click a dot to read why it moved, or a circle to read a theme's house view. These revisions improved accuracy for every agent we tested.

revised up revised down unchanged carried in from an earlier round dot size = probability points moved
The 24 biggest updates as a list