Through the use and validation of a novel, third-party social discounting task that directly assessed impartial altruism in real-world extraordinary altruists, we find that these individuals indeed show more impartial altruism compared to demographically similar controls. When offered the opportunity to allocate resources to a close other or to divide them equally between a close and more distant other, they were more likely to divide the resources equally than were typical adults.
Modifying the classic social discounting paradigm for third-party decision-making allows for a direct, veridical behavioral measure of impartiality using participants’ own social networks and employing real-world payouts in order to extend an otherwise hypothetical decision into a task with tangible outcomes. Previous measures of (im)partiality have relied on self-reported Likert responses, such as the fairness/reciprocity and ingroup/loyalty foundations in the Moral Foundations Questionnaire 26,27 or relied on self-reported levels of concern for different entities, as in the Moral Expansiveness Scale 28. Though useful, these scales do not directly assess behavior or real-world outcomes. Meanwhile, studies linking impartial versus partial altruism to anticipated relationship quality do so via third-party judgments of hypothetical characters 3,6,7. By contrast, participants in this novel third-party social discounting task made decisions that resulted in real-world payouts to actual others in their social networks based on their decisions. Responses in this third-party social discounting task followed the hyperbolic decay of the classic paradigm, validating that variation in social distance predicts generosity when respondents choose how to allocate resources between two individuals in their social network (as opposed to choices simply reflecting selfish versus selfless tradeoffs, as is portrayed in the class social discounting task) 8.
Prior work suggests such impartial generosity might conflict with the favoritism and prosocial prioritization of close others usually associated with close relationships and undermine the quality of these relationships 3,6. But, using two measures of impartial altruism––real-world altruistic kidney donation and increased impartiality in the third-party social discounting task––our results indicate that the propensity for impartial altruism need not come at the cost of the quality of close relationships. In the present study, relationship satisfaction for highly impartial participants was not significantly lower than that of controls, who showed relatively more favoritism for close others. This may be due to value homophily, as altruists’ close others also showed low levels of social discounting (were similarly generous and for distant relative to close others). In light of prior findings that only relatively impartial participants do not view impartial altruists as worse relationship partners 7, this suggests similar levels of generosity and impartiality within close relationships may serve as a protective factor for highly impartial, altruistic individuals to maintain high levels of relationship satisfaction with similar close others. By the same token, controls and their close others exhibited similarly lower levels of impartial altruism. This underscores our findings related to the mechanisms underlying impartial altruism. We found that first-person discounting decisions predicted impartial altruism. These results are consistent with consensus bias—that altruists may implicitly respond on others’ behalf as they would on their own behalf. Importantly, however, the accumulated patterns contradict a false consensus bias, as altruists’ decisions appear to correctly reflect how their close others would actually respond.
We did not find that explicit beliefs about impartiality or morality predict impartial altruism. Participants completed measures of normative moral beliefs to assess whether impartiality reflects explicit beliefs about impartial beneficence 29, moral tolerance (respect for moral priorities that differ from one’s own), or low moral relativism (impartial decisions for others based on the belief that the value of impartiality should be applied to all). However, none of these belief frameworks were significant predictors of impartial altruism. Although past work has found impartial beneficence to predict extraordinary, selfless altruism 13, it may be that a scale measuring selflessness does not predict impartiality that extends beyond the self, as in a third-party social discounting task assessing allocations between two others rather than allocations between the self and another person.
At face value, our findings might seem to contradict past work in moral psychology in which impartial characters are viewed as worse relationship partners 3,6,7. The finding that third-party character judgments do not corroborate real-life reports of relationship quality could reflect the effects of salience bias on judgments in such scenarios, including moral judgments 30. Salience bias is the tendency to over-inflate the importance of salient information simply because is it more noticeable or prominent than more relevant, less salient factors 31–34. This seeming discrepancy also likely reflects the different composition of the participant samples in third-party judgment studies versus the present study. The close others of altruists were not the random or convenience samples used in many moral judgment studies but were the closest friends or family members of a highly impartial person. These relationships have been selected and maintained over time. Others whom altruist have encountered who view their impartiality less favorably may be less likely to form or maintain close relationships with them.
