近年来，随着用户数量增长放缓，在线平台对于用户使用时长的争夺逐渐白热化，诸多平台选择将好友关系数字化，即将用户线下真实的好友关系或其他第三方平台的好友关系导入平台，一方面快速拓展用户群体，另一方面希望增加用户在平台的点评互动，进而增加用户活跃度及用户使用时长。国内外学者对于数字化好友关系进行了一定的研究，但过往研究主要从用户可以获取更多的社会资本角度，解释数字化好友关系对用户行为的影响，少有研究关注信息隐私对用户行为的影响，且过往研究少有关注用户发布内容的文本特征。本文以全球最大的本地生活服务平台及评论平台——大众点评为研究对象，基于社会临场感理论与信息隐私相关理论，通过对大众点评导入数字化的好友关系后实施“置顶好友评论”政策与“一键停用第三方全部社交关系”政策进行研究，分析用户在政策实施前后发布评论行为的变化，具体包括信息量及情绪的变化。首先，本文使用倾向得分匹配方法匹配了用户样本，并抓取了大众点评的相关用户评论。其次，本文使用基于离散情感词典的方法，计算了每条评论文本的情绪得分。第三，本文采用双重差分模型对“置顶好友评论”政策实施后用户发布评论行为进行分析，发现政策实施后用户评论中蕴含的信息量增加、积极情绪减弱、消极情绪增强，同时用户隐私披露程度较低的用户表达消极情绪的增强幅度更小。第四，本文使用断点回归模型对 “一键停用第三方全部社交关系”政策实施后用户评论行为进行分析，发现政策实施后用户发布评论中积极情绪增强，这一发现也支撑了信息隐私对用户情绪的影响研究。最后，本文使用双重差分模型对离散情绪进行了拓展研究，深入理解政策变化前后用户发布评论中蕴含情绪的变化。本文具有以下几点贡献：第一，本文补充了数字化的好友关系对在线评论产生影响的研究；第二，本文补充了信息隐私对在线平台用户行为的研究；第三，本文补充了用户在面对隐私侵犯时的情绪应对策略的研究；第四，本文补充了数字化好友关系分析中使用文本分析的方法与实践；第五，在情绪维度理论基础上，本文也结合了离散情绪理论对用户情绪变化进行了探究。本文的研究结果可以帮助在线平台管理者深入理解实施数字化关系对用户行为的影响，提示平台管理者重视保护用户信息隐私，并增强用户对平台内好友关系、社交互动的自主选择权，具有较强的现实意义。
In recent years, with the growth in the number of Internet users slowing down, online platforms are competing more fiercely for users' time spending. Many platforms choose to digitize friendship, which imports the real offline friendship network or third-party platform friendship network into the platform. On the one hand, platforms anticipate rapid expanding of user group after digitizing friendship. On the other hand, platforms hope to increase users' interactions on the platform, so as to increase users' time spending. Domestic and foreign scholars have done research on digital friendship, but previous studies mainly explained its impact on user behavior from the perspective of social capital. Few studies focused on the impact of information privacy on user behavior, and few studies put empahsis on linguistic features of user-generated content.This study considers two natural experiments at the world's largest local life service platform and UGC platform(Dianping.com). Building on social presence theory and information privacy theory, we leverage the exogenous policy change of "Top Friends' Comments" policy and "One Click to Stop All Third-Party Social Relations" policy to assess the impact of digital friendship on linguistic features(information and emotion) of user-generated content. First of all, we use propensity score matching method to match the user samples, and grab the user comments on Dianping. Secondly, we calculate sentiment score of each comment based on discrete sentiment dictionary. Thirdly, we estimate a DID model to assess the impact of "Top Friends' Comments" policy on user's comment behavior. We find that the amount of information contained in user's comments increases, positive emotion decreases and negative emotion increases. At the same time, the increase of negative emotions of users with lower self-disclosure is less. Fourthly, we estimate use a RDD model to assess the influence of "One Click to Stop All Third-Party Social Relations" policy on user's comment behavior. We find that positive emotion is enhanced after implementation of the policy, which also supports the impact of information privacy on user's emotion. Finally, we apply a DID model to assess the impact of change of "Top Friends' Comments" policy on discrete emotions, which helps platforms deeply understand the changes of emotions before and after policy changes.This study has the following contributions: first of all, this study supplements research on the impact of digitali friendship on online comments. Secondly, this study supplements research on the impact of information privacy on online platform user behavior. Thirdly, this study supplements research on users' emotional coping strategies in the face of privacy violations. Fourthly, this study supplements research on digital friendship with text mining methods. Fifthly, this study applies dimensional model of emotion and discrete emotion theory to assess the impact of policy change on user behaviour.Our findings can help online platforms deeply understand the impact of the implementation of digital friendship on user behavior. Our findings also suggest platforms to pay more attention to protecting user information privacy, and enhance user's independent choice of interaction on the platform, which has strong practical significance.
