Fake news: Overprecision and detection.

How to cite

RIAMBAU, Guillem; Bosch-Rosa, Ciril; Montolio, Daniel. 2025. "Fake news: Overprecision and detection.." AEA RCT Registry. AEARCTR-0014786. https://doi.org/10.1257/rct.14786-1.0

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

Guillem RIAMBAU* , Ciril Bosch-Rosa , Daniel Montolio

* Primary investigator

Registry ID
AEARCTR-0014786
DOI
https://doi.org/10.1257/rct.14786-1.0
First registered
Last updated
Status
In development
Trial dates
2024-11-17 – 2025-03-31
Countries
Private
JEL codes
C91, D83, P16
Registry record
http://www.socialscienceregistry.org/trials/14786

Abstract

This study investigates how individuals identify fake news in four European Union countries: Germany, Ireland, Poland, and Spain. The study consists of a two-wave survey, with all participants invited to participate in both waves. In the first wave, a representative sample of each country will be presented with 13 news headlines and asked to classify each as either true or fake. Participants are then randomly assigned to either the treatment group (two-thirds of the participants) or the control group (one-third). Participants in the treatment group receive feedback indicating whether they correctly classified more or less than half of the headlines, while those in the control group receive no feedback. All participants are then asked to classify a new set of 13 news headlines. Three weeks later, during the second wave, all participants are re-contacted and asked to classify another set of 13 news headlines as either true or fake. The primary goal of the intervention is to assess how feedback influences participants' ability to identify fake news. Additionally, the study aims to examine: (i) the long-term effects of the intervention, measured in the second wave conducted three weeks after the first; and (ii) the role of overprecision in participants’ ability to distinguish fake news.

Intervention

The intervention consists on providing information to participants on how well they perform on assessing the veracity of different news headlines.

Experimental design

The survey will be carried out simultaneously across all four countries. We will have a representative sample of 1,200 respondents in Spain, Germany, and Poland. For Ireland, we will collect 800. All participants who finish wave 1 will be re-contacted for wave 2 exactly 3 weeks after completing wave 1.

Allocation to treatment or control is random. In treatment, participants will be told whether they correctly classified or not more than half of the news in control there is no information on performance.

Design details

Randomization method
Randomization will be done by the company carrying the survey, YOUGOV (link here), by means of a computational algorithm.
Randomization unit
Randomization at the individual level within each country.
Clusters
There are 4 countries, but no clusters.
Observations
6,000 (1,600 per country with 1,200 in Ireland) for the first wave
Arms
6,000 (1,600 per country with 1,200 in Ireland) for the first wave. These are divided 2/3 into treatment (positive or negative signal) and 1/3 control.
Primary outcomes
1. Percentage of correctly assessed headlines 2. Confidence level for each headline (i.e., the probability assigned that a headline is true) 3. Overprecision measure 4. Whether the headline would be shared in social media All pre and post information shock and across waves.

Cited by

No citation data available for this registration yet.

Keywords