People want friends to be more prosocial toward oneself than toward others, and sometimes prefer friends who are more vicious than prosocial, for instance, toward one's enemies

Sometimes we want vicious friends: People have nuanced preferences for how they want their friends to behave toward them versus others. Jaimie Arona Krems et al. Evolution and Human Behavior, February 23 2023. https://doi.org/10.1016/j.evolhumbehav.2023.02.008

Abstract: Intuition and research alike suggest that people prefer friends to be prosocial—particularly kind and trustworthy. Here, we examine these preferences in light of the fact that dyadic friendships are embedded in wider social networks. Because our friends recurrently interact with other people, and these friend-other interactions can have various positive and negative effects on us, people should possess distinct preferences not only for how our friends behave toward us but also for how friends behave toward different other people (e.g., strangers, rivals). In six studies (N = 1183; two pre-registered) with complementary designs and cross-national samples (U.S. community, U.S. student, India community), we find: (a) When the targets of best friends' behavior are not specified, people's friend preferences track how one wants friends to behave toward oneself. Replicating patterns found in past work, (b) people generally want friends to be kinder and more trustworthy than not. But (c) people also want friends to be more prosocial toward oneself than toward others, and (d) people sometimes prefer friends who are more vicious than prosocial, for instance, toward one's enemies. These findings challenge some long-held conclusions about friend preferences, expand the known range of traits preferred in close relationship partners, and enrich our understanding of what it means to deem people, for example, “kind,” as such evaluative personality concepts may by default be indexed to the self.

Introduction

Most work on close relationships, especially friendships, tends to focus on the dyad and thus on how people want to be treated by their dyadic partners (i.e., their friends) (Hall, 2012; Huang, Ledgerwood, & Eastwick, 2020; Sprecher & Regan, 2002; Wiseman, 1986). What matters is that one's friend treats one well. In line with such thinking, robust evidence suggests that people prefer friends who are, for example, kind to them and disfavor those who are vicious to them (e.g., Cottrell, Neuberg, & Li, 2007; Fehr, 1996; Hall, 2011, Hall, 2012; Perlman, Stevens, & Carcedo, 2014; Sprecher & Regan, 2002). Somewhat similarly to such ‘canonical’ findings, ‘cooperative accounts’ of partner choice might predict that what matters is a friend's overall prosociality—to oneself or others—and so people should prefer friends who are maximally prosocial (to oneself and others) (for reviews—but not necessarily support for—such accounts, see Barakzai & Shaw, 2018; Hess & Hagen, 2006).1

Here, we integrate adaptationist theories of friendship, which emphasize friends' roles in providing one another preferential social support (DeScioli & Kurzban, 2009; Tooby & Cosmides, 1996), with our embedded dyad framework, and we test subsequent predictions about what people want in friends. Briefly, the embedded dyad framework emphasizes that dyads (e.g., friend pairs) exist embedded in wider and often densely interconnected social networks, wherein one's friends inevitably interact with other people (e.g., Basyouni & Parkinson, 2022; Dunbar, 2018, 2021). As implied by both this framework and adaptationist theories of friendship, friend-other interactions are not only a recurrent feature of the social landscape, but they can also have potentially profound effects on one's friends, one's friendships, and (thus) one's outcomes. Together, these lines of work suggest that friends should, on average, radiate positive effects on the self—both directly (via friend behavior toward the self) and also indirectly (via friend behavior toward others). If friends interacted only with oneself, then we would expect people to prefer friends who are maximally and solely prosocial. But because one's friends also interact with other people, including one's rivals, we suggest that people might sometimes prefer friends who behave with greater monstrousness than brotherly love—extending and sometimes challenging expectations from canonical work on friend preferences and cooperative accounts of partner choice.

