Four-year-old Children Align their Preferences with those of their Peers

Children express preferences for a wide range of options, such as objects, and frequently observe the preferences that others express towards these things. However, little is know about how these initial preferences develop. The present research investigated whether one particular type of social information – other children’s preferences – influences children’s own preferences. Four-year-old children observed, via video, two boys and two girls display the same preference for one of two stickers. Each child (peer) expressed liking for one sticker and dislike for the other. Then children completed two rounds of the Dictator Game, a classic resource distribution task. In each round, children distributed either 10 liked stickers or 10 disliked stickers (counterbalanced) between themselves and another child who was not present. If the preferences expressed by their peers influenced children’s own preferences, children should keep more of the liked than disliked stickers for themselves. In line with this prediction, more children kept more liked than disliked stickers, indicating their distribution patterns were influenced by their peers’ preferences. This finding suggests that children extracted informational content about the value of the stickers from their peers and used that information to guide their own preferences. Children might also have aligned their preferences with those of their peers to facilitate social bonding and group membership. This research demonstrates the strong influence of peers on children’s developing preferences, and reveals the effect of peer influence via video – a medium that young children are frequently exposed to but often struggle to learn from in other contexts.


Introduction
It is no secret that children are constantly learning from those around them, and that much of this learning occurs outside of formal teaching contexts. Any parent who has had the privilege of their child spouting a random curse word and wondering, "where did she learn that?" can attest to this. Indeed, there is a burgeoning literature detailing the subtle ways in which children acquire information, from eavesdropping (e.g., Floor & Akhtar, 2006) to observing the preferences of others (e.g., Hennefield & Markson, 2016). However, children are not indiscriminate social learners. Instead, they consider such factors as past accuracy (e.g., Birch, Vauthier, & Bloom, 2008;Jaswal & Neely, 2006;Koenig & Harris, 2005), expertise (e.g., Keil, Stein, Webb, Billings, & Rozenblit, 2008), and intention (e.g., Butler & Markman, 2012), when determining from whom to learn. Furthermore, there is reason to believe that, in addition to language, preferences might be particularly well suited to social learning. Shared preferences -whether for music, activities, or food -facilitate social bonding and group membership. Preferences can also convey information about the relative value of one option over another. Thus, as the present study will investigate, it is possible that preschoolers' incorporate one particular type of social information -other children's preferences -into their own developing preferences.
One strategy for selectively acquiring culturally transmitted information and strengthening social bonds is to copy the majority. There is emerging evidence that, in the absence of prior knowledge, children expect the behavior exhibited by a majority of individuals to be more reliable than that of a minority. For example, when faced with a group of adults labeling the same object, 3-yearold children tend to side with a majority of two or three over a lone dissenter, sometimes even developing distrust for the dissenter (Corriveau, Fusaro, & Harris, 2009). Even 2-year-old children are more likely to copy the functional strategy used by three peers over one used by a single peer (Haun, Rekers, & Tomasello, 2012). Thus, in situations with at least one objectively correct response, children seem to learn from statistical information available about others' actions and align their behavior with the majority. These findings support the possibility that children might also glean information from observing Hennefield and Markson: Peer Influence on Children's Preferences Art. 14, page 2 of 8 others' preferences, and align their own preferences with that of the majority.
In contrast to factual or functional knowledge, preferences are inherently subjective. Preferences could convey information about others' idiosyncratic likes and dislikes (e.g., she likes blocks and dislikes rocks) or about the value of options themselves (e.g., blocks are more valuable than rocks). It is thus remains an open question whether young children differentiate between objective and subjective information when learning from others. On one hand, the subjectivity of preferences might render it socially acceptable to hold and maintain a preference that is different from the majority. On the other hand, subjective information might be more malleable than objective information, potentially making preference information a stronger candidate for social influences. Further, it is plausible that young children might be most strongly influenced by other children -their peers -when it comes to subjective information. Indeed, empirical findings suggest that children consider the content of the material, for example, if they need information about toys or food, when determining whether a child or adult is a better source of information (VanderBorght & Jaswal, 2009). In addition, children appear to engage in behaviors that strengthen social connections and group membership with peers, such as conformity (Haun & Tomasello, 2011) and reputation management (Engelmann, Herrmann, & Tomasello, 2012;Leimgruber, Shaw, Santos, & Olson, 2012). Thus, in contrast to utilizing adults' general expertise and authority to acquire factual information, children might selectively learn from the preferences of their peers.
