Fig. 3. Social value scores predict choice preferences for specific social partners in one-shot social decision tasks with monetary and social outcomes.

Analyses involving this set of outcome variables were conducted by regressing each outcome onto the relevant pair of social value scores. For simplicity in visualization, figures in the top two rows plot the relationship between difference scores between social value score pairs and each outcome. ‘Forced Choice Spend Time’ refers to a one-shot question asking participants to choose one of two social partners with whom they would rather spend a free afternoon. Data for this visualization were drawn from confirmatory phase of the study (prolific sample, SONA-sourced weights). A greater value on the y-axis of the plots in the top two rows indicates greater preference for the first social partner listed in the axis label. Axis labels reflect differences between social value scores. The bottom two rows of plots depict the posterior distributions (the likelihood of observing a given coefficient value given the data) for each slope coefficient from the full multiple regression model. Posterior distributions were obtained via MCMC sampling. The shaded portion of the posterior distributions represent the 89% highest density credible interval (HDI). A hashed vertical line is shown over zero (null value); a solid vertical line within each posterior indicates the posterior mean. Evidence was judged to be robust if the HDI did not include 0 or the HDI fell outside of the Region of Practical Equivalence (ROPE) and moderate if part of the HDI fell outside of ROPE (see “Inferential Criteria” section of the main text). ROPE was defined as the range between -0.1 and 0.1. Here, there was robust evidence that social value scores predict choice preferences involving monetary outcomes for all social partners. There was also robust evidence that social value scores predict choice preferences involving social outcomes when pitting friends or acquaintances against parents. There was moderate evidence that social value scores predict choice preferences involving social outcomes when pitting friends against acquaintances. A logistic-link function was used for analyses with the binary forced choice item about spending time with a social partner. ‘Acq’ refers to acquaintance, ‘Fri’ refers to friend, ‘Par’ refers to parent, and ‘Val’ refers to the computed social value score. The sample size for this analysis is N = 233.