Social technologies for health have already become essential means for providing underserved populations greater social connectedness and increased access to novel health information. However, these technologies have also had negative unintended consequences. The resulting digital divide in social technology takes many forms - from explicit racism that excludes African American and Latinx populations from the resources enjoyed by White and Asian members of online communities, to self-segregation for the purposes of identity preservation and community-building that unintentionally results in limited informational diversity in underserved communities. The result is an often unnoticed, but highly consequential compounding of inequities.

This research seeks to use an online social network approach to address these challenges, in which the investigators demonstrate how reducing the online levels of network centralization and network homophily among African American community members directly increases their productive engagement with health-promoting information.

To investigate the causal effects of network structure and composition on the acceptance of new or unfamiliar behavior-relevant health information, the investigators propose a randomized controlled experiment that compares several independent populations to identify and address participants' endorsement of biased information, and engagement with novel behavior relevant information (e.g., regarding COVID-19 vaccination). Each population will have its own network structure (i.e., level of centralization) and composition (i.e., level of homophily).

To run each experimental trial, the investigators will recruit 240 African American participants, aged 18 to 40, collectively to answer behavior-relevant questions over a period of no greater than 8 minutes. Participants can respond asynchronously - i.e., when the participants' time permits. As with previous studies, the technical infrastructure will manage participants' progress through the study to ensure that all participants have the relevant information about each other's responses.

To ensure causal identification, each network graph will constitute a single observation of how individual decisions change under conditions of interdependent social information. Thus, each trial of 240 people (6 networks x 40 participants per network) produces 6 observations of a community-level social learning process. Power calculations indicate that 8 independent trials are sufficient to produce results of p<0.05 with 85% power, resulting in a desired population of 1920 participants for each health topic (e.g., COVID-19 vaccination is a single "health topic"), producing 48 independent observations of collective decision making per health topic.

The studies will target health topics for which there is substantial racial disparity in outcomes and behavior, such as acceptance of COVID-19 vaccination, and spreading of various categories of COVID-19 misinformation (e.g. beliefs related to assessment of personal risk, effectiveness of protective behaviors, methods of transmission, disease prevention, treatment, origins of the virus) and related health practices (e.g. choice of appropriate contraceptive methods, value of heart disease screenings, etc.).

Inclusion Criteria:

• Having internet access
• Aged 18 and above
• Living in the United States

Exclusion Criteria:

• Having no internet access
• Aged below 18
• Living outside of the United States

We will share the data collected from our online experiments. All sets of data are anonymous.

Data will be available when the primary intervention paper is published.

Data will be shared as a part of the published paper, in forms of supplementary materials. The public can access the data through the publisher's website.