The diffusion of online misinformation in India

How to cite

Narang, Jimmy. 2022. "The diffusion of online misinformation in India." AEA RCT Registry. AEARCTR-0006474. https://doi.org/10.1257/rct.6474-1.0

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

Jimmy Narang*

* Primary investigator

Registry ID
AEARCTR-0006474
DOI
https://doi.org/10.1257/rct.6474-1.0
First registered
Last updated
Status
In development
Trial dates
2021-12-01 – 2022-12-01
Countries
Private
Secondary IDs
JPAL CVI Grant GR-1192, Weiss Fund Berkeley’s Agreement No. 048745
Registry record
http://www.socialscienceregistry.org/trials/6474

Abstract

We conduct a set of lab-in-field experiments in India to understand how misinformation can spread through peer-to-peer messaging. We examine how sharers make sharing choices based on a story's (perceived) accuracy and the cost/benefit of sharing it; and receivers form an opinion about the story based on its content and their assessment of the sharers' discernment & motivations.

The experiments also provide estimates of Indian citizens' ability to discern misinformation of various topics (health, politics & identity, finance), and the types of stories they like to share. (N = 2000 participants expected)

Intervention

(No intervention)

Experimental design

We recruit participants in pairs, such that members of a pair know each other in real life. One partner in each pair is (randomly) assigned to be the "sharer", and the other, the "receiver". Participation is sequential: receivers begin when sharers finish their part of the study.

During the study, the sharer scrolls through a series of news stories -- formatted as they would appear in a Facebook newsfeed or a WhatsApp group -- and chooses to share as many (or as few) of these stories with their partner as they like. (There is no special incentive to do so: sharers are asked to match their real life behavior as much as possible). Additionally, sharers are asked to report how likely they think each story is true, whether they have seen it outside the lab, and a couple of other (randomized) questions.

Receivers also see a series of news stories on their screens, but do not have the option to share. Instead, they report how likely they think each story is true. Then, they are shown a "signal" about each story. The signal can be: (i) a prompt informing them if the story was shared by their partner (the sharer); (ii) the sharer's belief in the story (but not the sharing choice); (iii) a computer generated clue about whether the story is true or not. In each case, receivers report their posteriors after seeing the signal.
Both sharers and receivers are also asked for beliefs _about_ their partner's discerment or reasoning. For instance, receivers may be asked to guess how likely the _sharer_ thought the story was true. Sharers may be asked to guess if the receiver has seen the story before.

At the end of the study, participants are shown which stories were true and which were false, along with links to the actual news story or a third-party fact-checking website debunking the claim.

1-2 weeks later, participants answer an (optional) follow-up questionnaire that measures their recall of which stories were true (or not), and Provide some feedback on their experience of the study.

Design details

Randomization method
Randomization of participants into groups is done by a computer/program Randomization of the sequence of news-stories and questions is also programmatic.
Randomization unit
Individual.
Clusters
NA
Observations
40,000 (20 observations x 2000 participants)
Arms
2000
Primary outcomes
- beliefs in the accuracy of various news-stories / WhatsApp forwards, elicited as one of: percentages, likert scale, binary - decisions to share these stories with fellow participants (binary) - Beliefs about discernment and motivation: participants' beliefs about how good their partners are at guessing if a story is true, or how likely they are to share a specific type of story in real life. - Recall of the accuracy of stories 1-2 weeks after the experiment.
Sponsors
Name: JPAL Crime and Violence Inititative Url: https://www.povertyactionlab.org/initiative/crime-and-violence-initiative Name: Weiss Family Fund Url: https://bfi.uchicago.edu/the-weiss-fund/ Name: CEGA Location: UC Berkeley Url: https://cega.berkeley.edu/

Cited by

No citation data available for this registration yet.

Keywords