Elsevier

Games and Economic Behavior

Volume 107, January 2018, Pages 253-281
Games and Economic Behavior

Will any gossip do? Gossip does not need to be perfectly accurate to promote trust

https://doi.org/10.1016/j.geb.2017.09.015Get rights and content

Highlights

  • We study how gossip accuracy affects behaviour in a population of trust games.

  • We observe non-trivial levels of spontaneous inaccuracy in gossip.

  • This inaccuracy was largely due to gossipers' desire to punish untrustworthy players.

  • Endogenous inaccuracy did not adversely affect levels of trust and trustworthiness.

  • Even inaccurate gossip induces a degree of reputational concern in gossip targets.

Abstract

The fact that gossip can be inaccurate, intentionally or otherwise, has led to questions over its ability to build cooperation in large societies. We explore the impact of gossip accuracy on trust and trustworthiness in a population playing decentralized, two-player trust games. We observed non-trivial levels of spontaneous inaccuracy in gossip, and there was evidence that this was largely due to gossipers' desire to punish untrustworthy players. Although this endogenous inaccuracy did not adversely affect levels of trust and trustworthiness, introducing high levels of exogenous inaccuracy did. Importantly though, we observed greater trust and trustworthiness when highly inaccurate gossip was present than when communication was impossible. This suggests that even inaccurate gossip induces a degree of reputational concern in gossip targets and some willingness among gossip recipients to discriminate between partners on the basis of the gossip they received. Thus, gossip need not be perfectly accurate to effectively induce cooperation.

Introduction

One of the oldest problems facing organizations and societies has been how to induce good behaviour from their members or citizens, particularly when economic exchange occurs in large and/or geographically dispersed populations. In these circumstances, repeated interactions are rare and it is very costly for any one individual to observe and keep track of the behaviour of others over time. One way one can realign incentives between different parties to economic exchange and induce cooperative behaviour is by disseminating reputational information (Greif, 1993). There is evidence that when such information is available, people will make use of it, discriminating in favour of those with positive reputations. Indeed, the availability of reputational information has been the cornerstone of the success of online marketplaces such as eBay (Dellarocas, 2003, Lucking-Reiley et al., 2007).

However, the development of decentralized mechanisms that disseminate reputation information occurred well before the advent of the Internet. It has been suggested that one such mechanism is gossip, the class of speech that transmits information about the behaviours and attributes of third parties. By some accounts, gossip is said to account for the majority of human conversation; this raises the possibility that people's everyday conversations may be one of the most important contexts for the transmission of reputational information (Dunbar et al., 1997). The sheer ubiquity of gossip suggests that it may impact on economic exchange even in the presence of more formal reputational mechanisms, like those in online marketplaces. Indeed, word-of-mouth is still widely recognised as a powerful driver of new business (Trusov et al., 2009). However, there is (at least) one potential limitation to gossip serving this economic and social function: its inaccuracy. If the reputational information contained in gossip is inaccurate (through error or deliberate distortion), then recipients are more likely to make mistakes like choosing to trust an untrustworthy or unproductive partner. In these situations, people should be less willing to discriminate on the basis of gossip that they receive. Although there is no behavioural evidence for this claim, simulation work has shown that as gossip becomes less accurate, discriminating players are less successful (Nakamaru and Kawata, 2004; Roberts, 2008; Rauwolf et al., 2015).

In this paper, we explore the impact of gossip and its (in)accuracy on trust and trustworthiness in a population of individuals playing decentralized, two-player trust games. We suggest that gossip inaccuracy erodes the cooperative potential of gossip if it interferes with people's tendencies to discriminate on the basis of gossip and their partners' tendencies to anticipate this discrimination (Basu et al., 2009; Boero et al., 2009; Sperber and Baumard, 2012; Sylwester and Roberts, 2013). If gossip provides accurate information about a potential partner's past trustworthiness and competence, recipients are better able to choose to play with good players and avoid playing with bad players. If people receive inaccurate information, they are hampered in their ability to discriminate in this way. At the same time, gossip creates a demand for reputability, as long as targets are aware of the transmission of gossip, and believe that recipients are likely to act on it. If the accuracy of gossip is in question, the targets of gossip may infer that the recipients of the piece of gossip are unlikely to discriminate on the basis of the information it contains. This reduction in reputational concern is likely to reduce the motivation of the targets of gossip to engage in cooperative acts.

In each round of our experiment, we randomly match an Investor with an Agent; the Investor can send any proportion of her endowment to that Agent. Agents can be High-Productivity or Low-Productivity types: any amount sent to the former type is multiplied by six, while any amount sent to the latter is multiplied by three. Investors do not know what type they are facing when deciding how much to send. Agents can then return any amount they wish to the Investor. Before the round ends, each Investor sends a message summarizing the events of the current round: (i) the amount the Agent received; (ii) her type; (iii) the amount the Agent returned to the Investor. At the start of the following round, and prior to making their sending decision, Investors receive a message pertaining to the Agent they are matched with.