Friends are associated with many benefits to health and happiness (see, e.g., Dunbar, 2018, 2021). Friends may have also helped one another solve several recurrent fitness challenges, from ensuring sufficient access to resources for survival to winning agonistic conflicts (e.g., DeScioli & Kurzban, 2009; Tooby & Cosmides, 1996; Williams, Krems, Ayers, & Rankin, 2022). But presumably, such benefits depend(ed) on securing good friends—those able and willing to help one meet one's needs. Friend preferences may thus play an important role in the formation of such friendships. Friend preferences are thought to guide people to invest their finite time and energy on attracting and maintaining friends who fit this bill (e.g., Conroy-Beam & Buss, 2016; Krems & Conroy-Beam, 2020; but see, Huang et al., 2020).

What do people want in friends? Theoretically, there are myriad preferences that people could prioritize in friends—intelligence, left-handedness, physical attractiveness, dislike of cats, formidability, detached earlobes (e.g., Benenson, 2014; Eisenbruch and Roney, 2020; Hall, 2011; Lewis et al., 2011; Lukaszewski, Simmons, Anderson, & Roney, 2016; Williams et al., 2022). But decades of research seem to paint a clear picture of people's friend preferences2: Although not an exhaustive list, people tend to most prefer friends who are kind and trustworthy (e.g., Andreoni & Bernheim, 2009; Barclay, 2016; Cottrell et al., 2007; Erikson, 1950, Erikson, 1964; Gurven & Winking, 2008; Hall, 2011; Hatfield, Traupmann, & Sprecher, 1984; Holmes & Rempel, 1989; Panchanathan & Boyd, 2004; Rempel, Holmes, & Zanna, 1985; Schwartz & Bardi, 2001; Shaw, DeScioli, & Olson, 2012; Yamagishi & Yamagishi, 1994). For example, people sometimes compete for friends by advertising their kindness (Barclay & Willer, 2007; Reis & Gruzen, 1976). People also value trustworthiness even over other desired traits, including intelligence and attractiveness (Cottrell et al., 2007).

Although less work examines traits disfavored in friends, people tend to eschew those who seem vicious or indifferent (e.g., Benenson, 2014; Hall, 2011, Hall, 2012; Shinada, Yamagishi, & Ohmura, 2004; Walster, Berscheid, & Walster, 1973).3 For example, people not only prefer kind friends, but they also disfavor the appreciably unkind, such that disagreeable individuals tend to be befriended less (Jensen-Campbell, Knack, Waldrip, & Campbell, 2007; Nettle, 2006; Selfhout et al., 2010). People also strongly prioritize friends' reciprocation of valuation and caring. A lack of reciprocated care from friends can end relationships (Delton & Robertson, 2016; Delton et al., 2023; Kenny & La Voie, 1982; La Gaipa & Wood, 1981; Rose, 1984; Shaw, DeScioli, Barakzai, & Kurzban, 2017; Walster et al., 1973). Though more tentative, people might also disfavor exploitative or impartial friends. People detect and avoid cheaters (e.g., Cosmides & Tooby, 1992), and, even as people consider impartiality a virtuous and desirable trait (Tyler, 2000), people consider friends' impartiality toward them undesirable (Shaw et al., 2017).

We additionally note that classic work in social psychology suggests that relationship partners tend to be similar, familiar, and nearby (Barry, 1970; Bornstein, 1989; Byrne, 1971; Newcomb & Svehla, 1937; Zajonc, 1968). This does not imply that these traits are necessarily prioritized in friendship formation, however (DeScioli & Kurzban, 2012).

