As Predicted:Bridges to Counter Polarization (#34635)
Created: 01/28/2020 01:32 PM (PT)
This is an anonymized version of the pre-registration. It was created by the author(s) to use during peer-review.
A non-anonymized version (containing author names) should be made available by the authors when the work it supports is made public.
1) Have any data been collected for this study already?
No, no data have been collected for this study yet.2) What's the main question being asked or hypothesis being tested in this study?
Q1. Does positive social influence induced by closeness and non-political similarity with others facilitate convergence of opinions on politicized and polarized topics, such as income inequality and redistribution, climate change, mass migration, etc?
This preregistration is for the politicized topic of “government redistribution to reduce socio-economic inequalities” (in short: redistribution).
3) Describe the key dependent variable(s) specifying how they will be measured.
Stance on the politicized topic before and after interacting with a similar / dissimilar other on the political and personal dimensions. The stance is the sum (average) of the answers to a set of 7-point Likert scale questions focusing on the politicized topic.
The stance for redistribution is measured with 9 questions: one general question about the involvement of the government, and 8 specific questions about policies for health care, public education, foodstamps, taxes on billionaires, estate tax, aid to the poor, minimal wage, and public housing.
Participants’ stance S is labeled as:
- In Favor: if S > 0,
- Against: if S <= 0.
For the redistribution topic, the stance ranges from -27 to + 27; we define 4 sub-categories:
- Strongly Against: if S <= -12
- Mildly Against: if -12 < S <= 0
- Mildly in Favor: if 0 < S <= 12
- Strongly in Favor: if S > 12
Based on participants’ stance types, the following 4 match categories are considered for analysis.
- Against -> Against: 2 “Against” participants;
- In Favor - > In Favor: 2 “In Favor participants;
- Against -> In Favor: 1 “Against” participant matched with an “In Favor” partner;
- In Favor -> Against: 1 “In Favor” participant matched with an “Against” partner.
Similarity on views about the politicized topics is measured as the linear distance of the stances. For the matching,
Non-political similarity between is measured as the weighted sum of all common answers to non-political questions. In this compound score, questions are weighted based on how common a match in such question is plausibly expected to be, e.g., eye color is rated lower than favorite actor; hazel eye color is weighted higher than brown eye color, etc. Open-ended questions are matched with fuzzy logic. Location and Zip/Postal code are not used to compute the similarity score for privacy reasons. Full details about the weights are available in the anonymous timestamped repository: https://osf.io/7ghnj/?view_only=4f59af9f5bfb4e29aff2262dfa8aa66d
Closeness Expected and Perceived:
Closeness to the matched partner before (expected) and after (experienced) the interaction are measured with a single question on a 7-point Likert scale.
4) How many and which conditions will participants be assigned to?
MAIN SURVEY: Participants will complete a survey with questions about:
A. Non-political interests.
B. Political interests.
TEST: survey about the politicized topic (will determine the stance).
INTERVENTION: Participants will be assigned to 1 in 4 conditions, determining their interaction partner:
i) partner with similar views on the politicized topics and dissimilar personal characteristics;
ii) partner with similar views on the politicized topics and similar personal characteristics;
iii) partner with dissimilar views on the politicized topics and similar personal characteristics;
iv) partner with dissimilar views on the politicized topics and dissimilar personal characteristics.
A match score M > 600 is labeled of high-similarity and M < 300 is labeled of low-similarity (dissimilar). In case these constraints cannot be fulfilled, the next best match is taken.
All interventions include consists in two parts, in which participants will:
- PROFILE: read a profile page of their partner showing non-political commonalities as well as their rank similarity score (relative to other potential matches in the database),
- INTERACTION: read a short essay composed by their partner on the politicized topic.
No partner will be matched with more than 15 participants.
RETEST: same survey questions as in TEST.
In addition, participants will answer:
- Two questions to predict (after the PROFILE) and evaluate (after the RETEST) how much of a connection they will feel / felt with their partner.
- Two questions after the RETEST about what they thought the issue position and political leaning of the participant actually was.
- One question after the RETEST to state if they believed their match was a real vs fabricated participant.
Additional details on the MAIN SURVEY can be found at the anonymous timestamped repository: https://osf.io/7ghnj/?view_only=4f59af9f5bfb4e29aff2262dfa8aa66d
5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
We will run multilevel regressions with 1 level as the essay id predicting the stance update before and after the interaction with the following predictors (each one on a separate regression):
- Match (similarity) score
- Expected closeness
- Experienced closeness
- Match stance distance
We will repeat the analysis above with:
- the following controls: Gender, Income, Education, Social Media Usage, Dummy Matched Partner Perceived as Real Person, Party Affiliation, Initial Stance, Match Type.
- the interaction Match stance distance with: Match Score, Expected Closeness, and Experienced Closeness (each one on separate regression).
- the interaction of own stance with: Expected Closeness and Match Score (each on separate regression).
- the interaction Match Type with: Match Score, and Expected Closeness (with and without fixed terms; each one on separate regression).
The level of significance for all tests is 0.05.
6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
We will exclude any participant who fulfills any of the following:
- complete open-ended questions with non-sensical text;
- voluntarily declare that they did not pay attention to the questions;
- perform straight-lining on at least 2 blocks of the main survey;
- complete the interaction part of the survey too quickly, that is: spend less than 10 seconds reading the profile of the match or less than 15 during the interaction itself.
7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined.
We aimed to collect a total of 1250 observations.8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)
We have collected 158 short essays about the topic of redistribution to use as intervention.
In addition to Q1, we pre-register the following questions:
Q2. Do similarity scores on non-political dimensions predict expected and experienced closeness? Is the association stronger with expected or experienced closeness?
Q3. Does experienced closeness differ from expected closeness after interaction? If so, does holding opposite stances on the politicized topic increase or decrease the difference?
Q4. Do similarity and closeness affect participants with stronger vs weaker stances on the politicized topics differently?
Q5. Do participants who are strongly against a given politicized issue behave different than the rest of other participants with respect to closeness? Do they increase or reduce their support for redistribution when matched with somebody with the same views on the issue, but low closeness?