笑談政治？以《夜夜秀》為例分析YouTube留言區的審議互動

Abstract

本研究以政治脫口秀《賀瓏夜夜秀》為研究對象，探討其YouTube留言區政治對話的審議品質。研究蒐集2024年總統大選期間三個政治性單元——「欸！」、新聞亂報、來賓訪談的觀眾留言，依據審議民主理論建構「互動性」與「建設性」兩大核心意涵，並且設計形式互動、實質互動、垃圾留言、結構性主題相關性與互動性主題相關性等五項可量化操作的評估指標，結合傳統機器學習與大語言模型，提出一套新的審議性留言分析框架。 研究結果發現，68% 的留言屬於「有意義的政治對話」，但僅有 28% 同時具備互動性與建設性，另有 40% 為「單向政治留言」。不過值得注意的是，在這些缺乏互動性的政治留言中，其「互動性結構相關性」反而高於具互動性的留言，亦即這些單向留言在內容上與留言區的整體輿論形成呼應。此一結果反映出，當代公共領域的意涵已轉向將網路視為開放性的意見交流場域，人們可以在平台上自由表達個人政治立場與觀點。儘管此類留言未必能引發回應、形成討論串，仍具有推進有意義政治對話的潛力。

This study examines the deliberative quality of political discussion in the YouTube comment section of “The Night Night Show with Hello” during Taiwan’s 2024 presidential election. Grounded in deliberative theory, this study identify reciprocity and constructiveness as core dimensions, operationalized into five measurable indicators. The study employs a mixed computational approach that combines traditional machine learning with large language models to develop a scalable framework for assessing the quality of online political discussion. Results indicate that 68% of comments qualify as meaningful political discussion, yet only 28% display both reciprocity and constructiveness. A further 40% are one-way political comments which, despite lacking direct interaction, exhibit strong alignment with overall conversational trends. These findings suggest that contemporary online public spheres increasingly operate as a public communication network, where even non-interactive contributions can play a role in advancing meaningful dialogue.