In a first set of treatments, we explore whether inaccuracy in messages sent between Investors spontaneously occurs in populations playing our game, and we study what effect such inaccuracy has on levels of trust and trustworthiness. To achieve this, we generated different experimental treatments by varying who Investors sent their messages to and who they received them from. In the Gossip treatment, Investors sent their messages to the Investor who would play with their Agent in the following round. In the Truth treatment, Investors sent their messages to the experimenter, and received factual summaries of Agent behaviour from the experimental software. In the No Message treatment, Investors neither sent nor received messages. By comparing behaviour in Gossip to behaviour in No Message, we are able to understand to what extent Investors are willing to discriminate on the basis of the information they receive. Comparing Gossip to Truth allows us to understand if Investors are more likely to lie when communicating to one another than when communicating to the experimenter. This comparison also shows to what extent spontaneous message inaccuracy undermines the ability of Investors to discriminate between Agents and the resulting willingness of Agents to build positive reputations.

In an additional treatment, we study how exogenous message inaccuracy affects trust and trustworthiness. This Inaccurate treatment was a variation on the Gossip treatment. As before, Investors sent their messages to the Investor who would next play with their Agent. However, in this new treatment there was a 50% chance in a given round that an Investor would unknowingly receive a message intended for another Investor. Both Investors and Agents knew of this manipulation, although this was not common knowledge (i.e., Investors were not told that Agents were also aware of the potential for misdirected gossip and vice versa). In this treatment, Investors have weaker incentives to discriminate on the basis of the information they receive; this translates into a weaker incentive for Agents to be trustworthy. This treatment allows us to understand how high levels of inaccuracy affect Investors' reputation-based discrimination and Agents' reputational concern.

The rest of the paper is organized as follows. Section 2 contextualises our paper in the literature on reputation and gossip. Section outlines the research questions and summarizes the experimental design and procedures. Section 4 presents the results and Section 5 offers some concluding remarks.

Section snippets

Related literature

In populations where there is uncertainty about the type of player one is matched with, players may want to discriminate against those who have past histories of uncooperative behaviour, and discriminate in favour of those who have reliably cooperated in the past (Wilson, 1985). This means that if reputation-based discrimination is prevalent, strategic reputation building can be a sensible strategy. In these populations, people may find that the immediate costs of cooperation are outweighed by

Experimental design and hypotheses

In this pre-registered experiment (https://tinyurl.com/hc27ajw), we implemented a modified version of the trust game (Berg et al., 1995) in networks of 12 people. Participants played the game on networked computers that were separated by partitions to ensure anonymity. At the start of each session, equal numbers of participants were randomly assigned to be Investors (N=6) or Agents (High-Value N=3; Low-Value N=3), a role they retained for the duration of the experiment. We used the neutral

Message (in)accuracy

There was a considerable degree of spontaneous inaccuracy in the messages that were sent by Investors in the reputation treatments: 53 per cent of Investors sent one or more inaccurate messages, such that about 20 per cent of all messages had at least one inaccurate component. Notably, this approaches the levels of inaccuracy that have been shown to undermine the cooperative potential of gossip in simulation studies (Roberts, 2008; Rauwolf et al., 2015). This inaccuracy was more often seen in

Conclusion

It has been suggested that inaccuracy poses a major threat to the capacity of gossip to boost cooperation in society. Our results suggest that while inaccuracy is intrinsic to gossip—most Investors misrepresented their Agent's behaviour at least once—and perceived to serve a range of strategic purposes, it may be less of a threat to its functionality than previous simulation work suggests. Even though spontaneous inaccuracy in messages was widespread, meaning that gossip provided a weak

References (48)

  • M. Abraham et al.

    Reputation formation in economic transactions

    J. Econ. Behav. Organ.

    (2016)
  • K.J. Arrow

    The Limits of Organization

    (1974)
  • P. Barclay

    Harnessing the power of reputation: strengths and limits for promoting cooperative behaviors

    Evol. Psychol.

    (2012)
  • S. Basu et al.

    Recordkeeping alters economic history by promoting reciprocity

    Proc. Natl. Acad. Sci. USA

    (2009)
  • B. Beersma et al.

    How the grapevine keeps you in line: gossip increases contributions to the group

    Soc. Psychol. Pers. Sci.

    (2011)
  • J. Berg et al.

    Trust, reciprocity and social history

    Games Econ. Behav.

    (1995)
  • R. Boero et al.

    Reputational cues in repeated trust games

    J. Socio-Econ.

    (2009)
  • G. Charness et al.

    Trust and trustworthiness reputations in an investment game

    Games Econ. Behav.

    (2011)
  • C. Dellarocas

    The digitization of word of mouth: promise and challenges of online feedback mechanisms

    Manag. Sci.

