When new evidence kills one conclusion, do the decisions built on it change too? I tested six update strategies, then a local model.