Computational principles of belief system change
Abstract
People often fail to change their mind in response to evidence and may even become more strongly entrenched in their original position. Although this inflexibility can give rise to political polarization the world over, we have been lacking a formal, generalizable, cognitively sophisticated framework for understanding how differing ideologies can lead people to interpretthe same information in dramatically different ways.In this project, we propose to theoretically and empirically analyze belief system updating through the lens of Bayesian theory, a mathematical framework that has been used to precisely formalize probabilistic causal reasoning across countless domains of cognitive science. Using this framework, we propose to investigate how the persuasive power of evidence depends on the way it is filtered through ones worldview.Our technical approach will be to systematically develop and test Bayesian cognitive models of political ideology from survey data on a number of political issues. We will test whether these models can predict which beliefs will most resist change in the face of the facts and which kinds of information will be most convincing.The proposed work will inform us about why people believe misinformation and disbelieve facts, and if successful, will enable us to better explain the occurrence of polarization. It will also more broadly improve our understanding of how people learn about complex systems, which may help us improve the speed and quality of learning in a range of real-world domains and guide the development of decision support systems that complement human judgment.Approved for Public Release
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 15, 2021
- Source ID
- N000142112170
Entities
People
- Rahul Bhui
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy