Reputational Priors Magnify Striatal Responses to Violations of Trust

Abstract

Humans learn to trust each other by evaluating the outcomes of repeated interpersonal interactions. However, available prior information on the reputation of traders may alter the way outcomes affect learning. Our functional magnetic resonance imaging study is the first to allow the direct comparison of interaction-based and prior-based learning. Twenty participants played repeated trust games with anonymous counterparts. We manipulated two experimental conditions: whether or not reputational priors were provided, and whether counterparts were generally trustworthy or untrustworthy. When no prior information is available our results are consistent with previous studies in showing that striatal activation patterns correlate with behaviorally estimated reinforcement learning measures. However, our study additionally shows that this correlation is disrupted when reputational priors on counterparts are provided. Indeed participants continue to rely on priors even when experience sheds doubt on their accuracy. Notably, violations of trust from a cooperative counterpart elicited stronger caudate deactivations when priors were available than when they were not. However, tolerance to such violations appeared to be mediated by prior-enhanced connectivity between the caudate nucleus and ventrolateral prefrontal cortex, which anticorrelated with retaliation rates. Moreover, on top of affecting learning mechanisms, priors also clearly oriented initial decisions to trust, reflected in medial prefrontal cortex activity.

Why we stay with our social partners: Neural mechanisms of stay/leave decision-making

“…That is, previous studies have consistently demonstrated that neural pathways underlying social reward-processing and decisionmaking engage the reward-circuit (Behrens et al, 2008;De Quervain et al, 2004;Harris & Fiske, 2010;Hare et al, 2010;Jones et al, 2001;Moll et al, 2006;Phan et al, 2010;Rilling & Sanfey, 2011;Zaki & Mitchell, 2011) and that Reinforcement Learning models can be applied to social learning Fouragnan et al, 2013;Jones et al, 2001;Lin et al, 2012;Poore et al, 2012;Xiang, Lohrenz, & Montague, 2013).…”

“…On the contrary, vmPFC and striatum also encode social value (Behrens, Hunt, Woolrich, & Rushworth, 2008;Bhanji & Delgado, 2014;De Quervain et al, 2004;Fareri, Niznikiewicz, Lee, & Delgado, 2012;Hare, Camerer, Knoepfle, O'Doherty, & Rangel, 2010;Jones et al, 2001;Moll et al, 2006;Phan, Sripada, Angstadt, & McCabe, 2010;Rilling & Sanfey, 2011;Zaki & Mitchell, 2011), such as the value of an attractive face (Aharon et al, 2001;Kampe, Frith, Dolan, & Frith, 2001;O'Doherty, Winston, et al, 2003), the value of social approval/acceptance or having a good reputation (Izuma, Saito, & Sadato, 2008;Jones et al, 2011;Lin et al, 2012), the value of cooperation (Krill & Platek, 2012;Rilling et al, 2002;Stallen & Sanfey, 2013;Watanabe et al, 2014), and the value of charitable giving (Hare et al, 2010;Moll et al, 2006). Additionally, for both social and nonsocial contexts also discrepancy between expected and outcome values (e.g., reward prediction errors; Delgado, Nystrom, Fissel, Noll, & Fiez, 2000;Fareri, Chang, & Delgado, 2012;Fouragnan et al, 2013;Harris & Fiske, 2010;Jones et al, 2001;O'Doherty, Dayan, et al, 2003;Pessiglione, Seymour, Flandin, Dolan, & Frith, 2006;Phan et al, 2010;Poore et al, 2012;Schönberg, Daw, Joel, & O'Doherty, 2007;Schültz, Dayan, & Montague, 199...…”

The neurobiology of rewards and values in social decision making

“…Moreover, these trust-learning effects are strongly modulated by prior expectations (as is the case for non-social reward-learning effects 51 ): unexpected reciprocity in any trial induced stronger ventral striatum activity and higher investments in subsequent encounters compared with expected reciprocity 53 , suggesting that this activity may reflect a reward prediction error similar to that observed during non-social forms of reward learning. However, striatal prediction errors and behavioural learning during the outcome of trust decisions can also be suppressed when information about the opponents' trustworthiness is presented 54 , suggesting that strong priors can overrule the importance of new information gathered during single trials.…”

Neural signatures of trust in reciprocity: A coordinate‐based meta‐analysis

“…Indeed, the trustor may lose the amount of money she decides to send if the trustee defects and defection is highly desirable for the trustee since it yields him the highest outcome (Camerer and Weigelt, 1988; Fouragnan, et al, 2013). However, empirical studies have shown that trustors invest more than half of their initial endowment and trustees split the amount of money they received evenly (Camerer, 2003a; Fehr and Fischbacher, 2003; Johnson and Mislin, 2011).…”

“…Social preferences, such as prosociality or betrayal aversion, imply that subjects derive extra utility/disutility from trusting/distrusting behaviors (Bohnet, et al, 2008; Bohnet and Zeckhauser, 2004; Fehr and Camerer, 2007). Finally, beliefs about others built through social priors (e.g., implicit racial attitudes, group memberships) or reputation on the basis of previous experiences (via trial-and-error learning) help people estimate the utility associated with the decision options and adjust their behavior for future interactions (Chang, et al, 2011; Fareri, et al, 2012; Fouragnan, et al, 2013; King-Casas, et al, 2005). …”