Rational Polarization

Predictable polarization is everywhere. We can often predict the different directions that people’s opinions—including our own—will shift over time. Empirical studies suggest that this is so whenever evidence is ambiguous, a fact that’s often thought to demonstrate human bias or irrationality. It doesn’t. Bayesians will predictably polarize iff their evidence is ambiguous. And ours often is: the process of cognitive search—searching a cognitively-accessible space for an item of a particular profile—yields ambiguous evidence that can predictably polarize beliefs, despite being expected to make them more accurate. In principle, a series of such rational updates could lead to polarization that is predictable, profound, and persistent. Thus it’s theoretically possible that rational mechanisms drive predictable polarization. It’s also empirically plausible. I present a novel experiment confirming the polarizing effect of cognitive search, and then use models and simulations to show how such ambiguous evidence can help explain two of the core causes of polarization: confirmation bias and the group polarization effect