Facebook 粉絲專頁發布假新聞行為之網路外溢效果分析：以 2016 年美國總統大選為例

Abstract

本研究以 2016 年美國總統大選期間 Facebook 粉絲專頁之假新聞散播現象為觀察對象，探討粉專間是否存在假新聞發布行為的跨粉專外溢效應，並深入分析其背後的傳播機制與社群結構特性。本研究透過社會網絡分析與空間自迴歸（SAR）模型分析，利用 941 個 Facebook 粉絲專頁自 2016 年 7 月 31 日至 11 月 5日之縱橫資料，建構使用者與粉專間的二模網絡並進一步轉換為粉專間的一模網絡，並以特徵向量中心性分析粉專的重要性及互動結構特性。 研究結果顯示，假新聞的發布高度集中於少數特定粉專，且呈現明顯的政治不對稱性，Pro-Trump 陣營粉專的假新聞發布數量顯著高於 Pro-Clinton 陣營。此外，粉專之間透過高度共享的使用者形成明確且穩定的社群結構，呈現出顯著的政治分化特徵，即所謂的「同溫層效應」。中心性分析更進一步揭示，Pro-Trump陣營內部的互動結構較為密切與持續，Pro-Clinton 陣營則更傾向於短期且高度依賴外部政治事件的刺激。SAR 模型分析進一步證實，粉專發布假新聞的行為明顯具有跨粉專的模仿效應，特別是當同一政治陣營內的其他粉專增加假新聞發布頻率後，粉專本身也傾向跟進模仿此行為，進一步增加自身的假新聞發布數量。其中 Pro-Trump 陣營的外溢效果最強烈，而跨陣營間的外溢效應則相對較弱。此外，本研究發現假新聞的外溢傳播效果也受到粉專自身性質的影響，尤其是媒體型與組織型粉專具有較強的外溢性質。

This study examines the dissemination of fake news among Facebook fan pages during the 2016 U.S. Presidential Election, specifically investigating whether fake news posting behaviors exhibit cross-page spillover effects, and further analyzing the underlying dissemination mechanisms and community structures. Using social network analysis and Spatial Autoregressive (SAR) modeling, this research utilizes panel data comprising 941 Facebook fan pages from July 31 to November 5, 2016. The analysis begins by constructing a two-mode network representing interactions between users and fan pages, subsequently transforming it into a one-mode network that captures interactions among fanpages. Eigenvector centrality is then applied to evaluate the significance and interaction characteristics of these fan pages within the network. The findings reveal that fake news postings are highly concentrated among a small number of fan pages, displaying clear political asymmetry, with Pro-Trump pages posting significantly more fake news compared to Pro-Clinton pages. Additionally, fan pages form distinct and stable community structures through highly shared users, characterized by significant political polarization, often referred to as an ”echo chamber” effect. Centrality analysis further indicates that the internal interaction structures within Pro-Trump pages are tighter and more sustained, while those within Pro-Clinton pages tend to be short-term and heavily reliant on external political events. Moreover, the SAR modeling analysis confirms significant cross-page spillover effects in fake news posting behaviors, especially prominent within the same political camp,with the strongest spillover effect observed among Pro-Trump fan pages, whereas cross-camp spillover effects are comparatively weaker. This study also finds that the nature of fan pages influences the strength of spillover effects, with media-type and organization-type fan pages exhibiting stronger spillover properties.