Fig 1.
The current study aims to answer the question of why it is that repeated coordination may boost one’s willingness to cooperate with one’s partner, by testing separately for the emergence of altruistic motivation (Hypothesis 1) and trust (Hypothesis 2) towards one’s partner.
Table 1.
Analysis results from Experiment 1 using mixed-effects logistic regressions of partner condition and temptation level on subject’s choices.
Model 1 is used for inferring the effect of partner condition on altruism rates; it contains a by-subject random coefficient for this variable. Likewise, Model 2 is used for inferring the effect of temptation level on altruism rates; it contains a by-subject random coefficient for this variable.
Fig 2.
At the beginning of each trial, an image of the partner’s player number was displayed, which was either the same (Fixed Partner Condition) or different (Variable Partners Condition) on every trial. Then, the partner chose one of two values. The participant did not see what these two values were, and did not see what the partner had chosen, but was then herself presented with two values to choose between. One of these, indicated in green or blue, was the same value that the partner had chosen (coordination option); the other, indicated in orange, was an alternative value (alternative option). We varied whether, and to what extent, the alternative option constituted a temptation.
Fig 3.
Experiment 1 payoff structure.
Payoffs for each trial were determined as follows. If the participant chose the coordination option, then each received the corresponding amount (left box above). If the participant chose the alternative option, then the participant received the amount corresponding to the alternative option and their partner received no reward (right box above). Participants were thus aware that both their and their partner’s payoffs depended on the choice they made. Payoffs were displayed for 3000 ms before participants proceeded to the next trial. The amounts associated with the payoffs for participants’ coordination and alternative options varied, over pre-specified intervals and ranges, unpredictably across trials.
Fig 4.
Graphs show mean altruism rates (proportion of trials in which participants chose the coordination option) by Partner Condition for the corresponding temptation level (see: Methods) of participants’ alternative option. Points represent where empirical data was collected, and error bars represent 95% confidence intervals calculated using binomial tests. Regression results (see Table 1) show that altruism rates were significantly higher when participants coordinated with the same partner (Fixed Partner) on every trial than when coordinating with different partners (Variable Partners) on each trial. In addition, altruism rates are significantly decreasing with increases in the level of temptation of the alternative option.
Table 2.
Overview of differences in how participant and partner are impacted by each other’s choices in Experiments 1 and 2.
Fig 5.
At the beginning of each trial an image of the partner’s player number was displayed, which was either the same (Fixed Partner Condition) or different (Variable Partners Condition) on every trial. Next, two boxes (coloured in green and blue respectively) containing payoff options appeared on the screen, and the participant waited for a fixed duration of 4000 ms while the partner ostensibly selected one of these two boxes. Crucially, participants were led to believe that one of the options presented to the partner was a more or less tempting alternative option. Participants then chose whether to trust (i.e. to select the blue or green box) if they expected that their partner had previously chosen this same option, or to exit (i.e. to select the alternative option, in orange) if they did not trust their partner to have chosen the mutually beneficial option. Participants’ alternative options always entailed a lower reward.
Fig 6.
Experiment 2 payoff structure.
If the participant chose the alternative option (orange), then payoffs were determined as in the top half of the figure above: the participant received the guaranteed amount (300), and the partner received the amount corresponding to the option s/he had chosen: 600 if s/he had chosen the alternative option (blue) or 500 if she had chosen the coordination option (green). Conversely, if the participant chose the coordination option, the payoffs were determined as in the bottom half of the figure above: If the partner had chosen the coordination option (green), then each player received the corresponding amount (500); if the partner had chosen the then both chosen the alternative option (blue), the partner received the corresponding amount and the participant received 0. Participants should thus only choose the coordination option if they believed their partner had likewise done so. While payoffs for the coordination option and for participants’ alternative options were fixed, we varied the amount associated with the partner’s alternative option unpredictably across trials. Payoffs were displayed for a set amount of time before participants proceeded to the next trial.
Fig 7.
Graphs show mean trust rates (proportion of trials in which participants chose the coordination option) by partner condition for the corresponding temptation level of partners’ alternative option. Points represent where data was collected, and error bars represent 95% confidence intervals calculated using binomial tests. Regression results (see Table 2) show that trust rates were significantly lower when participants coordinated with the same partner (Fixed Partner) on every trial than when coordinating with different partners (Variable Partners) on each trial. In addition, trust rates are significantly decreasing with increases in the level of partners’ temptation of the alternative option.
Table 3.
Analysis results from Experiment 2 using mixed-effects logistic regressions of partner condition and partner temptation level on participants’ choices.
Regression models are identical to those used in Experiment 1. Again, Model 1 and Model 2 are used for inferring the effects of partner condition and temptation level, respectively, on trust rates, given the inclusion of their corresponding by-subject random coefficients.
Fig 8.
Our research investigated two competing hypotheses for why repeated coordination may boost an agent’s willingness to cooperate with their partner.
Our studies relied on the background assumption that either trust or altruistic motivation, or both, would increase agents’ willingness to cooperate with their partners. In each of our two separate studies, we tested a distinct hypothesis and found evidence that repeated coordination boosts altruistic motivation towards one’s partner but found no evidence of an effect of repeated coordination on trust.