Understanding online influence: Experimental evidence from the immigration debate on social media
How to cite
Samahita, Margaret; Conzo, Pierluigi; Gallice, Andrea; Morales, Juan; Taylor, Laura. 2021. "Understanding online influence: Experimental evidence from the immigration debate on social media." AEA RCT Registry. AEARCTR-0008381. https://doi.org/10.1257/rct.8381-1.6000000000000005
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
This project proposes to study the effect of social media on individuals' political attitudes through an online survey experiment. Our experiment is designed to disentangle some of the precise mechanisms through which online interactions lead to political influence. Our focus is on the most common feature of social media interactions: likes. Can the number of likes we see on social media messages affect its influence on us? Does liking a post make us more persuaded by the post's content? Does the salience of other participants' engagement matter? Does the fact that our likes are shown to others matter?
To answer these questions, we will conduct an online survey experiment in the context of the migration debate in the EU. We will recruit participants from four EU countries, Ireland, Italy, Hungary and France, which vary widely in their attitudes towards immigration.
Intervention
Subjects are shown 6 messages, formatted to look like social media posts, one at a time. Posts 2, 4, and 6 are randomly selected from tweets by EU members of parliament (with minor rewording for clarity) that have been pre-tested on another set of participants in a different study. Based on this pre-test we order posts from most pro- to most anti-immigration. The number of likes for each text will be randomised from two uniform distributions: low [1,50] or high [151,200]. For posts 1, 3 and 5, subjects are shown placebo texts.
Subjects are randomised into six treatments that vary in two dimensions. First, subjects are either told that the posts "have been liked by previous participants" or not. Second, we vary the ability to like the posts: in one condition subjects are not given the possibility to like, in another subjects can like privately, and in a third condition subjects can like publicly---thereby disentangling the effect of liking per se and the visibility of the like.
Experimental design
We first measure participants' attitudes towards immigration using a random subset of questions from the ESS.
We then randomise them into 1 of the 6 treatments described above, where they are shown 6 posts (3 on immigration, 3 placebos) with varying number of likes.
We then measure their post-treatment attitudes using the remaining ESS questions, elicit their willingness to donate to the pro-immigration charity and their willingness to sign a pro-immigration petition.
We finish with an attention check (asking participants if they remember how many likes they observed on 2 of the posts) and their trust in a randomly selected post.
Design details
Cited by
No citation data available for this registration yet.