Extensive interdisciplinary literatures have built on the seminal spatial dilemmas model, which depicts the evolution of cooperation on regular lattices, with strategies propagating locally by relative fitness. In this model agents may cooperate with neighbors, paying an individual cost to enhance their collective welfare, or they may exploit cooperative neighbors and diminish collective welfare. Recent research has extended the model in numerous ways, incorporating behavioral noise, implementing other network topologies or adaptive networks, and employing alternative dynamics of replication. Although the underlying dilemma arises from two distinct dimensions—the gains for exploiting cooperative partners (Greed) and the cost of cooperating with exploitative partners (Fear)–most work following from the spatial dilemmas model has argued or assumed that the dilemma can be represented with a single parameter: This research has typically examined Greed or Fear in isolation, or a composite such as the K-index of Cooperation or the ratio of the benefit to cost of cooperation. We challenge this claim on theoretical grounds—showing that embedding interaction in networks generally leads Greed and Fear to have divergent, interactive, and highly nonlinear effects on cooperation at the macro level, even when individuals respond identically to Greed and Fear. Using computational experiments, we characterize both dynamic local behavior and long run outcomes across regions of this space. We also simulate interventions to investigate changes of Greed and Fear over time, showing how model behavior changes asymmetrically as boundaries in payoff space are crossed, leading some interventions to have irreversible effects on cooperation. We then replicate our experiments on inter-organizational network data derived from links through shared directors among 2,400 large US corporations, thus demonstrating our findings for Greed and Fear on a naturally-occurring network. In closing, we discuss implications of our main findings regarding Greed and Fear for the problem of cooperation on inter-organizational networks.
展开▼
机译：广泛的跨学科文献建立在开创性的空间困境模型上，该模型描述了规则格上合作的演变，其策略通过相对适合度局部传播。在此模型中，代理人可以与邻居合作，支付个人成本以增强他们的集体福利，或者他们可以利用合作邻居并减少集体福利。最近的研究以多种方式扩展了该模型，包括行为噪声，实现其他网络拓扑或自适应网络以及采用替代的复制动态。尽管潜在的困境来自两个不同的维度-合作伙伴的开发收益（Greed）和与剥削伙伴的合作成本（Fear）-但是，空间困境模型所进行的大多数工作都认为或假定可以用以下两种方法来表示这种困境：单个参数：该研究通常单独检查“贪婪”或“恐惧”，或诸如“合作K指数”或“合作收益与合作成本之比”之类的复合词。我们从理论上对这一主张提出异议-表明在网络中嵌入交互通常会导致Greed和Fear在宏观层面上对合作产生分歧，互动和高度非线性的影响，即使个人对Greed和Fear的反应相同。使用计算实验，我们可以描述该空间区域中的动态局部行为和长期结果。我们还模拟了干预措施，以调查贪婪和恐惧随时间的变化，显示模型行为如何随着收益空间中的边界越过而非对称地变化，从而导致某些干预措施对合作产生不可逆的影响。然后，我们在来自2400家大型美国公司中通过共享董事的链接得出的组织间网络数据上的实验进行了复制，从而证明了我们在自然发生的网络上对Greed and Fear的发现。最后，我们讨论了关于贪婪和恐惧的主要发现对组织间网络合作问题的影响。
展开▼