Children appear sensitive to others' expression of their preferences early in life. By 18 months of age children understand that preferences are subjective (Graham, Stock, & Henderson, 2006;Repacholi & Gopnik, 1997), and by 3 years recognize when others share their own preference (Fawcett & Markson, 2010). Further, there is emerging evidence that children use others' preferences to acquire information about the relative value of options. When given the choice between two options, toddlers choose to play with the same objects for which adults have demonstrated a preference (Fawcett, Kushnir, & Markson, unpublished manuscript), and preschoolers pick toys and activities that were preferred by children over adults, and by individuals of the same gender as themselves (Shutts, Banaji, & Spelke, 2009). Preschoolers also avoid options that an adult does not choose, suggesting they devalue those options (Hennefield & Markson, 2016). However, it is not yet known whether children incorporate information acquired via the preferences of their peers into their own valuation of options and subsequent preferences.
Whereas previous studies on preferences have traditionally used binary forced-choice measures to determine which of two options a child prefers, these measures are limited in their ability to generate fine-grained distinctions in behavior. Resource distribution tasks like the Dictator Game (DG) have been used to examine fairness and prosocial behavior in young children (Aguilar-Pardo, Martínez-Arias, & Colmenares, 2013; Benenson, Pascoe, & Radmore, 2007;Benozio & Diesendruck, 2015;Blake & Rand, 2010;Chen, Zhu, & Chen, 2013;Gummerum, Hanoch, Keller, Parsons, & Hummel, 2010;Kogut, 2012;Lucas, Wagner, & Chow, 2008;Ongley & Malti, 2014) and typically generate greater variation in responses than binary forced-choice measures. Blake and Rand (2010) have also successfully used the DG to demonstrate that children use their own a priori explicitly stated preferences to guide how they distribute resources. Three-to 6-year-old children first identified their favorite and least-favorite stickers, and were then given either 10 of their favorite or 10 of their least-favorite stickers to distribute between themselves and another child. Children gave fewer of their favorite stickers to the other child, suggesting they assigned a higher value to their favorites stickers and found giving them away to be more costly. In contrast, children gave more of their leastfavorite stickers, indicating a willingness to share provided the cost was not too high. Further, when assessing distribution patterns for children's favorite and least-favorite stickers, Blake and Rand found that the majority of 4-year-old children distributed at least one sticker, and the proportion of favorite and least-favorite stickers distributed did not differ from that of older children. This finding suggests that the DG is a feasible measure for assessing value discrimination in 4-year-old children.
The present study used the Dictator Game to test whether 4-year-old children consider the preferences of their peers in their valuation of options by testing whether they differentially distribute two different resources -one their peers liked and one their peers disliked. Children watched a video in which four peers sequentially demonstrated the same preferences; each liked one specific sticker ("liked sticker") and disliked the other ("disliked sticker"). Children then played two rounds of the DG; one round with 10 liked stickers and one with 10 disliked stickers. If their peers' preferences influence children's own preferences, children should keep more liked than disliked stickers. Such a finding would suggest that children extracted informational content about the value of the stickers from their peers and used that to guide their own preferences. Alternatively, it is possible that children did not extract informational content from their peers' preferences, that this information did not influence their own preferences, or their distribution decisions were based on a priori aesthetic or idiosyncratic preference. If any of these cases, children should not differentially keep more liked than disliked stickers.

