Algorithms and Parasitic Content on Social Media

Last registered on January 05, 2026

Pre-Trial

Trial Information

General Information

Title
Algorithms and Parasitic Content on Social Media
RCT ID
AEARCTR-0017457
Initial registration date
December 22, 2025

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
January 05, 2026, 7:13 AM EST

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

Locations

Region
Region

Primary Investigator

Affiliation
MIT

Other Primary Investigator(s)

PI Affiliation
Columbia University

Additional Trial Information

Status
In development
Start date
2025-12-22
End date
2026-06-30
Secondary IDs
Prior work
This trial is based on or builds upon one or more prior RCTs.
Abstract
We conduct a lab experiment in which we study how social media users interact with social media content and how it in turn affects their well-being.
External Link(s)

Registration Citation

Citation
Solheim, Hannah and Nancy Wang. 2026. "Algorithms and Parasitic Content on Social Media." AEA RCT Registry. January 05. https://doi.org/10.1257/rct.17457-1.0
Experimental Details

Interventions

Intervention(s)
Intervention (Hidden)
This lab experiment is a follow-up to a previous field experiment Social Media, Body Image, and Economic Decisions (AEARCTR-0014958). In the field experiment, we find that social media algorithms suggest more weight-loss content to users who demonstrate an interest in body positivity. In this lab experiment, we will quantify the welfare effects of this policy. To do so, we will elicit stated preferences for types of social media content, after which we will randomize participants into different short simulated social media feeds that show different types of content. We will observe their watch behavior in this simulated social media feed and compare it to their stated preferences. Finally, we will assess how exposure to different social media feeds affects well-being.

First, we will survey participants about their social media use and preferences. After several screening questions, we will elicit willingness to pay to exclude certain types of content (e.g., weight loss, body positive, cryptocurrency/video games, advertisements). We will incentivize by implementing a small fraction of people’s choices in the next section (below), in which participants are asked to watch a stream of social media videos (skipping or replaying content as they like). We are most interested in their preferences for weight loss content and body positive content. The other genres may serve as placebos.

After, we will randomize participants into the following treatment arms:
(1) Mixed videos – a video stream including several genres (neutral, body positive, weight-loss, and other) (other could for example be video games, cryptocurrency)
(2) Body positive videos – a video stream including neutral and body positive videos (neutral + body positive)
(3) Weight-loss videos – a video stream including neutral and weight-loss videos (neutral + weight-loss)
(4) Control – a video stream made of all neutral videos
Videos are all taken from pools of TikTok videos of each genre. Videos are randomly drawn from the pool of videos in each genre. For example, if someone is assigned to treatment 2, they would be provided a video stream of 40 videos, some of which are randomly drawn from a pool of neutral videos (taken from the general TikTok For You Page) and some of which are randomly drawn from a pool body positive videos shown in the field experiment. The order of the videos is randomized. Once assigned to a treatment arm, the participant is asked to watch (skipping, replaying as they please) for 10 minutes. If they complete the stream of 40 videos early, they will continue to the next part of the survey early. During this section, we are interested in whether participants are able to avoid content they said they would prefer to exclude. For example, can people who want to avoid body positive content skip past it? Weight-loss content? We will use watch time for each video to estimate this.

Afterwards, we will ask several open-ended questions to understand effects on well-being. We will also use measures of body image and mood from the psychology literature. We will also ask several questions to understand how participants interpreted the content given to them. These questions are meant to understand how different genres of video content affect well-being.

Finally, we will offer participants an opportunity to debrief after the study. They can choose a category of videos (and with certainty, they will receive a short stream of videos from that category) to improve their mood.
Intervention Start Date
2025-12-22
Intervention End Date
2026-06-30

Primary Outcomes

Primary Outcomes (end points)
Primary outcomes include incentivized stated preferences for different types of content, watch behavior, and measures of well-being (i.e., body appreciation, open-ended text responses, mood).

More detail is provided in the attached document.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary outcomes include first stage measures of the feed (e.g., if the participants thought the feed was uplifting, what the participants report seeing).

More detail is provided in the attached document.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We will have three treatment arms and one control arm with randomization at the individual level. This is a one-shot experiment.
Experimental Design Details
Participants will be recruited via Prolific, an online platform that allows us to filter for participants who are young (ages 18-26) female social media users. More details are provided in the Intervention section.
Randomization Method
Randomization done in office by a computer
Randomization Unit
Individual. Furthermore, once randomized into a treatment arm, the individual will receive videos that are randomly drawn from a distribution of videos, where the distribution is determined by the treatment arm.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
No clusters
Sample size: planned number of observations
At least 1000
Sample size (or number of clusters) by treatment arms
We plan to randomize at least 1000 participants in fixed proportions: 55% in Treatment 1, 15% in Treatment 2, 15% in Treatment 3, and 15% in Control. We may change the sample size after hearing results from grant applications.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Given our current design and assuming 85% take-up, we calculate the following MDEs, reported in standard errors, for our continuous main outcomes. We plan for our main analyses to be between treatments 2 or 3 vs. control, or behavior in treatment 1 by baseline characteristics (e.g., stated preferences). • Treatments 2 or 3 vs. Control: 0.2 SD • Treatment 1 vs. Stated Preferences (assuming a 50/50 split): 0.26 SD
IRB

Institutional Review Boards (IRBs)

IRB Name
MIT Committee on the Use of Humans as Experimental Subjects
IRB Approval Date
2025-12-18
IRB Approval Number
E-7374
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials