Responding to Unfair Offers Made by a Friend: Neuroelectrical Activity Changes in the Anterior Medial Prefrontal Cortex

Discussion

A high rejection ratio was observed for unfair offers coming from the unknown proposer, confirming previous findings on inequity aversion and punishment of injustice in social interactions where the proposer is not known. This behavior stimulates communal fairness and altruism (Fehr and Gächter, 2002; Brosnan and De Waal, 2003). Our behavioral data indicate enhancement of utilitarian behavior when playing with a friend, i.e., an attenuation of altruistic punishment. In principle, this could be explained as a framing effect, in which friendship provokes a fundamental bias affecting the subjective perception of offer fairness, reflected in acceptance rates and proposer fairness ratings. Higher acceptance could also indicate an expectation that initial unfairness will be offset later in the game. Another explanation is that the participant sees no reason to altruistically punish their friend, and prioritizes personal gain as a consequence.

Similarly, it is possible that the participant attaches a certain subjective value to friendship, and wishes to avoid damaging it by sending a negative judgment of unfairness. This would be a long-term strategy in which the participants modulate their punishment responses as a function of the proposer targeting potential future gains. In this perspective, the species-wide benefit of extinguishing a maladaptive behavior would prevail when playing with unknown proposers, but be suppressed when playing with a friend to prevent offending them. In this direction, it is important to highlight that our findings are in agreement with and extend a previous study (Moore, 2009) showing that children share more with their friends than with others in a zero-sum game.

It is important to consider that the UG is typically played as a one-round game, implying that rejection cannot be due to self-interest as the proposer will have no chance to make new offers to a given recipient (van Dijk et al., 2004). In our experiment, however, the UG was played in several rounds with the same proposer/recipient pair. Thus, rejections might reflect bargaining behavior based on self-interest, and not only a response to offer fairness. At the same time, recipients played with a close friend and with an unknown person. With this, social distance might be an important variable to explain our results.

For instance, previous studies have already shown that identification increases cooperation on games such as Dictator Game and Prisoner's Dilemma as it decreases social distance (Frey and Bohnet, 1997; Bohnet and Frey, 1999). In our study, social distance combined with multiple rounds might have resulted in responses driven by a combination of retaliation, gain-related strategic behavior (when the proposer was the unknown person), and avoidance of subsequent conflict (when the proposer was the friend) [see Handgraaf et al. (2003) for a review on social utility in ultimatum bargaining]. Further, it is a well established notion that the expectation and prevention of future regret influences present decision making (Larrick and Boles, 1995; Zeelenberg et al., 1996; Zeelenberg and Beattie, 1997), and the fact that participants believed that they were playing with their friend might have promoted the acceptance of unfair offers as a way of minimizing the expectation of regret related to potential disappointment and conflict after the end of the game.

Our ERP findings are in agreement with the study by Boksem and De Cremer (2010), in which the amplitude of the MFN, peaking around 260 ms over frontal sites, indexed the perceived level of offer fairness (more pronounced for unfair offers), and was modulated by the participant's level of concern for fairness. Similarly to the present experiment, in their study participants played multiple rounds of the game with each proposer. Our data extend their results by demonstrating that the MFN is strongly responsive to social distance, reversing polarity when the proposer is a close friend rather than an unknown person. The MFN is generally assumed to originate from the anterior cingulate cortex (Gehring and Willoughby, 2002). However, the inverse solution in our case revealed engagement of anterior structures of the medial prefrontal cortex (MPFC). Specifically, the comparison between the two proposers revealed more negative current difference for offers from the unknown person within BA10 and BA11. This discrepancy may indicate that, as found with many mid- and long-latency components, the underlying cortical generators involve a network of regions, whose relative contribution on the inverse reconstruction maps may vary as a function of the specific task features and stability of the solution. It may also more simply be a consequence of inverse solution instability and represent activity actually originating from BA24/32.

