Elsevier

Cognition

Volume 192, November 2019, 103994
Cognition

Original Articles
The rare preference effect: Statistical information influences social affiliation judgments

https://doi.org/10.1016/j.cognition.2019.06.006Get rights and content

Abstract

Shared preferences—liking the same things—facilitate and strengthen bonds between individuals. However, not all shared preferences are equally meaningful; sharing a rare preference with someone is often more exciting and meaningful than sharing a common preference. Here we present evidence for the rare preference effect: Participants chose to interact with (Experiment 1), and endorsed interactions between (Experiment 2), individuals who shared a rare preference, rather than those who shared a common preference, and this tendency increased with the relative rarity of the preference. While having a preference usually implies knowing and liking something, the presence of shared knowledge alone was sufficient to give rise to the rare preference effect (Experiments 3 & 4). Further, we find that social affiliation judgments are modulated by the causal process by which individuals came to have shared knowledge: Participants preferred to interact with someone who acquired a shared preference deliberately rather than accidentally (Experiment 5). In addition to the many cultural and emotional factors that drive mutual attraction, these results suggest that people’s decisions about with whom to interact are systematically influenced by the statistics of the social environment.

Introduction

Preferences reveal far more than what a person likes or dislikes. Preferences reflect one’s personal history and social context, including one’s age, gender, socioeconomic class, culture, political affiliation, and personality (Bonneville-Roussy et al., 2013, Carney et al., 2008, Rentfrow et al., 2011, Rozin and Siegal, 2003, Van Eijck, 2001). Perhaps because preferences reveal so much about a person, we are curious to learn about what others like and readily broadcast our own preferences to others. People often bring up their music preferences when becoming acquainted with a stranger (Rentfrow & Gosling, 2006) and form fairly accurate impressions of a stranger’s personality traits based solely on their music preferences (Rentfrow and Gosling, 2006, Rentfrow and Gosling, 2007).

Preferences also shape our interactions with others. In particular, we are drawn to people who like what we like: We delight in finding overlaps in preferences and bond with those who enjoy the same hobbies, read the same books, or root for the same sports teams as us. Evidence suggests that people evaluate others who share their music preferences more positively than those who do not (Boer et al., 2011, Lonsdale and North, 2009) and tend to have similar preferences as their friends (Selfhout et al., 2009, Werner and Parmelee, 1979). Preschool-aged children also prefer to approach and learn from people who like the same toys as they do (Fawcett and Markson, 2010a, Fawcett and Markson, 2010b), and even infants expect people who have similar preferences to interact positively with one another (Liberman, Kinzler, & Woodward, 2014). Thus, discovering a shared preference is more than just a pleasant coincidence; it provides an opportunity to assess our compatibility with potential social partners and guide our future relationships with them.

However, not all shared preferences are equally meaningful. In the movie Everyone says I love you, a love-struck Woody Allen deliberately presents himself as a New Yorker who vacations in Bora Bora, listens to Mahler’s 4th symphony, and admires the Italian painter Tintoretto, just to make Julia Roberts (who is also a New Yorker who likes all these things) fall for him. And he succeeds: She not only finds these coincidences fascinating but also finds him attractive. Woody Allen’s strategy is not just a clever way to woo someone; it also raises a deeper question about the social significance of discovering someone who likes the same things as you. Among the many things one might like, what makes some shared preferences more meaningful than others?

Indeed, there may be many factors that make a shared preference meaningful. Among these factors, here we present evidence for a key factor that systematically influences how people perceive, interpret, and use shared preferences: their rarity. In the movie, Julia Roberts may not have fallen for Woody Allen if she routinely ran into people who also share these preferences. You may also have experienced firsthand the thrill of meeting someone who shares a very rare preference with you. We refer to this intuition as the rare preference effect: All else being equal, sharing a rare preference with someone may be more meaningful than sharing a common preference.

The current study presents initial empirical evidence for the rare preference effect and investigates its cognitive underpinnings. Below, we motivate the hypothesis that the prevalence of preferences can guide adults’ judgments about whom to approach or befriend. We then provide an empirical demonstration of the rare preference effect (Experiments 1–2) and further investigate the representations that give rise to this effect (Experiments 3–5).

At its core, detecting and making social judgments based on the prevalence of preferences requires a sensitivity to the relative probability of events—that is, distinguishing events that happen with low frequency (rare events) from those that happen with higher frequency (common events) within a given population. Prior research suggests that the roots of this ability may be present early in life. Even infants distinguish differences between common and rare kinds of objects in a population (Xu & Garcia, 2008) and use this information to draw further inferences about the person engaged in the sampling process (Gweon et al., 2010, Kushnir et al., 2010).