Method Participants
The study procedure was approved by the Institutional Review Board at Washington University in St Louis. Parental consent was obtained in advance from each participating child. Seventy-two 4-year-old children participated in either an Experimental (N = 48, M = 4; 7, Range = 4; 0-5; 0, 24 girls) or a Baseline Condition (N = 24, M = 4; 7, Range = 4; 1-5; 0, 12 girls). One child was replaced for failing to understand the study instructions. Children were recruited from a database of families that reside in a city in the Midwestern United States. The majority of children were white and from middle-class backgrounds.

Materials
Square stickers, 22 mm × 22 mm, were printed with one of two blue designs on a white background. The abstract designs, a swirl and snowflake, were chosen to be equally interesting to children, yet distinct (see Figure 1). Materials for each child consisted of 20 stickers (10 swirl, 10 snowflake), a felt game board, four envelopes, and a privacy screen to shield the child from the experimenter during test.
Social Influence Manipulation (SIM): Short video clips (11.6s each) in which four actors (two boys, two girls, roughly the same age and from the same community as the participants) demonstrated that they liked one sticker and disliked another sticker. For each clip, one actor picked up each sticker in turn and examined it while expressing a preference. They demonstrated liking by saying in an excited tone, "Oh, cool, I really like this one!" and disliking by saying in a negative tone, "Oh, no, I don't like this one". The "stickers" in the actors' hands were blank squares of paper that were positioned in front of the actor and then lifted in such a way that the front of the square was not visible to the camera, and thus impossible for the child to view. As the actor was expressing their first preference, a picture of the sticker they were "looking at" appeared onscreen, to the right of the actor, and remained visible for 4s while they spoke. Then that picture disappeared and a picture of the second sticker appeared onscreen to the left of the actor, and remained visible for 4s while they expressed the complementary preference. Thus, the same preference expressions (i.e., each instance of a preference being expressed) were counterbalanced to pair with each specific sticker, and all children viewed all the same preference expressions. This ensured that the actors own preferences could not influence their preference expressions, and had the added benefit of equating length of time allotted to each preference expression in the video. The order in which the actors appeared onscreen was counterbalanced with the constraint that two actors of the same gender did not appear in succession.

Procedure
Children in the Experimental Condition watched the SIM and then played the DG; children in the Baseline Condition only played the DG without first watching the SIM. All children played two rounds of the DG, one round with each sticker design (order counter-balanced). The Baseline Condition was included to test for differences between children who had watched the SIM and those who had not been exposed to preferences for the different stickers. To begin the DG, the child was seated across a table from the experimenter with the felt board between them. The experimenter placed 10 stickers (10 swirls or 10 snowflakes) in a circle on the board, and told the child that all the stickers belonged to the child and they would play a game with them. She placed one envelope to the right of the board and said, "Any stickers you want to keep to take home should go in this envelope". Then she placed another envelope to the left of the board and said, "Tomorrow there is another girl/boy coming here to play. She/he is just like you. Any stickers you want to leave for her/him should go in this envelope". The gender of the other child was matched to the child's own gender. Then the experimenter explained that when she placed the privacy screen on the table no one could see which envelopes the child put the stickers into. She verified that the child remembered the purpose of each envelope and the screen; if they did not she reminded them and asked again. Then she placed the screen on the table between herself and the board and told the child to "Go ahead and put all the stickers away".
After the child finished, the experimenter again verified that they remembered to whom each envelope belonged. One child failed this identification and was replaced. Then the child played the second round with 10 stickers of the other design. The only difference between the two rounds was that in the second round the child was told that the child they could elect to leave stickers for was a different child from the child in the first round (e.g., "another girl/boy is coming tomorrow"). Finally, in the Experimental Condition, after completing both rounds of the DG the experimenter placed one sticker of each design in front of the child and asked them to identify which sticker the kids in the video had liked. All children responded correctly.

