Generating a Deepfake Frame : A Text Mining Study Based on Reddit

This study investigates the understanding of deepfake, a highly realistic AI mimicry technique that is rapidly evolving to produce increasingly realistic videos and explores the construction of a deepfake framework through the lens of audience communication using framing theory. It identifies three key findings. First, the public discourse on deepfakes forms a concept hierarchy emphasising technology and its entertainment applications, with core concepts including AI, voice, actor and job, while peripheral concerns such as consent and company receive less focus. Second, employing the BERTopic algorithm, latent themes in public discussions were categorised into two dimensions: social dynamics and cultural phenomena. Third, sentiment analysis reveals predominantly neutral or negative attitudes, indicating concerns over the risks and societal impacts of deepfake technology. The deepfakes framework developed here provides a structured approach to understanding these impacts, highlighting the need for ethical considerations in technological development, regulatory measures and public education.