Narrative Belief Updating
How to cite
Albertazzi, Andrea; Pin, Paolo; Stimolo, Marco; Stringhi, Alessandro. 2025. "Narrative Belief Updating." AEA RCT Registry. AEARCTR-0015672. https://doi.org/10.1257/rct.15672-4.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
In this project, we examine how narratives influence the incorporation of new information into individuals’ beliefs. Specifically, we will replicate a classic Bayesian problem in a controlled experiment, where individuals face uncertainty and use new information to update their beliefs about the true state of the world. To this end, we will randomly assign participants to one of two conditions. In the baseline treatment, participants will engage with an abstract statistical urn problem: participants must identify from which of two urns, with different compositions, the colored balls are drawn. In the narrative treatment, participants will instead receive a fictional story—framed around identifying a culprit—that mirrors the same statistical problem. The only difference from the baseline is a more contextually rich framing.
By comparing how participants update their beliefs in these two conditions, we aim to assess whether the presence of a narrative systematically alters the way new informative and uninformative evidence is interpreted.
Intervention
Participants will complete a belief‐updating task based on a classic Bayesian inference problem. Each participant is randomly assigned to one of two between‐subjects conditions. In the Control (Abstract) condition, the task is presented as a neutral statistical problem involving two urns with different color compositions. Participants observe sequential draws of colored balls and must update their beliefs about which urn the draws come from. In the Narrative condition, the same statistical structure is embedded in a short fictional story in which participants must infer the identity of a culprit based on successive pieces of evidence. The only difference between treatments lies in the framing: while the underlying probabilistic structure and evidence are identical, the narrative condition provides a contextually rich storyline that may alter how information is interpreted and integrated into beliefs.
The information-generating process is the following: participants will receive a signal in favour of the true state (positive) with a 45% chance, in favour of the incorrect state (negative) with 30%, and will receive nondiagnostic information with a 25% chance.
Experimental design
Original experiment: In both treatments, participants will be presented with an initial "story". In the baseline, the story concerns two urns containing balls of three different colours. In the narrative treatment, the story is about two possible culprits of a series of thefts. Subjects are asked to state their prior belief about which one between the two urns/culprits is the correct one. Then, participants play 10 rounds in which they receive new information every time. The new information can be in favour of one of the two alternatives or uninformative. The ex-ante probabilities of receiving each piece of new information are public knowledge. Subjects are asked to state their posterior belief in every round upon receiving the new information.
Follow-up: The structure of this experiment is the same as the Narrative treatment in the original experiment. The only difference is that instead of receiving a nondiagnostic signal with a probability of 0.25, subjects will receive no information (a blank signal) with the same probability.
Design details
Materials
Cited by
No citation data available for this registration yet.