Previous work in social psychology, relationship science, and related areas has often tended to focus on dyadic relationship processes (see, e.g., Basyouni & Parkinson, 2022; Dunbar, 2018; Fehr, 1996; Merrie, Krems, & Sznycer, n.d.). In line with this, friend preference research has tended to focus on the friend dyad, explicitly or implicitly. For example, participants might respond to items assessing how friends should behave toward oneself: “can make me laugh” (Oswald, Clark, & Kelly, 2004), “will cheer me up when I am sad” (Zarbatany, Conley, & Pepper, 2004), “really listens to what I have to say” (La Gaipa, 1987), “goes out of his/her way to help me” (Bank, 1994). Other friend preference measures—for example, asking about whether an ideal friend “is helpful” (e.g., Krems & Conroy-Beam, 2020; Williams et al., 2022)—could also be inferred as implicitly asking about ideal behavior toward oneself (e.g., “is helpful to me”; see Lukaszewski & Roney, 2010). In all, this work has generated a rich body of knowledge about friend preferences (see Hall, 2012), which we refer to as canonical preferences. Again, this work generally suggests that people prefer friends who are maximally prosocial (toward oneself) and eschew friends who are antisocial (toward oneself).

Another body of work on partner choice has focused on the importance of a partner's prosociality or cooperativeness. These cooperative accounts typically assert that people select partners (e.g., friends) based on the cooperative benefits they could provide (e.g., Baumard, André, & Sperber, 2013; Kenny, Mohr, & Levesque, 2001; Rand & Nowak, 2013), which can be gleaned from reputational information. On this view, people should prefer partners who have—and/or have reputations for—maximal prosociality—both toward oneself (direct reciprocity; e.g., preferring those kind to us) and others (indirect reciprocity; e.g., preferring those kind to others; for a review of cooperative accounts, see Hess & Hagen, 2006). Such accounts might additionally predict that people prefer friends who behave with minimal viciousness, indifference, or other dimensions of antisociality—again, toward oneself or others.

We derived additional predictions from an integration of our embedded dyad framework with adaptationist models of friendship. Briefly, this framework emphasizes that dyadic relationships (here, friendships) are embedded in wider networks, wherein one's friends inevitably interact with and have their own relationships with other people. These interactions can affect one's friends, one's friendships, and (thus) one's own outcomes in potentially profound ways—both positive and negative (e.g., Ackerman, Kenrick and Schaller, 2007; Barakzai and Shaw, 2018; Benenson, 2014; Jordan, Sommers, Bloom, & Rand, 2017; Klein & Milardo, 1993; Krems, Williams, Aktipis, & Kenrick, 2021; Krems, Williams, Merrie, Kenrick, & Aktipis, 2022; Owens, Shute, & Slee, 2000; Parker, Low, Walker, & Gamm, 2005; Shaw et al., 2017; Sugiyama, 2004).

This social complexity is also implied by adaptationist models of friendship, which emphasize friends as social insurance for times of illness, injury, or conflict. Consider a situation of drought and starvation; whereas strangers are unlikely to invest resources in a starving person, because that person looks like a bad bet for reciprocity, the starving person's friends might indeed share their finite resources with them even over others facing similar need (Tooby & Cosmides, 1996). By doing so, the sharing friend ensures the continued survival of a person who has a stake in their own welfare, and who would thus help them in their future times of trouble. Likewise, consider an agonistic conflict between Alex and Benji—both of whom are Cam's friends. According to the Alliance Hypothesis of Friendship, Cam should side with the friend who is more likely to take Cam's side in later conflicts, which ensures the continued survival of Cam's likeliest supporter (DeScioli & Kurzban, 2009; DeScioli, Kurzban, Koch, & Liben-Nowell, 2011).

Taken together, the embedded dyad framework and these adaptationist models imply that the benefits of friendship depend, in part, on how much one's friends value oneself relative to others, and thus that a friend's behavior toward others can influence one's own outcomes. For example, if Cam takes Alex's side in the above dispute, Cam is also siding against Benji. If people are affected by friends' behavior toward the self and toward others, then people should possess preferences for how friends behave toward the self and toward others. Specifically, insofar as friends generally radiate positive effects on the self—not only directly (via how they behave toward oneself) but also indirectly (via how they behave toward others)—people's friend preferences should be systematically predictable, such that (1) the friendship value of a target to the self is a function of the effects the target has (or is expected to have) on the self, both directly and also indirectly, and (2) the value of a trait in the target (e.g., the value of a friend's viciousness) to the self depends on the net effects (direct and/or indirect) that trait will have on the self, which is affected by toward whom that trait is directed. In other words, good friends should benefit or at least not hinder us, even as via their behavior toward other people.