    (2003)
  • R.I.M. Dunbar

    Groups, gossip, and the evolution of language

  • R.I.M. Dunbar

    Gossip in evolutionary perspective

    Rev. Gen. Psychol.

    (2004)
  • R.I.M. Dunbar et al.

    Human conversational behavior

    Hum. Nat.

    (1997)
  • N. Emler

    A social psychology of reputation

    Eur. Rev. Soc. Psychol.

    (1990)
  • N. Emler

    Gossiping

  • D. Engelmann et al.

    Inequality aversion, efficiency, and maximin preferences in simple distribution experiments

    Amer. Econ. Rev.

    (2004)
  • M. Enquist et al.

    The evolution of cooperation in mobile organisms

    Animal Behav.

    (1993)
  • Fehr, D., Sutter, M., 2016. Gossip and the efficiency of interactions. IZA DP No....
  • M. Feinberg et al.

    Gossip as an effective and low cost form of punishment

    Behav. Brain Sci.

    (2012)
  • M. Feinberg et al.

    The virtues of gossip: reputational information sharing as prosocial behavior

    J. Pers. Soc. Psychol.

    (2012)
  • M. Feinberg et al.

    Gossip and ostracism promote cooperation in groups

    Psychol. Sci.

    (2014)
  • U. Fischbacher

    z-Tree – Zurich toolbox for readymade economic experiments

    Exper. Econ.

    (2007)
  • E.K. Foster

    Research on gossip: taxonomy, methods, and future directions

    Rev. Gen. Psychol.

    (2004)
  • D. Fudenberg et al.

    Maintaining a reputation when strategies are imperfectly observed

    Rev. Econ. Stud.

    (1992)
  • A. Greif

    Contract enforceability and economic institutions in early trade: the Maghribi traders' coalition

    Amer. Econ. Rev.

    (1993)
  • Cited by (43)

    • Smartphone use decreases trustworthiness of strangers

      2024, Journal of Economic Psychology
    • Online markets and trust

      2022, Journal of Economic Behavior and Organization
      Citation Excerpt :

      Our paper relates to the experimental studies on how formal reputation systems influence behavior in strategic interaction. Examples of experimental studies on noise-free reputation information and its impact on trust include: Bolton et al. (2004); Bohnet and Huck (2004); Fehr et al. (2009); Charness et al. (2011); Huck et al. (2012); Goeschl and Jarke (2017); Fonseca and Peters (2018). By contrast, there are also studies on various types of noisy reputation information.

    • More than one's negative ties: The role of friends’ antipathies in high school gossip

      2022, Social Networks
      Citation Excerpt :

      To answer this question, our point of departure is that gossip needs to be addressed beyond a pure sender-based perspective. Some explanations highlighted in the literature are directly related to the target’s behaviour, particularly those flagging norm violations, cheating, or free-riding (Fehr & Sutter, 2019; Feinberg et al., 2014; Fonseca & Peters, 2018; Samu et al., 2020). In this sense, gossip can be seen as an indirect and cheap way of punishing deviant actions (Becker, 1963; Feinberg et al., 2012; Giardini & Conte, 2012; Kniffin & Wilson, 2005; Malinowski, 1926).

    • When and why does gossip increase prosocial behavior?

      2022, Current Opinion in Psychology
      Citation Excerpt :

      Another question that currently remains unanswered is in which situations people completely reverse information about targets (i.e. depict cooperators as non-cooperators and vice versa) rather than merely exaggerate actual target behaviors, and how this influences prosocial behavior. One possibility is that if gossip senders expect gossip receivers to check gossip veracity by cross-validating it with multiple sources, this would lead them to refrain from lying very blatantly while gossiping (and perhaps to prefer using “exaggeration lies” rather than extreme lies as in Peters & Fonseca found [27]). The simplest structure in which gossip can take place is that of a triad consisting of a sender, a receiver, and a target of gossip [5,29].

    • Direct and indirect reciprocity among individuals and groups

      2022, Current Opinion in Psychology
      Citation Excerpt :

      These results suggest that gossip is used in daily life to influence and update reputations in a way that enables indirect reciprocity. In line with this, recent research also shows that even inaccurate gossip can promote trust among strangers via enhanced reputational concern [36]. Another factor promoting prosocial behavior in social interactions is third-party punishment (Figure 1b).

    View all citing articles on Scopus

    Author names are in alphabetical order. We thank the Editor, Associate Editor, and two anonymous reviewers, as well as Bill von Hippel, and participants at seminars at University of Exeter, University of Porto, University of Minho, University of Queensland, and the 2016 ESA meeting at Jerusalem for useful comments and discussions. The authors thank the University of Queensland, as well as ESRC grant ES/N00762X/1 for financial support. We thank Tim Miller for programming the experimental software.

    View full text