---

Presenter: Taoran Jin Ye

Faculty Sponsor: Andrew Cohen

School: UMass Amherst

Research Area: Psychology and Behavioral Sciences

Session: Poster Session 6, 4:15 PM - 5:00 PM, Concourse, B11

ABSTRACT

Belief revision, the process by which existing beliefs are updated to account for new information, plays an integral role in learning, reasoning, and decision-making. Rational belief revision can be modelled using Bayes’ Theorem, but it is unclear whether human belief revision is rational. Previous research has quantitatively measured belief revision, focused on explicit Bayesian reasoning, and highlighted the importance of causal explanations; however, the question of whether human belief revision is Bayesian, and how causality influences the process, has not yet been addressed experimentally. Employing the belief update task from Sharot et al. (2011), the current study investigates whether individuals revise their beliefs rationally, considering the confidence of their prior beliefs, and whether the process can be enhanced by providing causal explanations for new evidence. Participants were presented with neutrally worded, psychologically relevant reasoning questions and responded with a set of numerical estimates (best, minimum, maximum) both before and after numerical feedback was given. This feedback was accompanied by a causal explanation in the experimental condition and by a related non-causal fact in the control condition. If belief revision is Bayesian, in the case of low pre-feedback confidence (wide minimum-maximum ranges), participants will rely more on feedback during belief revision, resulting in more updating, and in the case of high pre-feedback confidence (narrow minimum-maximum ranges), participants will rely less on feedback, resulting in less updating. Furthermore, participants who receive causal explanations should revise their beliefs more than participants who do not.