Understanding the Emergence of Pro-Social Behaviourin Dynamic Social Networks

在社会困境中，个人利用集体资源，“合作者”为改善公共产品付出成本，而“脱北者”则搭便车，在不产生任何成本的情况下获得利益。虽然叛逃可能是一次互动中的个人理性策略，但我们知道，合作可以在重复的游戏中持续下去，在这种游戏中，会出现复杂的策略来对抗叛逃者。在这个项目中，我们希望了解动态社交网络中发生的事情，在动态社交网络中，面临社交困境的个人可以选择与谁互动，并可以观察(并可能模仿)他人的行为。那么，社交网络什么时候能促进合作呢？是容易建立联系、互动的类型，还是我们与他人的相似之处促使我们超越自己的利益？在最近的理论分析中，我们发现模仿速度与伙伴选择之间的比例是促进合作的关键。然而，我们的结果仍然局限于简单的伴侣选择和模仿规则，仍然与经验发现，即人类在实践中所做的事情没有联系。这个项目的目标是设计一个基于学习的社交网络合作的理论模型，并将其与我们收集的行为数据联系起来，目的是找出对我们在人类数据中看到的真正合作和叛逃模式的解释。这是一个跨学科的挑战，它同时使用分析和实验方法，提出一个严格的、有经验支持的理论模型，解释亲社会行为何时会出现在动态的社交网络中。项目承办方将采用理论分析，即通过研究动力系统和/或其算法复杂性，计算机辅助分析，即通过基于代理人的建模和模拟，以及经验分析，即通过人类实验。

In social dilemmas, where individuals make use of collective resources, "co-operators" pay a cost to improve public goods whereas "defectors" free-ride, reaping benefit without incurring any costs. While defection may be the individually rational strategy in one-shot interactions, we know that cooperation can be sustained in repeated games, where complex strategies emerge to counter defectors. In this project we want to understand what happens in dynamic social networks, where individuals facing a social dilemma can choose who to interact with, and can observe(and potentially imitate) others' behaviour. So, when do social networks promote cooperation? is it the ease of making connections, the type of interaction, or the similarity we have with others that motivate us to look beyond our self-interest? In recent theoretical analysis we found that the ratio between the speed of imitation and partner selection is key to promote cooperation. Our results are however still limited to simple partner selection and imitation rules and are still not linked with empirical findings i.e., what humans do in practice. The objective of this project is to devise a theoretical model of learning-based cooperation on social networks and link it to behavioural data we have collected, with the goal of identifying explanations for the patterns of real cooperation and defection we see in the human data. This is an interdisciplinary challenge that uses both analytical and experimental methodologies to come up with a rigorous and empirically backed theoretical model explaining when pro-social behaviour can emerge in dynamic social networks. The project undertaker will employ theoretical analysis, i.e., through the study of dynamical systems and/or their algorithmic complexity, computer-aided analysis, i.e., through agent-based modelling and simulation, and empirical analysis, i.e., through experiments with humans.