Yet tracking the statistics of people’s preferences may pose unique challenges. Unlike the proportion of objects in a box, preferences are not directly observable and must be inferred from others’ choices and testimony (Jern et al., 2017, Lucas et al., 2014) or, more recently, from proxy measures such as Billboard charts and streaming counts. Prior work has demonstrated biases and errors in adults’ estimates of the prevalence of abstract properties such as beliefs, opinions, attitudes, and habits. Adults systematically distort the prevalence of their own attitudes and behaviors (Monin and Norton, 2003, Nisbett and Kunda, 1985, Ross et al., 1977, Suls and Wan, 1987, Suls et al., 1988) and ignore base-rate information in favor of descriptions of specific traits when making judgments about others (Kahneman & Tversky, 1973). Even within these studies, however, there is evidence that adults use statistical information appropriately. For example, adults do use base rates when they are the only information provided (Kahneman & Tversky, 1973). Further, while adults tend to inflate the prevalence of their own attitudes, their estimates nonetheless accurately distinguish commonly held attitudes from rarer ones (Nisbett & Kunda, 1985). Numerous studies have shown that adults can also use sparse data to accurately report the distribution of parameters of real-world events, such as their frequency and duration (Griffiths and Tenenbaum, 2006, Hasher and Zacks, 1984, Hertwig and Gigerenzer, 1999, Peterson and Beach, 1967).

Collectively, prior work suggests that, despite some biases, people’s statistical intuitions do reflect the overall structure of the world. People track the “social statistics” of their environment, representing not only the prevalence of concrete, observable events but also the prevalence of abstract, unobservable qualities such as people’s attitudes and preferences. Although these statistics often manifest as intuitions rather than as exact, formal estimates, they do reflect the relative frequencies and distributions of abstract occurrences.

Building on prior work on intuitive statistical reasoning, the current work focuses on how statistical information might affect social judgments. More specifically, we propose that people’s beliefs about the prevalence of preferences systematically influence their decisions about with whom they want to interact. Here we outline our three overarching goals.

Our first goal is to address the most basic question: Is the rare preference effect a robust, systematic phenomenon that actually stems from rarity? It is possible that this effect is an illusion that only exists in personal anecdotes or movies. Even if the phenomenon itself is real, the appeal of sharing a rare preference may stem from factors that are confounded with rarity. For example, rare preferences may be considered more socially desirable (see Monin and Norton, 2003, Suls et al., 1988 for related work on attitudes and behaviors) or may be more strongly held than common preferences. Thus, our initial goal is to provide empirical evidence of the rare preference effect. We first demonstrate this effect in a personally relevant and ecologically valid context, where participants report their own preferences and their intuitions about the prevalence of their preferences (Experiment 1, first-person judgments). Here, participants generate statistical information about the real-world prevalence by reconstructing it from their own knowledge and past experience. We then replicate the effect in a minimal, tightly controlled paradigm, where participants see visual displays that explicitly convey statistical information about the prevalence of preferences for novel items in a novel population (Experiment 2, third-person judgments). Together, these experiments provide converging evidence for the rare preference effect and identify a contribution of rarity that is distinct from other attributes that come from participants’ prior knowledge of real-world items (e.g., social desirability, strength).

Having demonstrated the presence of the rare preference effect, our second goal is to investigate the scope of this effect: Are people indiscriminately and superficially drawn to rare events, or does the rare preference effect instead reflect a sophisticated use of statistical information? In fact, the rare preference effect may be just one of many ways in which people use information about the prevalence of preferences. For instance, statistical information may also guide inferences about new individuals. Suppose you meet someone whose favorite artists are unknown; in order to maximize the chances of finding a shared preference, you might bring up an artist that is widely liked (e.g., da Vinci) rather than an obscure artist that is less likely to be recognized, let alone liked. Thus, we use third-person judgments (Experiments 2 and 4) to test whether participants use information about the prevalence of preferences flexibly to make a wide range of social judgments. Rather than indiscriminately preferring social partners who have rare preferences, people may prefer someone who has a more common preference or even ignore prevalence information altogether, depending on the context.

Our final goal is to better understand the representations that underlie the rare preference effect. When someone says “My favorite artist is Tintoretto”, you learn two things about this person: (1) she knows about the artist and his work, and (2) she enjoys and admires his paintings. That is, preferences provide information about both what people know (henceforth referred to as knowledge) and what they like (henceforth referred to as affinity1). Although it is possible to simply know about Tintoretto without necessarily liking or enjoying his work, or to find his paintings appealing without knowing anything about the artist, knowing and liking often go hand in hand.

