Lab-in-the-field experiment in Brazil

Last registered on April 14, 2026

View Trial History

Pre-Trial

Trial Information

General Information

Title

Lab-in-the-field experiment in Brazil

RCT ID

AEARCTR-0018321

Initial registration date

April 09, 2026

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

April 14, 2026, 8:59 AM EDT

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

Locations

Country

Brazil

Region

Primary Investigator

Name

Rachel Berwald

Affiliation

UCLA Political Science Department

Contact Primary Investigator

Other Primary Investigator(s)

Additional Trial Information

Status

In development

Start date

2026-04-09

End date

2026-06-30

Keywords

Education

Additional Keywords

Communication

JEL code(s)

Secondary IDs

Prior work

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

Abstract

Amid a growing influence of misinformation and distrust in traditional news media, questions have been raised about how to improve media literacy education beyond the classroom. Prior studies in media literacy have been effective at improving students’ ability to identify misinformation but have only been conducted in K-12 classrooms (Amar et al., 2025) or have been less rooted in evidence-based pedagogy. Questions remain about how to improve media literacy education, particularly for adults, and to mitigate threats to democracy. This study addresses these challenges by bridging the gap between political science and education research. I argue that evidence-based pedagogical approaches, combined with interactive AI tools, can help individuals recognize biases in their media diets and encourage behavioral change beyond the intervention itself. As individuals become more willing to engage with ideas they may not agree with, they may also become more willing to engage with people they disagree with.

This study tests a series of approaches to teach people how to identify biases in their media diets and understand the value of exposure to cross-cutting media, varying exposure to an AI chatbot and evidence-based teaching methods. Participants will be assigned to one of four conditions that vary in exposure to pedagogy and the use of an AI chatbot.

Primary outcomes include selective exposure and willingness to interact with individuals holding opposing political beliefs, as well as data tracking respondents’ browsing history to see whether the effects measured in the intervention extend to behaviors outside the study. Effects will be measured immediately and at two weeks post-intervention to assess effect duration. I expect that the pedagogy treatment will have the strongest treatment effects, encouraging respondents to vary their media diets and interact with those who hold different political beliefs due to its high levels of interactivity, engagement, and critical thinking.

External Link(s)

Registration Citation

Citation

Berwald, Rachel. 2026. "Lab-in-the-field experiment in Brazil." AEA RCT Registry. April 14. https://doi.org/10.1257/rct.18321-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

In the experiment, respondents will be sorted into one of four groups. (1) The pedagogy group will interact with an AI chatbot that uses evidence-based teaching methods to teach respondents how to identify biases in their media diets. (2) The information treatment group will interact with a chatbot that simply tells them how to identify biases in their media diets. (3) The self-paced lesson treatment will interact with an evidence-based teaching method, but no chatbot, to teach respondents how to identify biases in their media diets. Finally, (4) the control will interact with a distraction task.

Intervention (Hidden)

To understand how AI and pedagogy-informed interventions can help individuals identify biases in their media diets, I will conduct a lab-in-the-field study in Brazil. The main survey will have three components. First, respondents will answer a series of basic questions related to, among other components, their media consumption, partisanship, media literacy, and AI literacy.

Next, respondents will be randomly sorted into one of four groups. In the pedagogy treatment, respondents will interact with a chatbot trained on an evidence-based teaching method called the Gradual Release of Responsibility, which guides users in identifying biases and developing personalized strategies to diversify their media consumption. In contrast, the information treatment delivers similar content via a scripted chatbot, without personalization or encouragement of critical thinking. The final treatment condition is the self-paced lesson condition, which incorporates pedagogy but does not involve a chatbot. Finally, respondents in the control group will engage in a distraction task.

Intervention Start Date

2026-04-13

Intervention End Date

2026-05-22

Primary Outcomes

Primary Outcomes (end points)

Selective exposure

Primary Outcomes (explanation)

Selective exposure: Respondents will select multiple headlines from a list with varying political slants (right, left, moderate).

Secondary Outcomes

Secondary Outcomes (end points)

Learning, acceptance, behavioral changes, mechanisms

Secondary Outcomes (explanation)

Learning: Asking respondents about content they learned about in the intervention
Acceptance: Asking respondents about their willingness to interact with people who have different political beliefs and sentiments toward the in-group vs. the out-group
Behavioral changes: measured using actual web-tracking data
Mechanisms: aspects that facilitate learning such as agency, interactivitiy, and enjoyment

Experimental Design

The intervention will be disseminated through a survey. Respondents will answer pre-treatment questions, interact with the intervention, and then answer post-treatment questions. Approximately two weeks later, respondents will be recontacted for a follow-up survey that measures certain outcomes. In between, nearly all respondents will have their web-browsing and app usage tracked.

Experimental Design Details

To understand how AI and pedagogy-informed interventions can help individuals identify biases in their media diets, I will conduct a lab-in-the-field study in Brazil. The main survey will have three components. First, respondents will answer a series of basic questions related to, among other components, their media consumption, partisanship, media literacy, and AI literacy.

Next, respondents will be randomly sorted into one of four groups. This will be the intervention.

After completing the activity, respondents will answer a series of outcome questions. These include, among other measures, selective exposure, learning, behavioral outcomes, acceptance, and mechanisms. Nearly all of the respondents will also have their browsing history recorded (through Netquest).

Approximately two weeks later, respondents will be recontacted to measure the long-term effects of some outcomes, including selective exposure, acceptance, and mechanisms.

Randomization Method

Simple random assignment with roughly 25% of respondents in each treatment condition. Randomization will be done via Qualtrics.

Randomization Unit

Individual

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

Around 3400 individuals in total will take the survey.

Sample size: planned number of observations

Respondents will be recruited through Netquest. Around 3400 individuals will take the main survey with a goal of having 3000 respondents in the recontact. All respondents will be in Brazil. Approximately 3200 of these respondents will be in the behavioral (web-tracking) panel.

Sample size (or number of clusters) by treatment arms

Approximately 25% of the sample will be in each treatment arm (25% in the pedagogy arm, 25% in the information arm, 25% in the self-paced learning arm, and 25% in the control).

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

Supporting Documents and Materials

IRB