This leads to predictions that extend, deviate from, and sometimes run counter to other accounts and intuition (e.g., people always eschew vicious friends). For example, imagine that your friend Amani demonstrates trustworthiness—but she does so by keeping in confidence your enemy's secret plan to harm you. Meanwhile, your friend Blanca demonstrates viciousness—but does so by deterring your enemy from harming you. As this example illustrates, in addition to the obvious and important (direct) effects that interactions with our friends can have on us, our friends' interactions with other people can also have major (indirect) effects on us as well. Note that, if you evaluate Amani and Blanca as friends via (a) intuition, (b) the inferences one might draw from canonical findings, or (c) cooperative accounts, you might conclude that Amani (trustworthy) is a better friend than Blanca (vicious). If you consider, however, the net (both direct and indirect) effects that Amani and Blanca have on your welfare, you would reach the opposite conclusion.

Previous work on friend preferences and cooperative accounts of partner choice generate multiple and sometimes competing predictions about how people will want ideal best friends to behave, as compared to our novel embedded dyad perspective.

First, work on friend preferences has, either explicitly or implicitly, addressed self-directed friend preferences—or how people want friends to behave toward them (see, e.g., Hall, 2012). We thus test whether (1) asking people how they want friends to behave (target-unspecified friend preferences) generates the same pattern of responses as asking how people want friends to behave toward them (self-directed friend preferences) (for a similar examination of mate preferences, see Lukaszewski & Roney, 2010). If unspecified friend preferences track self-directed (more than other-directed) friend preferences, this would be consistent with our argument that people prefer friends to behave differently toward the self versus others.

Second, and in line with canonical findings, we expect that people will generally prefer friends to be kind and trustworthy (but not vicious or indifferent)—both toward oneself and toward most others, as people tend to attach positive value to the welfare of other community members. Specifically, (2) for target-unspecified, self-directed, and neutral target-directed (i.e., stranger-directed) preferences, we predict that people will prefer friends to be kind and trustworthy, but not vicious, indifferent, or otherwise antisocial. Yet we also predict that (3) people will want friends to behave more prosocially toward oneself than toward others. After all, preferential prosociality may be part of the function of friends (e.g., DeScioli & Kurzban, 2009).

Further, the difference in preferred prosociality toward the self versus a stranger should be exaggerated when comparing preferences for how friends should behave toward oneself versus one's rivals. Indeed, people have rivalries and enemyships—relationships defined by competition and hatred (Adams, 2005; Holt, 1989)—which can harm one's outcomes (Aktipis et al., 2018; Günsoy, Cross, Uskul, Adams, & Gercek-Swing, 2015; Wiseman & Duck, 1995). Such relationships are often perceived as zero-sum; people believe they can be harmed when their enemies benefit (e.g., from others' kindness) and benefit when enemies are harmed (e.g., from others' viciousness; Aronson & Cope, 1968; Pietraszewski, 2016; Shaw, 2013). For a friend to achieve positive indirect effects on the self, that friend should not be kind toward one's enemy; rather, that friend should perhaps direct some degree of viciousness toward one's enemy. Thus, in some instances, we expect to see preferences for viciousness—counter to intuition, unlike in canonical findings, and at odds with cooperative account predictions. We test if (4a) people want friends who are appreciably more vicious toward one's enemies than they are toward oneself. We also test a stronger version of this prediction—(4b) that people prefer friends who are more vicious than kind toward one's enemies.4

We test these predictions in a two-wave study, in studies with varying designs, and in two nations. Methods were approved by university Institutional Review Board (IRB). All manipulations, focal (and exploratory) measures, and exclusions are noted. Data and syntax are on Open Science Framework (OSF) at https://osf.io/xeg48/.