In this paper we address how complex social communities emerge from local decisions by individuals with limited attention and knowledge. This problem is critical; if we understand community formation mechanisms, it may be possible to intervene to improve social welfare. We propose an interpretable, novel model for attributed community formation driven by resource-bounded individuals' strategic, selfish behavior. In our stylized model, attributed individuals act strategically in two dimensions: attribute and network structure. Agents are endowed with limited attention, and communication costs limit the number of active connections. In each time step, each agent proposes a new friendship. Agents then accept proposals, decline proposals, or remove friends, consistent with their strategy to maximize payoff. We identify criteria (number of stable triads) for convergence to some community structure and prove that our community formation model converges to a stable network. Ablations justify the ecological validity of our model and show that each aspect of the model is essential. Our empirical results on a physical world microfinance community demonstrate excellent model fits compared to baseline models.

翻译：本文探讨了有限注意力和知识的个体如何通过局部决策涌现出复杂的社会社群。这一问题至关重要：若能理解社群形成机制，便可能通过干预提升社会福利。我们提出了一种可解释的新型归因社群形成模型，该模型由资源受限个体的策略性利己行为驱动。在我们的简化模型中，具有属性的个体在属性和网络结构两个维度上采取策略行为。行动者被赋予有限的注意力，且通信成本限制了活跃连接的数量。在每个时间步中，每个行动者会提议建立新的友谊。随后，行动者根据最大化自身收益的策略，接受或拒绝提议，或移除已有朋友。我们识别了社群结构收敛的判据（稳定三元组的数量），并证明所提出的社群形成模型能够收敛至稳定网络。消融实验验证了模型的生态效度，并表明模型的每个方面都不可或缺。基于真实世界小额信贷社群的实证结果表明，与基线模型相比，我们的模型具有极佳的拟合效果。