Fig. 7: Relation of the number of issues participants had to decide about in each task (G, top panels) and the number of other participants they worked with (K, bottom panels), with the social balance (f, column 1) and the proportion of different triangles occurring in the group experiment (columns 2–5).

For each box, the central red line is the median, the edges are the 25th and 75th percentiles, the whiskers extend to the most extreme data points within another box length on each side, and the outliers are plotted individually as dots. Green Xs denote the mean predictions of the triadic model for q = 0.01, while green circles denote the mean predictions of the dyadic model for \(\alpha\)=0.9 (see Fig. 2). For both models, predictions are averaged over levels of attention \(\beta \in \{1,\,2\}\) (the areas of Fig. 2 denoted by green shaded areas). Shown are results for the 20 participant groups who received the relevant triangle information; patterns are similar for all 40 groups (see Fig. S9).