Figure 1
Diagram of gameplay. a. The Decision Tree of the Minnesota Trust Game shows possible outcomes for each player. The first mover first chooses between both players getting a small reward (S) or letting the second mover decide. The second mover can choose between both players getting a larger mutual reward (M) or an adverse payoff for the first mover (Ad) and a temptation T for the second mover. The table describes the values for S, M, T, and Ad for the two samples. The Suspiciousness condition is referred to as ‘SUS’ and the Rational Mistrust condition is referred to as ‘RMT’ when playing against a human partner. Participants also play against a fair coin, which are referred to as Risk aversion conditions or ‘RA(T)’, depending on the T amount. b. An example trial of the Rational Mistrust condition in the First Mover Game, with T = 25 and Ad = 0 (Sample 1). Participants saw question marks for the variable outcomes during the fixation, and then had a fixed amount of time to choose when the values were displayed. Their choice was confirmed during the confirmation window; however, they never received feedback about the choice made by the second mover.
Figure 2
Behavior and Model Simulations. a) Original Sample 1 aggregate behavior. b) Simulated data from the original Fehr-Schmidt model, which does not match the pattern of Sample 1 aggregate behavior well. c) Simulated data from the best-fitting Spite Sensitivity model. d) We recovered parameters based on simulated data created in 2c and then re-simulated data based on the recovered parameters to test the stability of the model. Behavior is very similar to that seen in Figures 2a and 2c. Note, risk increases along the X-axis, as indicated by the decreasing Adverse Payoff. RA15 is the low temptation condition against the coin. RA25 is the high temptation condition with the coin. SUS15 is the Suspiciousness condition with the human partner. RMT25 is the Rational Mistrust condition with the human partner. The shading indicates 95% CI.
Figure 3
Replication of model comparisons. a) Average performance seen in Sample 2, in which participants in the Suspiciousness condition showed higher trust than the other conditions. b) While our model separates out the two partner conditions and in general replicated the original data well, it underestimated actual trust in the Rational Mistrust condition. Note, risk increases along the X-axis, as indicated by the decreasing Adverse Payoff. RA10 is the low temptation condition against the coin. RA14 is the high temptation condition with the coin. SUS10 is the Suspiciousness condition with the human partner. RMT14 is the Rational Mistrust condition with the human partner. The shading indicates 95% CI.
Figure 4
From individual choices to thresholds. a. Example individual results for behavior, showing Heaviside threshold (dotted vertical line) determined for condition. The y-axis is a binary choice between trusting (1) and not trusting (0). The x-axis shows the adverse payoff in reverse order, such that lower thresholds (further to the right) indicate increased risk. b. Scatter plot of Sample 1’s thresholds comparing across the low and high temptation T values for coin conditions (blue) and partner conditions (red). Beside each axis, histograms show the relative count for each condition. The change in temptation had little impact on individuals’ thresholds when the second mover was a coin but did when it was a human.
Table 1
Correlations between Spite Sensitivity model parameters and MTG condition thresholds.
| PARAMETER | RA-LOW | RA-HIGH | SUS-LOW | RMT-HIGH |
|---|---|---|---|---|
| SAMPLE 1 | ||||
| Inverse temperature | –.26*** | –.23*** | –.24*** | –.11 |
| Envy | .38*** | .52*** | .15* | .53*** |
| Estimated spite-guilt | .02 | .01 | –.38*** | –.19** |
| General Risk Aversion | .70*** | .65*** | .34*** | .32*** |
| Social Risk Aversion | .1 | .13* | –.19** | .25*** |
| SAMPLE 2 | ||||
| Inverse temperature | .06 | –.13 | .17 | .11 |
| Envy | –.11 | –.15 | –.36** | –.03 |
| Estimated spite-guilt | –.11 | –.05 | –.53*** | –.03 |
| General Risk Aversion | .77*** | .72*** | .55*** | .43*** |
| Social Risk Aversion | –.13 | –.15 | –.15 | .37** |
[i] Note: Spearman correlations for each parameter (rows) by condition (columns). RA-LOW is the Risk Aversion condition against the coin partner when temptation T was below the mutual payoff M; similarly, RA-HIGH refers to the Risk Aversion condition against the coin partner when temptation T was higher than the mutual payoff M. SUS-LOW refers to the Suspiciousness condition when temptation was below the mutual payoff for the human partner. RMT-HIGH refers to the Rational Mistrust condition when temptation was higher than the mutual payoff for the human partner. The top half of the table shows results for Sample 1, and the bottom half is for Sample 2. Italicized values represent results replicated in Sample 2. p < .05*, p < .01**, p < .001***.
Figure 5
Sample 1 Behavior on First Mover game comparing Median Split of Alienation Scores. When comparing behavior in the first mover game between those with low (loAL) or high Alienation scores (hiAL), we found that those with higher Alienation scores are less trusting in the Suspiciousness condition (SUS15 loAL vs. SUS15 hiAL). There were not significant differences in between the other conditions when comparing high or low Alienation. Light and dark red lines are for conditions against the human partner, while light and dark blue lines represent conditions against the coin partner. Solid lines represent low temptation trials (T = 15), while dashed lines represent high temptation trials (T = 25). Light-colored lines represent high Alienation, while dark-colored lines represent low Alienation trials.