The Role of Social Identity for Learning in Networks

本研究旨在探讨社会认同对网络中社会学习动态的影响。如今，通过社交网络学习在个人、企业和政府的决策过程中发挥着关键作用。认识到社会学习对行为的关键影响，已经有大量的理论和实验文献旨在理解网络中的学习。综上所述，文献表明社会网络的结构对社会学习的动态具有深远的影响。实验研究主要集中在认知限制如何导致网络中贝叶斯学习的失败。迄今为止，还没有一篇实验论文深入研究社会认同对社会学习的作用。这将是这个项目的主要贡献。我们开发了一个实验装置来研究网络中的信念形成和社会学习。我们以国籍作为社会身份，通过两个国家之间的实验室互联来实现身份群体之间的互动。参与者在网络上互动，并面临回答来自不同专业领域的事实问题的任务。他们同时反复猜测正确的答案，观察他们的网络邻居在每个周期的猜测。在治疗过程中，我们改变了邻居的社会身份信息。在一系列研究中，我们分析了不同的网络大小以及外生和内生网络结构。我们用理论分析来补充实验研究。我们期望做出的主要贡献如下：(1)我们在一个干净和简单的环境中确定了社会认同对网络学习的影响。特别是，我们提供了一个测试，是否个人超重的意见，从分享他们的社会身份的人。(2)我们有助于理解社会认同是如何在社会学习中使用的，以及它如何影响结果。它是否提高了效率，特别是提高了学习的速度，还是会导致学习的扭曲和失败？例如，如果社会身份是专业知识的良好代表，那么歧视可以提高学习速度。(3)研究不同社会网络结构对社会学习动态和歧视行为的影响。社会认同和学习动态之间的相互作用将在允许测试各种不同网络配置的大型网络中得到解决。(4)我们研究了社会身份如何影响社会网络的形成，从而影响人们咨询的信息来源。特别是，我们感兴趣的是不同的社会网络是否会在存在或缺乏社会身份信息的情况下内生地出现，以及这如何影响社会福利。原则上，内源性网络的形成对社会学习是有利还是有害尚不清楚。

The proposed project aims to investigate the effect of social identity on the dynamics of social learning in networks. Learning through social networks today plays a key role in individuals’, firms’, and governments’ decision-making processes. Acknowledging the crucial impact of social learning on behaviour, there is already a large and growing theoretical and experimental literature aiming to understand learning in networks. In sum, the literature has shown that the structure of the social network has profound influence on the dynamics of social learning. Experimental studies mostly focused on how cognitive limitations can lead to failures of Bayesian learning in networks. None of the experimental papers have so far deeply studied the role of social identity for social learning. This will be the main contribution of this project.We develop an experimental setup to study belief formation and social learning in networks. We implement the interaction between identity groups by interconnecting laboratories between two countries, using nationality as social identity. Participants interact on a network and face the task to answer factual questions from various areas of expertise. They simultaneously and repeatedly guess the true answer, observing the guesses of their network neighbors in each period. Across treatments we vary information on the social identity of the neighbors. In a sequence of studies, we analyse different network sizes as well as exogenous and endogenous network structures. We complement our experimental studies with theoretical analysis.The main contributions we expect to make are the following: (1) We identify the effect of social identity on learning in networks in a clean and simple setting. In particular, we provide a test whether individuals overweight opinions from people who share their social identity. (2) We contribute to an understanding how social identity is used in social learning and how this affects the outcome. Does it improve efficiency, in particular increase the speed of learning, or does it lead to distortions and failure to learn? Discrimination can, for example, increase the speed of learning if social identity is a good proxy for expertise. (3) We will study how different social network structures influence the dynamics of social learning and discriminatory behaviour. The interplay between social identity and learning dynamics will be addressed in larger networks that allow to test a variety of different network configurations. (4) We study how social identity affects the formation of social networks and thus the source of information that people consult. In particular, we are interested in whether different social networks will emerge endogenously in the presence or absence of information on social identity and how this influences social welfare. In principle, it is not clear whether endogenous network formation is beneficial or detrimental to social learning.