Our results suggest a reason that altruism may not come at the cost of close relationship quality is value homophily––the phenomenon of preferential formation and maintenance of close relationships with those who share similar values, attitudes, and beliefs 17,18,20,21. Though altruists were less partial toward their closest others compared to controls, the closest others of altruists were similarly less partial and more generous when compared to the closest others of controls. Similar values have been found to benefit cooperation 18, and moreover, similar levels of cooperative behavior have been found to extend to up to two degrees of separation 35. Yet how value homophily influences close relationship quality has not been previously assessed, to our knowledge. More, value homophily provides insight into why such impartial altruism does not negatively affect close relationships. Rather than one party viewing impartiality favorably and the other party viewing impartiality negatively, our dyadic data suggests that patterns of generous decision-making are similar within dyads. That close others respond similarly during the discounting task aligns with a growing body of neuroscience findings that social closeness predicts similarity in how people respond to events in the world at neural level 36. Future work should seek to investigate the causes of value homophily in close relationships. Although close relationships research suggest similarity in close relationships results from selection effects rather than partners becoming more similar over time 37, altruism research suggests that altruistic behavior and preferences can be socially contagious 38,39. Future work should also investigate how more distant, less relationally satisfied dyads vary (or do not vary) in impartiality, generosity, and other values, as including a more diverse range of relationships might reduce ceiling effects in relationship satisfaction scores.
Our findings should be considered in light of some limitations. One limitation is that efforts to recruit controls who were demographically matched with altruists resulted in a sample that was relatively racially and ethnically homogenous and on average, richer and more educated than the average American. This in part reflects the selection criteria applied to living kidney donors, who must typically meet stringent health criteria before donating, including low risk for kidney disease, which disproportionately affects lower-income, Black, and Hispanic adults in the United States. However, it is reasonable to assume our findings would apply to more demographically representative samples, as prior research has found that more diverse altruistic samples, such as bone marrow donors and heroic rescuers, 16,40,41 display similar reductions in first-party social discounting as altruistic kidney donors 10 as well as other similar personality and behavioral traits. It could be argued that the behavior of altruists in this study may reflect their awareness that they were recruited for a study on altruism, introducing an experimental demand bias. Several pieces of evidence argue against this possibility. First, altruists do not score highly on self-report altruism scales more directly related to altruism and more susceptible to demand effects 10,42. Second, prior work finds that the traits and behaviors that most reliably distinguish altruists from controls are not those predicted by the average person, suggesting that altruists are not simply conforming to stereotype-consistent behaviors 10.
Taken together, our findings address fundamental questions regarding the potential real-world social costs of impartial altruism. We used a novel behavioral paradigm to veridically model impartial altruistic decisions in real-world altruists and demographically similar controls, along with their closest friends and family members. We found that impartial altruists tend to have impartial friends and family members, and similar close relationship quality as controls. This study contradicts prevalent beliefs that care for others is limited in quantity, a belief associated with reduced empathy and caring behaviors toward distant others 43. Rather, we find that altruism toward socially distant others need not prevent impartial altruists from enjoying the benefits of close social relationships.
Materials, Methods, and Data Availability Statement
Recruitment procedures are described in the pre-registration and approved by the Georgetown University Institutional Review Board. Access to data, materials, analysis code, and the preregistration can be found in the study’s online repository [dataset] 44. The study was performed in accordance with approved guidelines and regulations, and informed consent was provided by all participants.
Participants
Participants were recruited through our database of verified altruistic kidney donors in North America and demographically similar controls. Additional controls were recruited via the online recruitment database ResearchMatch. Interested participants invited their socially closest other, described as their “closest friend, partner, or relative,” to contact the researchers independently. Once both participants in a dyad agreed to participate, they were instructed to separately complete a survey using Qualtrics.
