Experimental Design Details
We will post a short description of our study on Facebook and invite U.S. college students to participate in our study. We will explicitly state up front that the experiment includes discussion of socioeconomic and political topics, some of which may be sensitive. The main experiment will be conducted online in the following steps
1. Baseline survey: We will recruit 500 participants. First, we will elicit students’ private beliefs on a set of socioeconomic and political topics, some of which are politically sensitive. Students can choose to answer “yes”/“no” and can also provide a one-sentence explanation of their beliefs. Students will also be incentivized to guess how other students privately answer each topic. We will also collect demographic data, including gender, ethnicity, age, major, and self-identified political affiliation, and personally identifiable information, specifically full name, email address, and profile picture. All information collected in the survey will be kept private and confidential and only the two researchers will have access to the personally identifiable information provided in the survey.
2. Randomly split participants into two subsamples of 100 and 400 people respectively. The 100 participants first choose whether to post their opinions publicly on an online message forum created by us on Slack. As admin of the Slack group, we will control display settings, such as hiding participant emails so that they are not publicly displayed and editing participant’s full name and display name so that they are uniform (e.g. all participants will use their actual full name as their user name and display name as well as the provided profile photo in the survey as their profile photo).
3. Public expression (Round 1): the 100 participants will be asked a subset of the socioeconomic and political topics asked in the baseline survey and each respondent can choose to answer publicly “yes”, “no”, or skip the question and can also choose to provide explanations of their beliefs via publicly posted comments within the Slack group. Respondents’ full names and profile pictures will be posted alongside their messages if they choose to publicly express their opinions. Participants will only have 10 minutes to post their comments in the Slack group, to limit actual back-and-forth online conversations between participants.
4. Randomize into information treatments: For the remaining 400 participants, we will randomize participants into one of two information treatments:
a. Control: Provide information that summarizes the public expression of the 100 participants. For example, suppose that among 100 participants, 30 publicly answered “yes,” 20 publicly answered “no,” and 50 remained silent on a certain topic. We will show the control group a summary data visualization (such as a pie chart) that 30 respondents publicly posted “yes” and 20 respondents posted “no” in answer to the question.
b. Treatment: Provide information that draws attention to the fraction of people who remained silent during the group chats. Take the same example where among 100 participants, 30 publicly answered “yes,” 20 publicly answered “no,” and 50 remained silent on a certain topic. For the Silence Treatment group, we will show a summary data visualization that 30 respondents publicly posted “yes,” 20 respondents posted “no,” and 50 respondents remained silent in answer to the question.
5. Follow-up survey: After the information treatment, we will elicit students’ private beliefs on the same set of socioeconomic and political topics from the baseline survey. We will also elicit second-order beliefs by incentivizing accurate guesses about the belief distribution of other participants within their specific groups.
6. Randomization into new discussion groups: Based on the information treatment assignments, we will randomize participants into two groups of all control individuals and all treatment individuals.
7. Public Expression (Round 2): In these two groups with all control or all treatment individuals, students will again be asked a set of socioeconomic and political topics. Participants will make public expression decisions in two rounds. In each round, half of the participants can choose to answer publicly “yes”, “no” or skip the question and provide explanations for their beliefs via publicly posted comments, and the other half will act as “observers”. In this way, each participant expresses their opinions only once and we will have minimum experimenter demand effect.
8. Endline survey (actions): After the public expressions and information treatments, we will ask all respondents whether they are willing to take certain political expression actions, such as signing a petition or donating to a charitable cause.