Manipulation of Pro-Sociality and Rule-Following with Non-invasive Brain Stimulation

Decisions are often governed by rules on adequate social behaviour. Recent research suggests that the right lateral prefrontal cortex (rLPFC) is involved in the implementation of internal fairness rules (norms), by controlling the impulse to act selfishly. A drawback of these studies is that the assumed norms and impulses have to be deduced from behaviour and that norm-following and pro-sociality are indistinguishable. Here, we directly confronted participants with a rule that demanded to make advantageous or disadvantageous monetary allocations for themselves or another person. To disentangle its functional role in rule-following and pro-sociality, we divergently manipulated the rLPFC by applying cathodal or anodal transcranial direct current stimulation (tDCS). Cathodal tDCS increased participants’ rule-following, even of rules that demanded to lose money or hurt another person financially. In contrast, anodal tDCS led participants to specifically violate more often those rules that were at odds with what participants chose freely. Brain stimulation over the rLPFC thus did not simply increase or decrease selfishness. Instead, by disentangling rule-following and pro-sociality, our results point to a broader role of the rLPFC in integrating the costs and benefits of rules in order to align decisions with internal goals, ultimately enabling to flexibly adapt social behaviour.

. Interface screenshots. Participants repeatedly had to drag a ball to either the blue or orange box. In the 'me' block and 'other person' block, the decision had real financial consequences for the participant or another person, respectively, that changed across trials (see a for an example-trial). In the 'me vs. other person' block, participants had to decide between allocating a sum of money between themselves and another person. The sum, as well as the allocation-choice changed across rounds (see b for an example-trial). Table S1 and S2 show the regression results, estimating the accumulated money for oneself ('me' block), or another person ('other person' block) across the three tDCS conditions.  Table S3 shows the regression results for the fairness evaluations depending on tDCS condition, the hypothetical transfer, and its interaction.

Additional Analysis
We specifically wanted to test the role of the right LPFC in rule-following when the rule did not coincide with what participants would choose in the 'free' part (i.e. rules that demanded to financially hurt oneself or the other person), while showing that behaviour is unchanged when internal goals and the rule coincide (i.e. rules that are beneficial or neutral).
However, due to aggregating the data across consequences we lost possibly valuable variability related to the degree of how beneficial or harmful following the rule really was (see experimental setup & design).
We therefore also fitted two more complex models to the non-aggregated data, using the binary trial-by-trial response variable (0 = not following the rule, 1 = following the rule). As predictor, we used the continuous rule consequence variable, that varied between -30 (following the rule would lead to a loss of 30 cents) and +30 (following the rule would lead to earning 30 cents more than violation the rule). In this regression, we included the observations of all participants and dummy-coded unconditional rule-following (participants who followed the rule across all trials without being influenced by its consequence at all).
To account for the dependencies within subjects, we fitted two (Bayesian) random intercept binomial regression models using JAGS/R to the 'me'-trial and 'other person'-trial data, respectively. Non-informative Gaussian priors (m = 0, sd = 100) were used for each predictor and non-informative uniform priors (range 0 to 100) for the error terms. We used three parallel chains. For every estimated coefficient, the potential scale reduction factor (Gelman and Rubin Diagnostic) was below 1.05, indicating good mixing of the three chains and thus high convergence. Regression tables reported below show estimated coefficients (log-odds) together with the 95% confidence interval (CI, also called highest density interval in the Bayesian framework). Note that, since noninformative priors were used, a 95% CI that only contains negative or positive values can be interpreted as significant at a p = .05 two-sided threshold from a frequentist perspective. Fitting the models using restricted maximum likelihood (REML) as implemented in the lme4 package in R revealed similar estimates and resulted in the same statistical inferences. Table S4 and Figure S2a show the fitted model for 'me'-trials. As can be seen, the probability to follow the rule increased the more beneficial the rule was, up to 100% for rules that would yield beneficial outcomes to the participant (consequence ≥ 0) in all three tDCS conditions. However, participants under cathodal and sham tDCS, compared to anodal tDCS, had a higher likelihood to follow harmful rules and, therefore, had a steeper increase in rule obedience towards more beneficial consequences.
Table S5 and Figure S2b shows the fitted model for 'other person'-trials. Again, the probability to follow the rule increased the more beneficial the rule was, up to 100% for rules that would yield beneficial outcomes to the other person (consequence ≥ 0) in all three tDCS conditions. However, participants under cathodal, compared to anodal, tDCS had again a higher likelihood to follow harmful rules.

Selfish vs. pro-social rules
In the main manuscript, we focus on selfishness (money accumulated for oneself at the expense of another person) when participants were free to decide across tDCS conditions and looked at how selfishness was attenuated by a rule that dictated to take the pro-social choice in half of the trials.
To further disentangle rule-following in the 'me vs. other person' block, we separately looked at rule-following when the rule was selfish (e.g. the rule dictated to take 10 cents and give 0 cents to the other person instead of taking 5 cents and giving 5 cents) vs. when the rule was pro-social (e.g. the rule dictated to take 5 cents and give 5 cents to the other person instead of taking 10 cents and giving 0 cents). We compared this to the intrinsic behaviour of participants in the 'free' part (i.e. percentage of selfish vs. pro-social choices). Figure S3 separately shows average selfishness when freely deciding and selfishness when the rule dictated to be selfish (a), and average pro-sociality when freely deciding and pro-sociality when the rule dictated to be prosocial (b), across tDCS condition.
When freely deciding, participants under cathodal tDCS chose the selfish option more frequently (64.7%) compared to participants under anodal tDCS (54.4%), with sham tDCS in the middle (57.7%). All participants frequently followed the rule, when it dictated to take the selfish option, especially under cathodal tDCS. Participants under cathodal tDCS followed a 'selfish rule' 98.3% of the time, followed by sham tDCS with 91.7% and anodal tDCS with 89.3% (see Figure S3a). Interestingly, across all conditions, participants increased their selfish choices when the rule dictated them to be selfish, indicating that a selfish rule can serve as an excuse to act selfishly.
When the rule was to take the pro-social option, participants under cathodal tDCS changed their intrinsic behaviour the most, as can be seen by the difference between voluntary pro-social choices vs. rule-induced prosocial choices ( Figure S3b). On the contrary, participants under anodal tDCS chose the pro-social option in 45.5% of the trials on average when freely deciding and only marginally deviated from their voluntary behaviour when a rule dictated them to choose the pro-social option (44.6%).
Taken together, this pattern led to the highest attenuation of selfishness under cathodal tDCS. While participants accumulated more money for themselves at the expense of the other person under cathodal vs. anodal tDCS when freely deciding, participants under cathodal tDCS accumulated significantly less money when confronted with a rule that dictated pro-social choices in 50% of the trials, while participants under anodal tDCS stayed more consistent with their free choices ( Figure S4).