Fig 1.
The Bayesian assumption of BTS holds in all treatments for each experiment.
Along with each plot, we provide the Pearson correlation coefficient, and associated p-values in parentheses, between the endorsed responses and the predictions made by responders. (A) In the Coin Flip experiment, the average predicted percentage of heads increases with the number of heads reported by the responders. (B) In the Dice experiment, the average predicted dice total increases with the sum of the reported dice rolls from the responders. (C) In the Pricing experiment, the average predicted reward selected by responders for completing the questionnaire increases with the selected rewards of the responder.
Fig 2.
BTS improves honesty in the Coin Flip experiment.
(A) 57% of reported coin flips were heads (43% tails) in the control treatment. (B) 53% of reported coin flips were heads (47% tails) in the BTS intimidation treatment. We expect 50% of coin flips to be heads if coin flips were reported honesty (represented by the black dashed line in both plots).
Fig 3.
BTS treatments improve honesty in the Dice experiment.
Probability mass function (PMF) of reported dice outcomes in (A) the control treatment, (B) BTS intimidation treatment, and (C) transparent BTS treatment. Honest reporting would produce a uniform distribution with values 1/6 in expectation (represented in each panel by the black dashed line).
Table 1.
The Pearson’s χ2 goodness-of-fit statistic for pairwise distribution comparisons in the Dice experiment.
Pairs of distributions are significantly different except for the distributions resulting from the two BTS treatments. Significance p-values are provided according to * = 0.1, ** = 0.01, and *** = 0.001.
Fig 4.
The distribution of the dice sums by treatment group.
The expected distribution of dice totals if dice outcomes were reported honestly is presented in black.
Table 2.
BTS produces statistically distinguishable distributions of dice totals.
The table displays the Pearson’s χ2 goodness-of-fit statistic for pairwise distribution comparisons. Each pair of distributions was significantly different from each other. Significance p-values are provided according to * = 0.1, ** = 0.01, and *** = 0.001.
Fig 5.
BTS treatments alter reward selection in comparison to the control treatment.
We present the probability mass function of selected rewards for task completion in the (A) control treatment, (B) BTS intimidation treatment, and the (C) transparent BTS treatment.
Table 3.
The pairwise Pearson’s χ2 goodness-of-fit statistic comparing selected rewards in the Pricing experiment.
Each pair of distributions is significantly different. Significance p-values are provided according to * = 0.1, ** = 0.01, and *** = 0.001.