Interacting Mental Models

Abstract
People rely on mental models to make predictions, yet these models often differ across individuals—even when based on the same information—contributing to persistent disagreement. In practice, however, individuals rarely form models in isolation: they observe and interact with others’ forecasts. We study whether and how such social exposure reshapes mental models using a controlled rule-learning experiment. Specifically, we ask whether disagreement leads to convergence toward the true model, persistence of heterogeneous beliefs, or coordination on a misspecified model. We further test whether attentional constraints during initial learning limit subsequent model revision due to social learning by varying whether participants can revisit the original data after social interaction. We find that participants learn from others, but that exposure to divergent predictions alone is insufficient to induce substantive model revision. Joint paper with Benjamin Bushong and Tristan Gagnon-Bartsch.