初探生成式AI風險感知的成因與第三人效果：以ChatGPT為例

Exploring the Antecedents of Risk Perception toward Generative AI and Its Third-Person Effect: A Case Study of ChatGPT

隨著ChatGPT等生成式AI愈來愈融入人類生活，人工智慧科技對於人類生活的負面影響也值得關注。本研究以第三人效果（the third-person effect）作為理論基礎，一方面探討民眾的ChatGPT使用經驗、AI素養對生成式AI風險第三人感知的影響，另一方面也檢驗上述因素是否影響其支持政府規範AI科技，並將樣本群體劃分為「X世代」、「千禧世代」及「Z世代」，剖析不同世代對生成式AI風險的認知與反應。研究以《2024台灣網路報告》的數據為分析資料（N = 979），結果發現，三個世代的民眾普遍展現出第三人效果傾向，認為AI科技風險對他人的影響遠大於對個人的影響；不過，ChatGPT使用經驗在三個世代中均無法顯著預測第三人感知，而AI素養也僅對X世代產生顯著負向影響。此外，ChatGPT使用經驗對千禧世代與X世代的法律規範態度有顯著負向影響，而AI素養對Z世代法律規範態度有顯著負向影響，至於三個世代的第三人效果則無法顯著預測對AI科技法律規範的態度。研究結果有助於增進了解當前台灣民眾對於生成式AI的風險認知，並可作為政府制訂相關政策時參考。

With the increasing integration of generative AI, such as ChatGPT, into daily life, the potential negative influence of this technology on human lives warrants attention. The current study employs the third-person effect (TPE) as a theoretical framework to investigate the impact of individuals’ChatGPT use experience and AI literacy on their third-person perception of generative AI risks. Furthermore, it investigates whether these factors affect the support for government regulation of AI technology. The sample was divided into Generation X, Millennials, and Generation Z to explore the variations in AI risk perceptions and responses across these generations. This study analyzed data from the 2024 Taiwan Internet Survey (N = 974). The results revealed that all three generations exhibited a tendency toward the third-person perception, perceiving that the risks of AI technology as having a greater impact on others than on themselves. However, ChatGPT usage did not significantly predict the third-person effect across any generation. AI literacy had a significant negative impact on the third-person only within Generation X. Furthermore, ChatGPT usage exerted a significant negative influence on attitudes toward AI regulation among Millennials and Generation X. In contrast, AI literacy had a significant negative impact on such attitudes among Generation Z. Notably, the third-person perception failed to exert a significant impact on attitudes toward AI regulation across three generations. These findings contribute to a better understanding of Taiwanese citizens’risk perceptions regarding generative AI and can serve as a reference for policymakers in formulating relevant policies.