Given that expressing a preference for something usually implies both knowledge and affinity, one possibility is that either of these properties is sufficient to give rise to the rare preference effect. Infrequent stimuli are seen as particularly salient and attention-grabbing (McCarthy et al., 1997, Sutton et al., 1965), and people tend to value rare or scarce items more than common items (Verhallen & Robben, 1994). Thus, all else being equal, people might not only prefer those who share rare knowledge with them, but also those who share rare affinities with them in the absence of prior knowledge. Yet preferences—a stable liking of some activities or items over others—are often the result of a complex chain of events. In order to come to like Tintoretto, one must have had experiences that led one to discover his work in the first place, perhaps by studying Renaissance art, living in Venice, or surrounding oneself with people who frequent art museums. Thus, preferences reflect various aspects of one’s cultural knowledge and past experiences. It is possible that the presence of shared knowledge (in the absence of explicit information about affinity) may be sufficient to give rise to the rare preference effect.

In fact, other people’s knowledge and affinity can provide qualitatively different information about them. Knowledge is often a reflection of one’s current interests, prior background, and social history. This may be especially true for relatively rare or obscure preferences for music, hobbies, or activities; someone who came to learn about Tintoretto presumably sought out particular kinds of artists or was close to people in that niche, whereas someone who came to know about da Vinci could have learned about the artist through many different channels. Thus, shared knowledge between two individuals, when it is rare, can be a good indicator of a broader, meaningful common ground (Clark, Schreuder, & Buttrick, 1983) or even signal the presence of a latent social group to which both individuals belong (Gershman, Pouncy, & Gweon, 2017). On the other hand, affinity—as operationalized here—is separated from a person’s prior background or social history. While knowledge in a particular domain implies an active effort to learn about it, affinity reflects a predisposition to find something attractive even without having deliberately sought it out. Although discovering a shared affinity with someone can be just as delightful as discovering shared knowledge, it may not necessarily support further inferences about shared cultural background or social history.

Prior work has found empirical support for the privileged status of shared knowledge over shared affinity in interpersonal relationships. Adults and children prefer people who share personally relevant beliefs with them over those who share arbitrary beliefs (Heiphetz, Spelke, & Banaji, 2014). Preschool-aged children prefer peers who share their knowledge (but not preferences) over peers who share their preferences (but not knowledge) (Soley & Spelke, 2016). Further, children selectively attribute shared knowledge—but not shared preferences—to members of the same cultural group (Soley & Aldan, 2018). These studies suggest that both adults and children readily represent different aspects of preferences that can be shared between individuals (i.e., prior knowledge and spontaneous affinity) and prioritize shared knowledge. These results provide indirect support for our hypothesis that the presence of shared knowledge between individuals may drive the rare preference effect more strongly than shared affinity in the absence of prior knowledge.

The present work explores how adults use statistical information about the prevalence of preferences to choose with whom they would rather interact. We first provide initial evidence for the existence of the rare preference effect (Experiments 1 & 2). We then test the hypothesis that this effect is driven more strongly by the presence of shared knowledge than by shared affinity (Experiments 3 & 4). Finally, Experiment 5 provides further support for the importance of shared background knowledge in people’s interpersonal decisions, by asking whether the means by which agents acquired a preference can influence people’s decisions, even when rarity is held constant. Our experiments provide converging evidence from two complementary paradigms. In one, we ask participants to provide statistical information about the prevalence of their own preferences (Experiments 1, 3, 5). In the other, we take advantage of a minimal, rigorously controlled context where we provide explicit statistical information about the prevalence of novel preferences in a novel population (Experiments 2, 4).2

Section snippets

Experiment 1

In Experiment 1, we asked participants to list their favorite bands, books, or movies and to estimate the prevalence of people’s preferences for these items. When given the choice between two potential social partners, we predicted that participants would favor an agent who shares a rarer preference with them over someone who shares a more common preference, and that this tendency would increase as the relative rarity of the preference increased.

Experiment 2

In Experiment 2, participants made third-party judgments where they introduced potential friends to a target agent, based on the agent’s preferences for novel items and on explicitly provided information about how prevalent preferences for these items are among a novel population. We manipulated the prevalence of preferences for each novel item (between subjects) and the preferences of the target agent (within subjects). In this task, participants only had access to information about the

Experiment 3

In Experiments 3 and 4, we further explored the nature of the representations that underlie the rare preference effect. We hypothesized that, although preferences provide information about what a person knows (knowledge) and what they like (affinity), the rare preference effect may be driven more strongly by shared knowledge than by shared affinity. In Experiment 3, we adapted the first-person paradigm in Experiment 1 to test whether participants prioritize shared rare knowledge (without an

Experiment 4

As above, we adapted the third-party paradigm in Experiment 2 to test whether participants use statistical information flexibly across contexts, and whether these effects manifest when agents share knowledge (without a stated preference) or affinity (without prior knowledge).

