Social Media, Well-being, and Political Attitudes: The Causal Effects of Recommender Algorithms

Last registered on June 06, 2023

View Trial History

Pre-Trial

Trial Information

General Information

Title

Social Media, Well-being, and Political Attitudes: The Causal Effects of Recommender Algorithms

RCT ID

AEARCTR-0011464

Initial registration date

May 22, 2023

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

June 06, 2023, 9:09 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Country

Switzerland

Region

University of St.Gallen

Primary Investigator

Name

Roland Hodler

Affiliation

University of St.Gallen

Contact Primary Investigator

Other Primary Investigator(s)

Additional Trial Information

Status

In development

Start date

2023-05-22

End date

2023-08-31

Keywords

Behavior, Electoral

Additional Keywords

JEL code(s)

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

We study the causal effects of social media recommender algorithms on well-being and political opinions, specifically on Twitter.

External Link(s)

Registration Citation

Citation

Hodler, Roland. 2023. "Social Media, Well-being, and Political Attitudes: The Causal Effects of Recommender Algorithms ." AEA RCT Registry. June 06. https://doi.org/10.1257/rct.11464-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Intervention Start Date

2023-05-22

Intervention End Date

2023-08-31

Primary Outcomes

Primary Outcomes (end points)

The subjects' Twitter activity, well-being, and political attitudes (regarding political orientation and current events, e.g., the current situation in Ukraine). Variables are constructed based on our pre- and post-treatment surveys and the subjects' publicly accessible Twitter activities.

Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes (explanation)

Experimental Design

Each subject is randomly assigned to one of two groups at the end of the pre-treatment survey. Subjects in the treatment group are asked to use Twitter's algorithmic newsfeed setting; those in the control group to use the chronological newsfeed setting. All subjects are asked to use only this one newsfeed setting for six weeks. We then conduct our post-treatment survey.

Experimental Design Details

Each subject is randomly assigned to one of two groups at the end of a pre-treatment survey. Subjects in the treatment group are asked to use Twitter's algorithmic newsfeed setting; those in the control group to use the chronological newsfeed setting. All subjects are asked to use only this one newsfeed setting for six weeks. We then conduct our post-treatment survey.
In addition, we let some subjects run a Google Chrome extension (on their own computers) to collect what they would see under the two different newsfeed settings. Moreover, we collect information on all the subjects' Twitter activity (tweets, retweets, their profile description, liked tweets, who they follow, and who follows them). The data on their Twitter activity are publicly available; we can access them since the subjects indicate their Twitter username in the survey (for this purpose).

Randomization Method

Randomization done by a computer program in Swiss student sample. Stratified randomization done by survey company in US population sample.

Randomization Unit

Individuals

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

No clustering

Sample size: planned number of observations

For the Swiss student sample, we invite all students of a large introductory economics lecture to participate in our survey experiment. We expect around 100 students to participate. For the US population sample, we target around 6,000 subjects who are U.S. residents, Twitter users, and at least 18 years old.

Sample size (or number of clusters) by treatment arms

For the Swiss student sample, we expect around 50 students per treatment arm.
For the US population sample, we expect around 3,000 subjects per treatment arm.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Supporting Documents and Materials

IRB