Uncovering strategies for identifying deepfakes

Abstract

Introduction. The proliferation of generative artificial intelligence tools capable of producing high-quality videos that can masquerade as genuine content has raised concerns about online misinformation. This study investigates human ability to identify deepfake videos, with a focus on identification performance and the strategies employed.

Method. Data was collected through an online survey. Participants were young adults aged 21 to 35. They were shown four videos and asked to identify them as real or deepfake, followed by questions about the identification strategies used.

Results. Our results revealed the diverse range of strategies utilised. Predominant strategies centre around assessing the authenticity of traits pertaining to the video's subject as opposed to peripheral details. Furthermore, we uncovered preferences for intuition and strategies that relate to individual decision-making over consulting other individuals or online materials.

Conclusion. Our results help enhance understanding of how people identify deepfake videos, adding to existing knowledge. These findings also inform initiatives aimed at educating the public about spotting deepfakes.