 |
| October 2026 |
 |
78 |
|
| 專輯論文Special Issue Articles |
| 有依據就可信嗎？探討引用信源對人工智慧生成內容（AIGC）可信度感知的影響 |
| When Transparency Backfires: Source Citations, Confirmation Bias, and the Perceived Credibility of AI-Generated Misinformation |
|
|
|
 |
(9)
|
| 作者 |
閔晨、趙攀、余文婷 |
| Author |
Chen MIN, Pan ZHAO, Wenting YU |
| 關鍵詞 |
生成式人工智慧、引用信源可信性、AIGC可信度感知、AIGC虛假錯誤資訊 |
| Keywords |
generative artificial intelligence, source credibility, perceived credibility of AIGC, AI-generated misinformation |
| 摘要 |
生成式人工智慧的快速發展加劇了全球範圍的資訊失序風險，如何提高公眾對人工智慧生成虛假資訊的辨識能力，已成為社會各界亟待解決的重要議題。近年來， New Bing、文心一言等平台透過為人工智慧生成內容（artificial intelligence generated content, AIGC）標註引用信源來增強內容透明度與可信度。然而迄今為止，尚未有研究探討提供引用信源能否有效提升公眾對AI生成內容的辨識能力，特別是當面對AI可能生成的虛假錯誤資訊時。基於啟發式－系統加工模型（heuristicsystematic model, HSM）和信源可信度理論，本研究考察了AIGC內容正確性、引用信源可信性、以及態度一致性對個體感知AIGC可信度的作用。透過兩項組間線上實驗（N = 848），研究發現：引用信源會提高AIGC虛假資訊的誤導性。在無引用來源時，公眾對正確資訊的可信度感知顯著高於錯誤資訊；而在有引用來源時，正確資訊與錯誤資訊的可信度感知差異不顯著。此外，當AIGC為虛假錯誤資訊時，提供高可信度引用來源反而會提升其可信度感知。同時，確認偏誤效應在公眾接觸AIGC時依然顯著存在。研究結論對AIGC的治理具有重要的理論與實踐啟示。 |
| Abstract |
The rapid development of generative artificial intelligence (AIGC) has exacerbated the risk of information disorder worldwide, raising an urgent question of how to improve the public’s ability to detect AI-generated misinformation. In recent years, platforms such as New Bing and Ernie Bot have attempted to enhance the transparency and credibility of AIGC by providing source citations. However, little is known about whether citing sources effectively improves the public’s ability to discern AI-generated content, particularly when facing fabricated or misleading information. Drawing on the heuristic-systematic model (HSM) and source credibility theory, this study examines the effects of content accuracy, source credibility, and attitude congruence on individuals’ perceived credibility of AIGC. Based on two between-subjects online experiments (N = 848), the findings show that source citations increase the misleading potential of AIGC. Without citations, participants rated accurate information as significantly more credible than inaccurate information; with citations, this difference became nonsignificant. Moreover, when AIGC presented false information, citing highly credible sources further increased perceived credibility. In addition, confirmation bias remained robust when individuals processed AIGC. These findings provide important theoretical and practical implications for the governance of AIGC.
本文引用格式：
閔晨、趙攀、余文婷（2026）。〈有依據就可信嗎？探討引用信源對人工智慧生成內容（AIGC）可信度感知的影響〉。《傳播與社會學刊》，第78期，頁79–110。
Citation of This Article:
Min, C., Zhao, P., & Yu, W. (2026). When transparency backfires: Source citations, confirmation bias, and the perceived credibility of AI-generated misinformation. Communication and Society, 78, 79–110. |
|
|
 |
| No.77 2026 July |
 |
| No.76 2026 April |
 |
| No.75 2026 January |
 |
| No.74 2025 October |
 |
| No.73 2025 July |
|
