新聞媒體對於生成式人工智慧的感知風險、規範與實踐

生成式人工智慧（Generative Artificial Intelligence，簡稱GenAI）在2022年底商用化並影響新聞產業，促使新聞媒體機構思考如何使用，並在很短的期間內推出了相對應的規範，但規範方向是否正確、規範的目的是否明確，對新聞業與學術界都是一個尚待開拓的新領域。人工智慧有許多已知的風險，應用在新聞領域中，可能造成新聞的錯誤、偏誤、失真、幻想等等，雖然規範基於對風險的認知，提出廣泛的應對方向，但媒體是否真能夠遵循此方向與人工智慧共處，令人好奇。本研究深入探討GenAI在新聞媒體行業中的規範，特別是同業規範與機構規範的異同、實務與規範間的差距、風險與應對策略，以及臺灣媒體對於GenAI的特殊關係。本研究透過Nvivo質性分析軟體，分析了41份來自16個國家的新聞機構所制定的GenAI規範，並加上七個臺灣媒體的深度訪談結果，提供詳細的分析。本研究發現，在未來不確定的情況下，對人工智慧的感知易用性會提升感知的風險，並促使新聞機構模仿其他機構制定規範，來應對新技術的不確定性，而同業組織與媒體機構在規範目標及內容上也有顯著的差異。臺灣媒體過往較少使用人工智慧，因此面對許多新的挑戰。本研究結果驗證了制度理論中模仿同質性與規範同質性的區別，也擴充了對新聞GenAI規範的理解，為新聞機構和政策制定者提供了具體的參考方向。

The rapid proliferation of generative artificial intelligence (GenAI), notably marked by the widespread adoption of tools like ChatGPT since late 2022, has ushered in a transformative era for the news industry. This comprehensive research examines how news media organizations are adapting to this technological shift, highlighting the potential for GenAI to revolutionize news production and storytelling. It specifically examines the development and implementation of GenAI guidelines, the nuanced differences between industry-wide norms and individual and organizational practices, the practical challenges in aligning regulations with real-world applications, and the distinct concerns voiced by Taiwanese media regarding GenAI. The experiences and risk from Taiwanese media are also useful for media in other countries that only have had a chance to use AI before after this wave of generative AI before 2022. While the use of AI in journalism is not a novel concept, the advent of GenAI has significantly accelerated its adoption throughout the news industry. GenAI tools offer enticing possibilities, including heightened efficiency, innovative storytelling formats, and the potential to revolutionize news production. However, these advancements also raise apprehensions about the authenticity, transparency, and ethical dimensions of news reporting in this AI era. The literature has identified recurring themes in GenAI guidelines, such as the importance of transparency, adherence to ethical standards, and the indispensable role of human oversight in ensuring the responsible use of AI. Nevertheless, researchers still need to conduct more investigations into the practical implementation of these guidelines and the specific hurdles that media organizations face within diverse cultural and technological landscapes. This study adopts a rigorous methodological approach, analyzing 41 GenAI guidelines sourced from sixteen countries and encompassing various media entities, such as commercial and non-profit organizations, public broadcasters, news agencies, and journalist associations. To enrich the analysis and provide real-world context, researchers conducted in-depth interviews with nine practitioners representing seven Taiwanese news organizations. These interviews illuminate discrepancies between established regulations and actual practices and reveal issues that defy easy regulation. The meticulous analysis of GenAI guidelines reveals five overarching categories: ethical values, internal personnel, systems, audiences and users, and the external competitive environment. Each category is further dissected into sub-categories and specific codes, offering a holistic framework for understanding the multifaceted considerations involved in GenAI regulation.