Broadly, it has been shown that BA10 and BA11 are engaged during tasks involving cooperation in two-person decision-making games (McCabe et al., 2001), and subserve person perception as well as mentalization (for review, see Amodio and Frith, 2006). As confirmed in a meta-analysis by Steele and Lawrie (2004), these regions are also crucially related to emotion. In our task, efforts to understand the other's intentions and motives (Frith and Frith, 2003; Frith and Singer, 2008) might have had an important role in driving activity in BA10 and BA11 and might have guided the behavioral responses. Indeed, as proposed by Amodio and Frith (2006), the anterior MFC can play a role in “meta-cognitive processes of reflecting on feelings and intentions.”

More specifically, previous experiments point to a close relationship between the MFN and dopaminergic activity. In the studies by Holroyd et al. (2004) and Nieuwenhuis et al. (2004), undesirable outcomes were associated with strongest negativity, while positive outcomes elicited weaker MFN responses. In parallel, converging results indicate that unexpected rewards elicit positive responses in the MFN time window (Potts et al., 2006; Holroyd et al., 2008, 2011; Pfabigan et al., 2011). Together, these findings closely follow the known pattern of dopaminergic system response to reward and reward prediction error. In fact, it was previously proposed by Holroyd and Coles (2002) that frontal negative responses may represent increased deep-layer cortical activity due to attenuated inhibitory dopaminergic input. By the same logic, a positive scalp potential could result from deep-layer inhibition during enhanced dopamine release, e.g., reward delivery.

From this perspective, our results would imply that unfair offers dampen dopaminergic activity with respect to fair offers when originating from an unknown person, but boost it when made by a friend. This effect cannot be explained in terms of altruistic punishment: our behavioral data indicate more pronounced altruistic punishment when playing against an unknown person, and the well established notion that altruistic punishment is associated with increased dopaminergic activity (e.g., Strobel et al., 2011) would lead to predicting the opposite difference. An alternative account is that, as proposed by Boksem et al. (2011a), the MFN may represent an abstract marker of “social pain,” i.e., may reflect negative outcomes of a social nature such as unfair treatment by others. For example, Boksem et al. (2011a) evaluated the MFN during a game in which the player's performance did not affect the other participant (an unknown person), and reported that the MFN was largest in trials during which the participant lost and the other player won, leading to the idea that the MFN might index the comparison between one's and other people's experienced outcomes, i.e., fairness related to social structures (Handgraaf et al., 2003). Since the unfair offers originating from the friend were perceived as less unfair than those from the unknown person, the difference in “social pain” might have reflected in MFN amplitude. Speculatively, this account might even explain the observed inversion in polarity: if receiving an unfair offer from an unknown person is more “painful” (i.e., deactivates the dopaminergic system more strongly) than receiving it from a friend, once average reward expectations have built up over a few trials, the dopaminergic system may generate a positive reward prediction error signal for the less “painful” situation of receiving an unfair offer from the friend and a negative signal for more “painful” unfair offers by the unknown person.

Clearly, one limitation of this study is that the experimental comparison, i.e., friend versus unknown person, contained multiple elements such as social distance, similarity, sympathy, and trustfulness. As a consequence, we are unable to directly map the observed effects to a specific factor. Further work is necessary to try to disentangle the effects of these individual factors, for example by isolating knowledge, degree of sympathy, etc. with specific regressions; this is likely to require a much larger population to attain adequate statistical power. Another limitation is the absence of real monetary endowment. While we cannot exclude that this may have introduced differences with respect to experiments in which financial reward was available, other studies in this field have obtained consistent findings without paying participants (e.g., Polezzi et al., 2008; Boksem et al., 2011b), and it has been established that small monetary incentives do not necessarily boost task performance and may even have paradoxical effects (Gneezy and Rustichini, 2000).

More than two millennia ago, Aristotle wrote: “… we prize friendship for itself, even though nothing else is likely to come to us from it.” Our data demonstrate that this tendency has a neurobiological correlate at the level of anterior MPFC engagement, likely reflecting dopaminergic system activity, which accompanies the strong behavioral effects observed during a social decision-making task. There are fundamental differences in the way we process offers coming from a friend or an unknown person. These can be either seen as beneficial, since unfair offers do not elicit a punishment response among friends, or as socially maladaptive, as we become lenient about unfair proposals. Further experiments are needed to disentangle the exact dynamics underlying the observed behavioral and neuroelectric effects, in particular to verify whether the MFN amplitude can be modeled as a reward prediction error signal in the context of “social pain.”