Results
If children incorporate the preferences expressed by their peers into their own preferences, then they should more liked than disliked stickers. The findings support this hypothesis (see Figure 2). Twenty-one of the 48 children in the Experimental Condition kept more liked than disliked stickers, whereas only 7 children kept more disliked than liked stickers. This distribution is significantly different from what would be expected by chance, χ 2 (1, N = 28) = 7, p = .008. Of the remaining 20 children, 14 children kept an equal number of liked and disliked stickers, and 6 kept all the stickers. If a category for the 14 children who kept an equal number of stickers is added to the above analysis, the distribution remains significantly different from chance, χ 2 (2, N = 42) = 7, p = .03. It is unclear how to classify the 6 children who kept all the stickers. Indeed, some researchers differentiate between prosocial (distributing at least one) and non-prosocial (keeping all) children, and exclude those non-prosocial children from subsequent analyses (e.g., Blake & Rand, 2010;Kogut, 2012). Whereas it is possible that some children are not inclined to give regardless of the value of the resources, it is also possible that these 6 children did not differentially value the stickers, or that keeping all the stickers masked any differentiation they did make. However, even if these 6 children are conservatively added to the 14 children who kept an equal number of stickers, the distribution remains significantly different from chance, χ 2 (2, N = 48) = 7.625, p = .022. There were no significant differences or interactions with regard to resource distribution as a function of round order. Overall, these analyses indicate that children were influenced by their peers' preferences such that more children systematically kept more stickers liked by their peers than disliked, and this finding persists even when conservatively including all children (i.e., those who did not distribute any stickers) in the analyses.
Of the 24 children in the Baseline Condition, only 8 children kept a different number of stickers between the two rounds, 10 children kept an equal number, and 6 children kept all the stickers. Further analyses revealed that children's distribution patterns in the Experimental Condition differed from their distribution patterns in the Baseline Condition. First, the number of children who kept a different quantity of stickers across both rounds (Experimental = 28, Baseline = 8) was compared to the number of children who kept an equal quantity (or all) stickers across both rounds (Experimental = 20, Baseline = 16). These distributions are marginally different from each other, Fisher's Exact p = .079, suggesting that children who viewed the SIM were marginally more likely to distribute different quantities of stickers in the two subsequent rounds of the DG than children who had not viewed the SIM. Further, in the Baseline Condition, all 8 children who kept a different quantity of stickers kept more in the second round than the first round. However, in the Experimental Condition, 14 children kept more stickers in the first round (67%, 11 were children who distributed the liked-stickers first). That finding suggests that the SIM overrode children's tendencies to keep more stickers as the game progressed.
There were no gender differences or differences as a function of sticker type in either condition. Interestingly, whereas there is clear evidence that children differentiated between the liked and disliked stickers in their distribution patterns, this difference is only weakly reflected in the mean number of stickers they kept. A Wilcoxon Signed-Rank test indicates that, overall, children kept marginally more liked (M = 6.5) than disliked (M = 6.1) stickers, (Z = -1.799, p = .072). In the Baseline Condition children kept an average of 6.75 stickers, which is not significantly different from the overall quantities of either the liked or disliked stickers. Further, the modal number of stickers kept in all conditions was 5 (31.25% of all distributions), indicating that many children in the present sample possessed a strong desire to distribute resources equally. This preference for equal distribution is commonly observed in third party resource distribution tasks where children divide resources between two recipients (e.g., Kenward & Dahl, 2011;Olson & Spelke, 2008), but is often less robust in the DG (e.g., Benenson et al., 2007;Blake & Rand, 2010). Thus, the main finding that more children kept more liked than disliked stickers is especially notable because it indicates that peers' preferences play a significant yet subtle role in influencing children's behavior.

