Facebook 貼文之爭議性與討論度關係的調節變數分析

Abstract

隨著社群媒體的快速發展，內容創作者積極的尋找能提升使用者參與度的關鍵因素，但由於網路平台的特色迥異，不僅影響到了創作者的營運方向，也吸引到了不同特質的使用者群體，使得單一的標準無法適用於所有社群平台，因此相關研究百花齊放。此外，文章往往因為本身主題的關係，而帶著與生俱來、或高或低的爭議性，對使用者討論度造成顯著影響。本研究以留言數作為討論度的衡量指標，旨在並透過分析各式調節變數，找出能左右爭議性與討論度的關係之因子。

本研究以Facebook上的文章為對象，透過其留言內容來計算文章的爭議性，並探討調節變數的介入將如何使爭議性與討論度的關係產生變化，希望能協助社群媒體上的內容創作者制定創作策略。本研究利用大型語言模型GPT-4o以及Gemini 2.0 Flash進行爭議性計算，並透過迴歸分析驗證爭議性與討論度的關係。

As social media continues to evolve rapidly, content creators have been actively seeking key factors that can enhance user engagement. However, due to the distinct characteristics of different platforms, the operational strategies of creators are affected, and varying user groups are attracted, making it impractical to apply a single standard across all social media. As a result, related studies have been flourishing. Among the factors influencing engagement, controversy plays a significant role, as posts often intrinsically carry different degrees of controversy depending on their topics, which in turn shapes the level of user discussion and overall engagement. This study conceptualizes user discussion as the number of comments, and further seeks to identify the key factors that are able to influence the relationship between controversy and discussion through the analysis of various moderating variables.

The target of this study is the posts from Facebook pages. We quantify post-level controversy based on the content of the comments and examine how the intervention of moderating variables affects the relationship between controversy and discussion, in the hope of assisting content creators on social media to develop more effective content strategies. This study utilizes large language models, such as GPT-4o and Gemini 2.0 Flash, to facilitate the quantification of controversy, and applies regression models to validate the influence of controversy on discussion.