Debiasing Motivated Reasoning: Unpacking Mechanisms

How to cite

Thaler, Michael; Ngangoue, Kathleen; Sticher, Vanessa. 2023. "Debiasing Motivated Reasoning: Unpacking Mechanisms." AEA RCT Registry. AEARCTR-0012638. https://doi.org/10.1257/rct.12638-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Michael Thaler* , Kathleen Ngangoue , Vanessa Sticher

* Primary investigator

Registry ID
AEARCTR-0012638
DOI
https://doi.org/10.1257/rct.12638-1.0
First registered
Last updated
Status
In development
Trial dates
2023-12-13 – 2024-12-13
Countries
Private
JEL codes
C91, D83, D91
Registry record
http://www.socialscienceregistry.org/trials/12638

Abstract

Motivated reasoning, the tendency to selectively interpret information in ways that support beliefs people would rather hold, is a pervasive phenomenon that has been shown to contribute to belief biases, political polarization and excessive trust in ``fake news.'' In this project, we will experimentally study whether feedback interventions can induce people to become less susceptible to motivated reasoning and improve the accuracy of how they update beliefs. In particular, we test the effects of giving people two forms of feedback that increase their awareness when they incorrectly assess the truthfulness of a piece of news. Treatment A tells them about the costs of their inaccurate assessment, and Treatment B has people introspect about whether the bias of motivated reasoning impacted their assessment. We will test whether these interventions lead people to change how they process information about news on subsequent political issues. We will analyze whether they induce people to become less biased, more accurate, and less polarized about how they process future pieces of news.

Experimental design

(See attached analysis plan)

Design details

Randomization method
Randomization for all treatments done by computer.
Randomization unit
Pro-Party / Anti-Party news randomized at question level. Control / treatments randomized at individual level.
Clusters
Target: 1800 participants
Observations
Approximately 20,000 political news assessments
Arms
Target: 600 participants control, 600 participants in treatment A, 600 participants in treatment B
Primary outcomes
(See attached analysis plan)

Cited by

No citation data available for this registration yet.

Keywords