General Discussion
The present study investigated whether other children's preferences would influence 4-year-old children's own preferences. After viewing four peers express the same preference for one of two types of stickers, more children kept more of the stickers liked by their peers than stickers disliked by them. Importantly, this pattern was not due to children holding an a priori preference for one of the stickers or an artifact of playing multiple rounds of the Dictator Game. Rather, children's resource distribution was influenced by the preferences of their peers, likely via children's extraction of informational content about the value of the stickers. Children's responses could have also been driven by the social dynamics of the experimental context, either independently or in conjunction with value assignment. Thus, peers' preferences might influence one's understanding of what others value, and children could have used this information to provide a socially appropriate response. Prior research shows that by 4 years of age children will publically conform to their peers, even when they know their peers are wrong (Haun & Tomasello, 2011), and can strategically manage their reputations, sharing more with recipients who could reciprocate later (Engelmann, Over, Herrmann, & Tomasello, 2013;Kenward, Hellmer, Winter, & Eriksson, 2015). However, if children's responses in the present study were primarily guided by reputational concerns, then children should have given away more liked than disliked stickers to curry favor with the recipients. Children could also have considered the disliked stickers to be undesirable, and thus considered not giving them away to be prosocial. However, if children believed the disliked stickers to be truly undesirable to others, then they should have kept more disliked than liked stickers as they would not benefit from giving any disliked stickers away. Thus, whereas children might have considered their own reputation and the preferences of the recipient when determining how many resources to distribute, their assignment of different values to the stickers (based on peer preferences) likely drove children's resource distribution in the present study.
The finding that children aligned their preferences with those of their peers raises several additional questions. The first concerns whether children's preferences were influenced by their observation of their peers expressing liking or dislike, or the contrast between both types of information. The DG could have offered insight into this question if children's distribution patterns for the liked or disliked stickers differed significantly from either the Baseline Condition or norms typically reported in the literature. However, the differences in children's distribution patterns in the Experimental Condition were only weakly reflected in the mean number of stickers they kept. Furthermore, in general, children in the DG tend to keep 65-75% of the resources for themselves, which is similar to the distribution patterns reported in the extant adult literature (see Engel, 2011, for a review). Younger children also tend to keep even more resources than older children (e.g. Benenson et al., 2007), and younger children are found to keep all of the resources more frequently than older children (Blake & Rand, 2010;Gummerum et al., 2010;Ongley & Malti, 2014). However, in the present study few children kept all the stickers, and the most common distribution pattern was to divide the stickers equally between themselves and the other child. It is not clear whether these patterns were an artifact of the methodology or the sample, but it does suggest that the present study might not have fully captured children's preferences toward the options. To address this limitation, future studies could employ complementary resource distribution tasks that disentangle children's desire to share resources from their valuation of those resources, and include children from diverse socioeconomic backgrounds and cultures. Those methods would serve to elucidate the scope and magnitude of peer influence, including key questions about whether expressions of like and dislike are equally influential.
The distinction between liked and disliked information is important because there is considerable evidence that both adults and children use positive and negative information in substantially different ways to reason about the world (e.g., Rozin & Royzman, 2001). Specifically, this difference manifests in a negativity bias in which negative information is learned and used to a greater extent than positive information (see Vaish, Grossmann, & Woodward, 2008, for a review). If the negativity bias extends to preference information, children should avoid disliked options to a greater extent than they increase their affinity toward liked options. In contrast, if children consider preferences to be accurate, reliable indicators of value and/or social norms, they should equally avoid disliked options and increase their affinity toward liked options. As noted above, future research is needed to disentangle these two possibilities.
A second question concerns the consistency and quantity of input needed for children to acquire information from observing others' preferences. In the present study children received consistent information from all four peers. There is substantial evidence that young children readily use statistical information to inform their learning, including to infer others' preferences (Kushnir, Xu, & Wellman, 2010;Ma & Xu, 2011). However, whether children attended to the internal consistency of the preference information (i.e., within-individual consistency) or the proportion of individuals who expressed a particular preference (i.e., between-individual consistency) is not known. Further, classic work on social influence and conformity has revealed that a minimum of three informants is necessary to elicit conformist behavior in adults (e.g., Asch, 1956). Four informants were used in the present study to equate gender, however it is unclear whether children needed to view four different peers in order to be influenced by their peers' preferences. In addition, because the peers in the present study were strangers, children it is likely that children could have been more attentive to the quantity and consistency of the preferences than if observing familiar peers. Future studies are needed to clarify the impact of informant attributes on children's option valuations and preferences.
