RAPID: A Quasi-Experimental Study of Corporate Policy Effects

为了回应公众对社交媒体中辱骂和负面信息的强烈抗议，社交媒体正在着手一系列政策变化，以澄清现有标准，并促进更多的报道和删除辱骂内容。这项研究将通过对任何政策转变前后发送的在线消息进行情绪分析，来检验这些规则变化对负面消息的影响。通过找出规则变化影响的个人和地理差异的来源，它将揭示影响对公司规则的反应的社会和文化因素，以及导致网络攻击模式的社会和文化因素。这项研究还将确定有助于或未能帮助限制日常在线互动的负面影响的组织实践。这项研究将利用不同学科的专业知识来实现其目标。数据将来自三个来源：1)根据人口统计特征以个人为目标的在线消息，2)包含常见诽谤的消息，以及3)包括发布地点的消息(地理标记)。这些信息将通过扩展地理可视化和分析技术以及信息检索方法来捕获和处理，并通过使用计算方法来计算情绪和评估社交媒体消息的积极和消极。统计分析将显示，新的公司规则是否会导致情绪负面程度显著下降或可能上升。它们还将显示，在公司规则改变后，辱骂沟通是否会变得更加含糊其辞，更难识别。

In response to public outcry over abusive and negative messages in social media, social media outlets are embarking on a set of policy changes to clarify existing standards and facilitate more reporting and removal of abusive content. The research will examine the effect of these rule changes on negative messages through sentiment analyses of online messages sent before and after any policy shifts. By identifying sources of individual and geographic variation in the effects of the rule changes, it will shed light on the social and cultural factors that shape responses to corporate rules and that contribute to patterns of cyber aggression. The research also will identify organizational practices that aid, or fail to aid, in limiting the negativity of everyday online interactions.This research will use a convergence of expertise in different disciplines to accomplish its goals. Data will come from three sources: 1) online messages that target individuals on the basis of their demographic characteristics, 2) messages that contain common slurs, and 3) messages that include the location from which they were posted (geo-tagged). These will be captured and processed by extending geographic visualization and analysis techniques and information retrieval methods, and by using computational methods to compute sentiment and to assess the positivity and negativity of social media messages. Statistical analyses will show if new corporate rules result in significant decreases or possibly increases in level of sentiment negativity. They also will show whether abusive communications become more evasive and less identifiable following corporate rule changes.