HCL+PrioR mtg. Kevin Dorst. Monday 12/8 1:30-2:30pm ET

[Antonio Alonso Arechar]

Good morning everyone,

Kevin Dorst will present next week (12/8 1:30-2:30pm ET; see blurb below).

Lab meeting will be hybrid, with a zoom link (https://mit.zoom.us/j/91655448125) and in person in the conference room on the 15th floor of E94 (1579; 245 First Street).

Best regards,

Antonio

---

Title: Accurate Judgments Drive Polarization

Abstract: Confirmation bias is pervasive, important, and misunderstood.  People do not simply dismiss incongruent evidence: arguments work, shifting beliefs regardless of prior opinions.  But nor are people convinced by bad arguments: quality matters in how they react.  I propose that people exhibit confirmation bias when they predictably shift their beliefs in the direction that they search for evidence. This violates the `martingale' principle of standard Bayesian updating.  But I show that once we permit ambiguity—uncertainty about our own probaiblities—even Bayesians violate martingale when searching for evidence leads to ambiguity-asymmetries.  This Bayesian theory of confirmation bias predicts (1) that people should be able to predict their own persuasion, (2) that searching for evidence induces ambiguity asymmetries, and (3) that asymmetric increases in accuracy should drive polarization in both (3a) simple search tasks and (3b) real-world argumentation.  I report on experiments that confirm each prediction.