In the present study, it is highly unlikely that children had an a priori preference for one of the stickers due to unfamiliar design of both stickers. Indeed, the novel design on the sticker the actor was looking at did not appear on screen until after the actor had picked up the sticker and begun to express their preference. This design intentionally minimized the time in which children were given to form a preference for one of the stickers based on something other than the experimenters' preference. Thus, whereas the present findings demonstrate that children develop new preferences based on the preferences of their peers, another interesting question is whether children alter a priori preferences to align with those of their peers. We expect, in line with the above predictions, that the quantity, quality, and consistency of information provided by peers, in conjunction with more specific attributes of the peers (e.g., friend vs. stranger), and also the strength of the child's preexisting preference, will all be relevant factors in determining the contexts in which children alter a priori preferences to align with their peers.
A final point worth noting is the finding that children aligned their preferences with those of peers when they viewed these peer preferences via video. There is substantial evidence for a "video deficit" in early childhood -that is, infants and toddlers have more difficulty learning information from video compared to when that same information is presented live (e.g., Troseth, Saylor, & Archer, 2006;Troseth & DeLoache, 1998). Although there is evidence that by three years of age children can learn some information solely from video (e.g., Roseberry, Hirsh-Pasek, Parish-Morris, & Golinkoff, 2009), learning is enhanced when the videos are combined with social engagement (Nussenbaum & Asmo, 2016) and interactive parent-facilitated questions about the programming (Strouse, O'Doherty, & Troseth, 2013). As neither of these techniques were utilized in the present study, it is thus likely that the current findings underestimate the extent to which preschool-age children incorporate their peers' preferences into their valuations. Further, children's lack of familiarity with both the peers in the video and the peer recipients of the stickers increases the likelihood of underestimating the extent to which children align their preferences with those of their peers in trying to facilitate social bonding and group membership.
Social learning, including social transfer of preference, is widespread throughout the animal kingdom, and includes both imitative and non-imitative learning from conspecifics. For example, social transfer of food preferences is widely established in non-human animals and is typically characterized by animals approaching a food that a conspecific previously ate, or avoiding a food that made a conspecific ill (e.g., Galef, 1977). Applied to the present study, this type of approach and avoidance behavior could result in children approaching (via assigning a higher value to) the sticker liked by their peers, and avoiding the sticker disliked by their peers. Although the value transmission that occurred in the present study was robust, because children assigned that value based on the observation of their peers' preferences and that value was then used to inform their decisions in a distinct resource distribution task, we abstain from making any claims about the specific social learning mechanism(s) that drive this behavior.
There is a wealth of timely and pertinent information about option values and social norms available to children via the social domain, and strategically extracting and utilizing such information could yield adaptive advantages. The present research indicates that young children use social information -specifically, peers' preferences -to inform their relative valuation of options and subsequent preferences. Research on children's learning in social contexts has primarily focused on information that children learn from adults (e.g., Sobel & Kushnir, 2013). However, peers comprise a large and essential component of children's social networks, and understanding what information children are learning from their peers is crucial. The present study offers initial insights into the influence of peers on children's developing preferences. The findings suggest that, indeed, children do consider their peers' preferences to provide value-laden information. Further, whereas peers' preferences might denote universally agreed upon values, another possibility is that such preferences inform children of culture-specific values (e.g., Csibra & Gergely, 2011). Thus, acquiring information about the value of options via social contexts might also serve to facilitate the cultural transmission of information and strengthen social connections. That children can learn the value of a resource from one group of peers and use that information to facilitate an interaction with a new peer supports this possibility, and underscores the need to consider the complexities of children's social spheres in accounts of preference development. By revealing the effects of peer influence via video, the present study highlights the consequential influence of peers on children's developing preferences.

Data Accessibility Statement
Data from this manuscript is available via Washington University's Open Scholarship data repository: DOI: https:// doi.org/10.7936/K7KP810V