Figure 2

Probability distributions of the eccentricity of the posted ideas at different popularity levels (number of likes, i.e., amount of attention they attracted). Popularity levels are represented in different color bins, low (blue) and high (orange) (and, in case of GAB data, medium (green)). The vertical dashed lines show the average value of eccentricity for each popularity level. (a) Plots for the dataset collected in the laptop tagline writing experiment (high collaboration task). Two-sample Anderson–Darling test with Bonferroni correction on unequal sample sizes (n_>2 = 59, n_<=2 = 815) (b) Plots for the dataset collected in the short story writing experiment (low collaboration task). Two-sample Anderson–Darling test with Bonferroni correction on unequal sample sizes (n_<=2 = 669, n_>2 = 38) (c) Plots for the dataset collected from GAB. Two-sample Anderson–Darling test with Bonferroni correction on unequal sample sizes (n_<=10 = 130,234, n_11-100 = 1866, n_>10090). In the plots of high popularity levels, the tail of the probability density function becomes broader in all three data sources. The average eccentricity also increases as the popularity level goes up in all cases.