Updating Beliefs When Evidence is Open to Interpretation: Implications for Bias and Polarization
Roland G. Fryer Jr.
Harvard University - Department of Economics; National Bureau of Economic Research (NBER); American Bar Foundation; University of Chicago
Institute of Mathematical Stochastics
Matthew O. Jackson
Stanford University - Department of Economics; Santa Fe Institute; Canadian Institute for Advanced Research (CIFAR)
May 1, 2016
We introduce a model in which agents observe signals about the state of the world, some of which are open to interpretation. Our decision makers use Bayes' rule in an iterative way: first to interpret each signal and then to form a posterior on the sequence of interpreted signals. This 'double updating' leads to confirmation bias and can lead agents who observe the same information to polarize: there can be a greater distance between their beliefs after observing a common sequence of signals than before. Such updating is fully optimal if agents have bounded memory and sufficiently discount the future. If they are very patient and have bounded memory, then a time-varying random interpretation rule (still double-updating) is optimal. We explore the model in an on-line experiment in which individuals interpret research summaries about climate change and the death penalty and report beliefs. Consistent with the model, there is a significant relationship between an individual's prior and their interpretation of the summaries. More than half of the subjects exhibit a polarizing behavior - shifting their beliefs further from the average belief after seeing the same summaries as all other subjects - something that is inconsistent with standard Bayesian updating, or even naive updating, but consistent with our model.
Number of Pages in PDF File: 44
Keywords: beliefs, polarization, learning, updating, Bayesian updating, biases, discrimination, decision making
JEL Classification: D10, D80, J15, J71, I30
Date posted: June 5, 2013 ; Last revised: May 12, 2016