Inference by Experience and Cognitive Sophistication
These graphs plot over- and underinference by task experience and cognitive sophistication.
Weak signals correspond to those that would move a Bayesian from a prior of 0.5 to a posterior between (0.5, 0.6); strong signals correspond to those that would move a Bayesian to a posterior between (0.7, 1). The y-axis measures how much weight subjects put on the signal, relative to a Bayesian. Values greater than 1 indicate overinference; values less than 1 indicate underinference.
Subjects are asked to repeat the task with different signals strengths; the top panel shows that they infer less from weak signals, and infer more from strong ones, as they gain experience. The bottom panel shows that subjects who perform better on a cognitive reflection task infer less from weak signals and infer more from strong ones. These heterogeneities are consistent with a theory of cognitive imprecision in which experience and sophistication lead to more precise inference.
Error bars indicate 95% confidence intervals.