Experiment 5

In Experiment 5, we adapted the first-person paradigm in Experiments 1 & 3 to test whether participants are sensitive to the causal process by which people arrived at a preference. In this task, participants were given the choice between two agents who shared the same music preference with the participant. Critically, one agent arrived at the preference by deliberately seeking out similar songs; thus, their preference is potentially diagnostic of other, hidden similarities. By contrast, the

General discussion

The current work presents support for the hypothesis that people use statistical information about preferences to guide their social decisions, especially when such information signals the presence of a broader shared background. We first experimentally demonstrated the rare preference effect, confirming the intuition that discovering a shared preference with someone is more meaningful when it is rare than when it is relatively common. This effect is found in both first-person and third-person

Acknowledgments

This work was supported in part by a National Science Foundation Graduate Research Fellowship Program grant to NV and a Varieties of Understanding grant from the John Templeton Foundation and James S. McDonnell Scholar Award to HG.

References (47)

  • D. Boer et al.

    How shared preferences in music create bonds between people: Values as the missing link

    Personality and Social Psychology Bulletin

    (2011)
  • A. Bonneville-Roussy et al.

    Music through the ages: Trends in musical engagement and preferences from adolescence through middle adulthood

    Journal of Personality and Social Psychology

    (2013)
  • D. Byrne et al.

    Attraction as a linear function of proportion of positive reinforcements

    Journal of Personality and Social Psychology

    (1965)
  • D.R. Carney et al.

    The secret lives of liberals and conservatives: Personality profiles, interaction styles, and the things they leave behind

    Political Psychology

    (2008)
  • C.A. Fawcett et al.

    Children reason about shared preferences

    Developmental Psychology

    (2010)
  • S. Gershman et al.

    Learning the structure of social influence

    Cognitive Science

    (2017)
  • T.L. Griffiths et al.

    Optimal predictions in everyday cognition

    Psychological Science

    (2006)
  • H. Gweon et al.

    Infants consider both the sample and the sampling process in inductive generalization

    Proceedings of the National Academy of Sciences

    (2010)
  • L. Hasher et al.

    Automatic processing of fundamental information: The case of frequency of occurrence

    American Psychologist

    (1984)
  • L. Heiphetz et al.

    The formation of belief-based social preferences

    Social Cognition

    (2014)
  • R. Hertwig et al.

    The ‘conjunction fallacy’ revisited: How intelligent inferences look like reasoning errors

    Journal of Behavioral Decision Making

    (1999)
  • D. Kahneman et al.

    On the psychology of prediction

    Psychological Review

    (1973)
  • H. Kim et al.

    Deviance or uniqueness, harmony or conformity?: A cultural analysis

    Journal of Personality and Social Psychology

    (1999)
  • Cited by (15)

    • Interpersonal utility and children's social inferences from shared preferences

      2023, Cognition
      Citation Excerpt :

      When inferring friendships, 4-year-olds are more likely to select people who are similar in meaningful ways than people who have an arbitrary or coincidental similarity (e.g. it snowed on both of their birthdays; Afshordi, 2019). In adulthood, shared preferences are also seen as a weaker sign of affiliation when they are common, compared with when they are rare (Vélez, Bridgers, & Gweon, 2019). By one cognitive account of these data, what children may be doing when they observe similar behavior is either deploying a homophily-based rule, or deciding not to deploy this rule.

    • Attitude similarity and interpersonal liking: A dominance of positive over negative attitudes

      2022, Journal of Experimental Social Psychology
      Citation Excerpt :

      Of course, our findings also do not imply that shared likes bond stronger than shared dislikes in all social contexts. For example, recent research has found that the similarity effect in interpersonal attraction is stronger for rare shared attitudes than for common shared attitudes (Alves, 2018; Vélez et al., 2019). Accordingly, people are especially attracted to others who share those attitudes that are not shared by many other people.

    • The social meaning of common knowledge across development

      2021, Cognition
      Citation Excerpt :

      For instance, around the age of 5, children correctly infer what knowledge can and cannot be self-acquired and what knowledge is easy or difficult to acquire (Lockhart et al., 2016), and expect generic factual knowledge to be shared more widely in comparison to non-generic knowledge (Cimpian & Scott, 2012). Further, children use group membership cues to make inferences about others' cultural knowledge, but not about their knowledge of generic facts (Soley, 2019) and adults make stronger affiliative inferences about others when the knowledge they share is relatively rare (Velez et al., 2019). These findings raise the possibility that different kinds of knowledge vary in their social significance for children as well as adults, such that they might use knowledge selectively as an indicator of one's social group membership.

    • Understanding preferences in infancy

      2023, Wiley Interdisciplinary Reviews: Cognitive Science
    View all citing articles on Scopus
    View full text