260 participants completed all laboratory tasks, exceeding our target sample size (N = 208), but in accordance with our stopping rule (N = 260). The sample size was determined through a power analysis conducted in R using G*Power (‘pwr’ package) 45. The sample included 59 altruists, 59 close others of altruists (nominated by altruists as their closest other), 71 controls, and 71 closest others of controls. Altruists and controls were closely matched in gender distribution and race/ethnicity (p = .509; see SI Tables S1-S6 for full demographic reports). However, following exclusions, the altruist sample was slightly older (Maltruists = 54, Mcontrols = 47, p = .03), higher in income (Maltruists=$90,000–179,999, Mcontrols=$60,000–89,999, p = .0002), and more educated (% ≥ 4-year degree = 81.3% for altruists; % ≥ 4-year degree = 73.26% for controls; p = .023). Altruists’ closest others and controls’ closest others were similar in gender distribution, education level, and race/ethnicity. Altruists’ closest others, however, were older (Maltruists’ friends = 56, Mcontrols’ friends = 47, p = .003) and had higher household income (p = .012) compared to controls’ closest others. Thus, we controlled for gender, age and income in all analyses. Because income and education are typically highly correlated, our preregistered analyses did not include both factors in statistical models in order to reduce collinearity 46. Altruists and controls did not differ significantly in relationship type (i.e., spouse, friend, roommate, etc.; see Figure S1 in the SI) across nominations.
Procedures
Social Discounting. All participants first completed a standard social discounting task 8,9,47. They next completed a novel social discounting task created for this research protocol to assess third-party social discounting. Its design closely parallels a standard social discounting task, except participants allocate money between closer and more distant others (e.g., allocating $145 to N = 1 (favoritism) versus splitting $150 evenly between N = 1 and N = 50 (impartial). Because the task requires choosing to favor a close other versus impartiality, the number of choices decreases as the social distance of the close other increases (Fig. 4).
Before the experiment, participants were requested to provide the names of individuals from their personal social network who corresponded to various social distances, namely N = 1, 2, 5, 10, 20, and 50. They were given the following instructions: "Imagine a list of 100 people in your social circle, with person #1 representing the closest relationship and person #100 being someone completely unfamiliar to you." While it's possible for more than one person to represent a given social distance within social networks, participants were instructed to select only one person for each distance 48. The experiment also included social distance level 100, representing strangers whose names might not be known 8. Subjects were not required to provide a name for this level. Therefore, a total of seven social distances were considered.
Participants made a total of 252 decisions across the first-party and third-party social discounting tasks, each involving the seven previously described social distances. The task was structured into seven blocks, each containing nine trials. Following established protocols 48, participants were instructed to mentally envision the seven individuals on the list (N = 1, 2, 5, 10, 20, 50, or 100) and make nine binary decisions concerning retaining and/or sharing specific amounts of money with each of these individuals. In each trial, participants had to indicate whether they preferred to keep a sum of money exclusively for themselves (the selfish option) or to retain a sum of money for themselves while also sharing an identical amount with the Nth person on their list (the generous option).
The selfish options involved keeping varying amounts of money, ranging from $155 down to $55, decreasing in $10 increments. On the other hand, the generous option consistently entailed keeping a fixed amount of money and sharing that same amount with the Nth person (i.e. “$75 for you, $75 for [N]”). Participants were required to confirm their knowledge of each N's identity before task initiation.
Prior to commencing the task, participants were informed that their payments were linked to their task responses, ensuring that the experiment adhered to ethical standards and avoided any form of deception, meeting the requirements of behavioral economics. Based on their actual choices, participants received 10% of the total from a randomly selected trial. If a participant chose the generous option, the designated person also received 10% of the amount allocated for them. If necessary, participants were asked after the task to provide an email address or phone number for contacting the relevant person to facilitate payment through PayPal or an Amazon gift card. In cases where the randomly selected trial involved a person at a social distance of 100, a random recipient received the payment from our database of future participants.
Participants then completed a battery of moral belief questionnaires, which included: the Moral Relativism Scale, Moral Tolerance Scale (MRS, MTS) 49 and the Oxford Utilitarianism Scale (OUS) 29. These scales assessed explicit moral beliefs in order to evaluate the role of normative moral beliefs in impartial altruism.
Participants also completed the McGill Friendship Questionnaire Respondent’s Attachment and Friend’s Functions subscales (MFQ-RA) MFQ-FF) 50 to assess the quality of relationships between altruists and controls and their closest others. See the Supplemental Text in the SI for more information regarding these self-report measures, as see Table S13.