Abstract
Since its emergence in 2017, deepfake technology has evolved from a niche innovation into a global concern with significant implications for politics, security, ethics, and privacy. Its ability to generate synthetic yet hyper-realistic content—including video, audio, text, and images—has made it a powerful tool for both creative applications and malicious activities such as disinformation, fraud, and sexual exploitation. Taiwan, which has been repeatedly targeted by deepfake-driven disinformation campaigns and non-consensual content, presents a particularly critical case for understanding how academia engages with the challenges posed by this technology. This study conducts one of the first systematic analyses of academic literature on deepfakes in Taiwan, examining the characteristics, evolution, and thematic focus of the research. Using Structural Topic Modeling (STM) and web-scraping techniques, 143 academic studies—including journal articles, master’s theses, doctoral dissertations, book chapters, and institutional reports—were analyzed to identify dominant research trends and their development over time. The results indicate that academic interest in deepfakes in Taiwan has grown rapidly since 2019, with the majority of publications written in Chinese. Ten major thematic clusters were identified, primarily focusing on detection algorithms, machine learning applications, and legal frameworks for regulating deepfakes. However, the analysis also revealed a relative lack of interdisciplinary studies addressing psychological and sociopolitical aspects—areas more prevalent in global deepfake research. Comparatively, Taiwanese scholarship demonstrates a strong emphasis on technological and legal countermeasures rather than on societal impacts or public perception. Overall, the study highlights Taiwan’s increasing but technically focused academic engagement with deepfakes and emphasizes the need for expanded cross-disciplinary collaboration. Strengthening policy-oriented, ethical, and sociotechnical research will be essential for developing comprehensive national strategies to mitigate the multifaceted risks posed by deepfake technology.
