Author(s) Gordon Kraft-Todd (Boston College) - gordon.kraft-todd@bc.edu Max Kleiman-Weiner (Massachusetts Institute of Technology) - maxkw@mit.edu Liane Young (Boston College) - liane.young@bc.edu
Pre-registered on 2021/11/12 - 07:13 AM (PT)
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? Here, we are exploring how virtuous (generous vs. impartial) actors' moral goodness and trait generosity are affected by the observability of the actors' behavior. Specifically, we predict:
• H1) Public displays of virtue will be perceived as worse than private displays of virtue (on both moral goodness and trait ratings dependent measures; i.e. virtue discounting).
• H2) Virtue discounting of generosity will be greater than virtue discounting of impartiality (i.e. differential virtue discounting).
• H3) Observers will infer more selfish motivations for public displays of generosity compared to public displays of impartiality.
Note: Though our main question regards whether these hypotheses are borne out in our manipulation employing a set of example virtuous behaviors, we also hypothesize these in our manipulation employing virtue labels (i.e. generous vs. impartial) as stimuli.
3) Describe the key dependent variable(s) specifying how they will be measured. We have two primary dependent measures: moral goodness ("How morally good is [your friend]?" measured with a 100-point unmarked slider scale with anchors at: 0 "extremely morally bad"; 50 "neither morally bad nor morally good"; and 100 "extremely morally good"), and trait ratings ("How [generous/impartial] is [your friend]?" measured with a 100-point unmarked slider scale with anchors at 0 "not at all" and 100 "very much"). We have six secondary dependent measures which will be answered on 100-point unmarked slider scales with anchors at 0 "not at all" and 100 "very much". Two assess the perceived benefit participants perceived to various parties relative to the actor: 1st-party benefit ("How much do you think [your friend] will personally benefit from behaving this way?"); and 3rd-pary benefit ("How much do you think another person would benefit from interacting with [your friend]?"). Four assess participants' perceived motivations for the actor's behaviors ("How much do you think [your friend] is motivated to act [generously/impartially]…"): reputational ("…because she is trying to make others think she is [generous/impartial]?"); authentic ("…because she wants to be [generous/impartial]?"); norm-signaling ("…because she wants others to be [generous/impartial], and she is trying to lead by example?"); and moral ("…because she thinks it is the right thing to do?").
4) How many and which conditions will participants be assigned to? We will randomly assign participants to one of 8 between-subjects conditions in a 2 (stimuli: set of example virtuous behaviors vs. virtue labels) x 2 (virtue: generosity vs. impartiality) x 2 (observability: public vs. private) factorial design.
5) Specify exactly which analyses you will conduct to examine the main question/hypothesis. For H1 and H2, we will conduct multivariate regressions. A significant main effect of the observability manipulation would provide support for H1. If we observe a significant interaction of the virtue and observability manipulations, we will test for H2 by comparing coefficients of the observability manipulation for generosity compared to impartiality by computing pairwise comparisons of estimated marginal cell means corrected for multiple comparisons using Scheffe's adjustment. Observing a coefficient on observability for generosity that is significantly larger for than impartiality would provide support for H2. For H3, we will construct multiple moderated mediation models using structural equation modeling with standardized variables, and indirect effects will be calculated using percentile bootstrapped standard errors with 1,000 replications. Depending on the outcome of a factor analysis of the six secondary dependent measures with varimax rotation and iterated principal factors, if a sufficiently small (<=3) number of factors explain a sufficiently large (>=70%) proportion of variance, we will enter these factor scores as mediators of our observability manipulation on both of our primary dependent measures for each virtue that has a significant simple effect of observability in the multivariate regressions described above. We may also construct a second set of models (if this will help clarify our results from our initial mediation models), entering all six secondary dependent measures separately as mediators of our observability manipulation on both of our primary dependent measures for each virtue that has a significant simple effect of observability in the multivariate regressions described above.
6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations. We will exclude observations that fail basic attention checks (e.g. "What is your favorite color? [Please select blue]"). As a robustness check, we will test to ensure the results of the tests reported above hold in the subpopulation of our sample who correctly answer comprehension questions.
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. For power calculations, we employed the Superpower package in R software (Lakens & Caldwell, 2021) based on pilot data employing the design described here. With a desired effect size of d=.22 (for the virtue*observability interaction), we determined a sample size of N=250 per cell was required to achieve power at 93% with an alpha level of .05 (total N=2000).
8) Anything else you would like to pre-register? (e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?) We will also collect basic demographics (gender, age, race, income, education, and political affiliation).
