Data from: The price of being seen to be just: an intention signalling strategy for indirect reciprocity

Cooperation amongst strangers is a marked characteristic of human sociality. One prominent evolutionary explanation for this form of human cooperation is indirect reciprocity, whereby each individual selectively helps people with a ‘good’, but not ‘bad’ reputation. Some evolutionary analyses have underscored the importance of second-order reputation information (the reputation of a current partner’s previous partner) for indirect reciprocity as it allows players to discriminate justified ‘good’ defectors, who selectively deny giving help to ‘bad’ partners, from unjustified ‘bad’ defectors. Nevertheless, it is not clear whether people in fact make use of second-order information in indirect reciprocity settings. As an alternative, we propose the intention signalling strategy, whereby defectors are given the option to abandon a resource as a means of expunging their ‘bad’ reputation. Our model deviates from traditional modelling approaches in the indirect reciprocity literature in a crucial way—we show that first-order information is sufficient to maintain cooperation if players are given an option to signal their intention. Importantly, our model is robust against invasion by both unconditionally cooperative and uncooperative strategies, a first step towards demonstrating its viability as an evolutionarily stable strategy. Furthermore, in two behavioural experiments, when participants were given the option to abandon a resource so as to mend a tarnished reputation, participants not only spontaneously began to utilise this option, they also interpreted others’ use of this option as a signal of cooperative intent.

陌生人之间的合作是人类社会性的显著特征。针对这类人类合作形式，一种重要的进化论解释是间接互惠（indirect reciprocity）：个体可选择性地帮助拥有“良好”而非“不良”声誉的他人。部分进化分析强调了二阶声誉信息（second-order reputation information，即当前合作对象的过往合作对象的声誉）对于间接互惠的重要性，因为该信息可帮助博弈者甄别正当的“善意”背叛者——这类个体选择性地拒绝向“不良”声誉的对象提供帮助——与不正当的“恶意”背叛者。然而，目前尚不清楚人类在间接互惠场景中是否真的会利用二阶声誉信息。为此我们提出意图信号策略（intention signalling strategy）：允许背叛者通过放弃某项资源来消除自身的“不良”声誉。本研究的模型与间接互惠领域的传统建模思路存在关键差异：我们证明，若允许个体传递自身意图的信号，仅需一阶信息（first-order information）即可维持合作。值得注意的是，本模型对无条件合作策略与不合作策略的入侵均具备鲁棒性，这是证明其作为进化稳定策略（evolutionarily stable strategy）可行性的关键第一步。此外，在两项行为实验中，当参与者被赋予通过放弃资源修复受损声誉的选项时，他们不仅会自发地使用该选项，还会将他人使用该选项的行为解读为合作意图的信号。