We preferably copy the behavior of our own group’s members, even if group the group is arbitrary and novel, and even if one's peers are less competent than others

Copy the In-group: Group Membership Trumps Perceived Reliability, Warmth, and Competence in a Social-Learning Task. Marcel Montrey, Thomas R. Shultz. Psychological Science, December 23, 2021. https://doi.org/10.1177/09567976211032224

Abstract: Surprisingly little is known about how social groups influence social learning. Although several studies have shown that people prefer to copy in-group members, these studies have failed to resolve whether group membership genuinely affects who is copied or whether group membership merely correlates with other known factors, such as similarity and familiarity. Using the minimal-group paradigm, we disentangled these effects in an online social-learning game. In a sample of 540 adults, we found a robust in-group-copying bias that (a) was bolstered by a preference for observing in-group members; (b) overrode perceived reliability, warmth, and competence; (c) grew stronger when social information was scarce; and (d) even caused cultural divergence between intermixed groups. These results suggest that people genuinely employ a copy-the-in-group social-learning strategy, which could help explain how inefficient behaviors spread through social learning and how humans maintain the cultural diversity needed for cumulative cultural evolution.

Keywords: cumulative cultural evolution, intergroup dynamics, social-learning strategies, transmission biases, open data, preregistered

Discussion

Although previous studies have found an apparent in-group bias in social learning, they have failed to resolve whether this constitutes a genuine social-learning strategy or a mere confluence of other factors (Buttelmann et al., 2013; Howard et al., 2015). Our study disentangled group membership from similarity and familiarity by assigning group membership at random. We found that rather than eliminating the preference for in-group members, this approach resulted in a robust in-group-copying bias, which (a) was bolstered by a tendency to observe in-group members, (b) overrode participants’ stated beliefs, (c) grew stronger when social information was scarce, and (d) even caused cultural divergence between intermixed groups. Taken together, our findings suggest that people genuinely employ a copy-the-in-group strategy and that group membership has both a direct and indirect effect on copying.

Why might a copy-the-in-group strategy have evolved in the first place? One reason could be that it allowed humans to rapidly adopt and vigorously maintain group norms that enhance coordination (McElreath et al., 2003) or promote cooperation (Boyd & Richerson, 2009). Another reason could be that social learning is useful only to the extent that adopting other people’s behavior yields similar payoffs (Laland, 2004). For example, copying out-group members could be less efficient or even counterproductive if groups differ in terms of what behavior is punished or rewarded. Finally, such a strategy could also have evolved because it minimized the risk of deception. Because social learning is essentially information scrounging (Kameda & Nakanishi, 2002), in which the copier benefits from other people’s knowledge without incurring the same costs, knowledgeable individuals have an incentive to mislead others. However, this incentive is minimized when observed individuals have a vested interest in the copier’s success. This holds true in kin relationships (Laland, 2004) and likely generalizes to other settings, such as intergroup competition.

Although the in-group-copying bias may be adaptive, we were careful to control for factors that could justify such a bias or evoke a similar-looking one. First, we limited the role of similarity and familiarity by assigning participants to groups at random. Second, we ensured that there was no inherent advantage to copying either group by giving participants access to the same information and by discouraging deception. Third, we deterred intergroup competition by framing the game in terms of individual performance and by obscuring group membership whenever participants saw others’ scores. Fourth, we prevented selective copying from advancing social goals (Over & Carpenter, 2012), such as inclusion (Watson-Jones et al., 2016), by limiting social interactions to unidirectional observations of previous participants. Finally, we matched everyone with an equal number of participants from each group to ensure that fellow in-group members did not form a majority. That being said, because cultures could conceivably vary in their reliance on group membership, the generalizability of our findings may be limited by our sample consisting solely of U.S. residents. Furthermore, although online recruitment platforms often yield a more diverse pool of participants than traditional student samples, they raise other concerns about data quality. For a brief overview of these concerns and our analysis of their impact, see “Data Quality” in Supplemental Information About Data Quality, Preregistration, and Other Results.

In addition to controlling for potential confounds, we also tested several proposed explanations for why people may preferentially copy in-group members. One is that we attend to them more often (Kinzler et al., 2011). This seems sensible, given that native-language speakers attract more attention from a very young age (Kinzler et al., 2007) and that adults bias their attention toward in-group members, whether that group is preexisting or novel (Kawakami et al., 2014). However, developmental studies have failed to reveal any obvious in-group attentional bias in social learning (Oláh et al., 2016), and both looking times (Buttelmann et al., 2013; Pető et al., 2018) and eye-tracking data (Howard et al., 2015) contradict this hypothesis. It is therefore curious that most participants in our experiments (70% and 67%) observed the in-group more often and that this drove much of the in-group bias in copying. One possibility is that humans become more likely to attend to in-group members over the course of development. For example, instead of attention guiding children’s copying, biases in copying could shape their attention. Another possibility is that attention played a greater role in our study because participants explicitly chose whom to observe and paid a clear opportunity cost for each selection. Our findings may thus be particularly relevant to contexts in which people have strong and explicit control over their sources of social information, such as on social media.

Another proposed explanation for why people may prefer to copy in-group members is that we ascribe greater competence to the in-group (Kinzler et al., 2011) or view out-group members as less reliable (Oláh et al., 2016). This hypothesis is rooted in the classic social-psychology finding that in-group members are evaluated more favorably, even when groups are arbitrary and novel (Brewer, 1979). Indeed, we found that the in-group enjoyed advantages in perceived reliability, warmth, and competence. However, none of these beliefs had much effect on whom participants copied. On the contrary, because the in-group-copying bias arose even among participants who viewed the out-group as more competent, our results seemingly contradict the axiom that social-learning strategies exist solely to help people identify and adopt the most effective behavior (Kendal et al., 2018).

In fact, if people are predisposed to copy in-group members, perhaps even when their perceived competence is low, this could help explain the spread of inefficient or even deleterious behaviors. For example, opposition to vaccination is often disseminated through highly clustered and enclosed online communities (Yuan & Crooks, 2018) who use in-group-focused language (Mitra et al., 2016). Likewise, fake news tends to spread among politically aligned individuals (Grinberg et al., 2019), and the most effective puppet accounts prefer to portray themselves as in-group members rather than as knowledgeable experts (Xia et al., 2019). Our research also sheds light on why social media platforms seem especially prone to spreading misinformation. By offering such fine-grained control over whom users observe, these platforms may spur the creation of homogeneous social networks, in which individuals are more inclined to copy others because they belong to the same social group.

Finally, the fact that the in-group-copying bias produced some amount of cultural divergence in both of our experiments is of particular interest from a cultural evolutionary point of view. This is because the exceptional complexity of human culture and technology (Montrey & Shultz, 2020) likely depends on integrating and recombining diverse cultural traits (Mesoudi & Thornton, 2018). Theory suggests that for cultural evolution to be cumulative (i.e., for complexity to increase over time), populations may have to be fragmented to some degree so that unique traits have the opportunity to flourish (Derex et al., 2018). Otherwise, cultural traits may homogenize, leaving learners with little to recombine. Although the intergroup differences in behavior we observed do not rise to the level of cultural traditions, our findings could help explain how cultural differences persist in intermixed groups (McElreath, 2004). A copy-the-in-group strategy could thus be one mechanism for achieving the cultural diversity needed for cumulative cultural evolution.