Real and Assumed Information

We consider a simplified version of the model in Augenblick et. al (2025), in which people use models to translate data into beliefs and predictions. We posit that they make correct inferences given a model but not unconditionally, forgetting model uncertainty, and demonstrate two implications: first, that new information can decrease confidence rather than increase it, if it prompts reconsideration of assumed models; second, that people with models may misunderstand what kind of information is persuasive to share with others.