我們與假新聞之間的距離：以電腦化字詞分析取向解析真假新聞的差異

When Digital Deception is not the Extension of Real-life Deception: Computerized Text Analysis of Digital Deception in Day-to-day News

過往研究均致力於探索人際溝通中如何辨識出真實與欺瞞之間的線索，然在社會大眾越來越倚賴網路訊息傳遞的情況下，找出傳播真假訊息在語言使用風格上的差異就相形重要。本研究回顧有關測謊的理論及研究，發現欺瞞行為會在不同層面語文風格上洩露出線索。因此，本研究嘗試透過語文風格剖析的角度探究真假新聞之間的差異，我們以LIWC作為語文風格分析工具，配合中文版辭典（TC-LIWC）研究真假新聞。研究一以接近測謊理論預設之情境為主，透過實驗操弄邀請研究參與者仿寫真假新聞，藉此了解欺瞞心態反映在語文風格上的變化。研究二以網路蒐集而來之真假新聞為分析主體，並進一步發展可能之預測模式。研究結果發現，在測謊理論預設之情境下，研究參與者似乎傾向在其語文軌跡當中保守地表露情緒感受，此一心理意義為個人試圖避免情緒詞彙引起關注而洩漏說謊線索，且在仿寫假新聞情境下，確實造成認知負荷而難以撰寫複雜之文本。而在自然情境下之假新聞分析，研究發現假新聞傾向以相對較高頻率聚焦負向情緒詞彙的方式喚起閱讀者近似的情緒狀態，且為了引導社會大眾理解的方向而使用較多之認知歷程詞彙來誤導。本研究期待能增進社會大眾對真假新聞的辨識能力有所貢獻。

Previous research has been devoted to exploring ways to identify deception in interpersonal communication. However, as the public increasingly depends on the Internet, discriminating language use between lies and truth is paramount. This study reviews the theory of lie-detection and finds that different levels of linguistic style can provide clues in deceiving. Therefore, this study attempts to explore the differences between authentic and fake news from the perspective of linguistic style analysis. We use LIWC as an analysis tool, along with the Traditional Chinese version dictionary (TC-LIWC). Study 1 focused on the assumptions of the deception theory. Through experimental manipulations, participants were invited to imitate authentic and fake news, so as to understand the difference in deceiving reflected in the linguistic style. Study 2 focuses on the analysis of authentic and fake news collected online and further develops possible prediction models. In study 1, the results showed that under the presupposition of theory, participants tend to conservatively express their emotional feelings in their language use. This means that individuals are trying to avoid using emotional words to get attention and leaking clues. On the other hand, the result of study 2 demonstrates that fake news in natural circumstances demonstrates more cognitive loading which leads to less complex text. This research also suggested that fake news tends to evoke readers' emotional states, especially focuses on negative emotion words at a relatively high frequency, and uses more cognitive processes to mislead the direction of social understanding. This research is expected to increase the ability of people to recognize authentic and fake news.