Research

Experiment on Narratives and Information in Persuasion

With Alisher Batmanov

[draft coming soon!]

We experimentally study how the elements of cheap-talk communication and narrative persuasion interact in the same setting. In an inference task, receivers observe a dataset and make predictions after receiving a message from a sender who may have additional information and who communicates either through a simple recommendation or a narrative explanation. By varying the sender's information and communication mode, we isolate the effects of new information versus new interpretation. We examine how the effects of these message types differ based on whether the sender and receiver have aligned or misaligned preferences, allowing us to test predictions from models of strategic information transmission and to assess when persuasion works through informational advantage versus framing. We find that receivers' guesses land roughly midway between the true grade and the sender's recommendation. Both informational advantage and narratives increase persuasive power, though information plays a much stronger role: a sender with an informational advantage moves guesses approximately 39% closer to their recommendation, while adding a narrative under equal information yields a smaller, statistically insignificant effect of approximately 9%. This asymmetry holds across all levels of task complexity — varying the number of variables and conditional versus unconditional DGP structure. When preferences are aligned, both channels move guesses more strongly toward the recommendation, but information continues to matter more.