Towards a Positive Thinking About Deepfakes: Evaluating the Experience of Deepfake Voices in the Emotional and Rational Scenarios

This study conducted a prospective exploration to investigate humans’ experiences with self-deepfaked voices, specifically the differences between AI-generated, self-deepfaked, and authentic human voices in the hedonic and pragmatic contexts. Our result shows that, despite participants consistently preferring human voices across all tasks, their experiences between authentic human voice, self-deepfaked, and AI-generated voices, and between hedonic and pragmatic contexts can be different. Self-deepfaked voices outperformed in hedonic scenarios, providing enriched listening experiences. On the contrary, self-deepfakes and AI-generated voices exhibit their different potentials in pragmatic contexts. The findings could inspire the potential application of deepfakes to create a proper social experience between humans and AI.