Presentation
Three Attempts to Improve Human Deepfake Detection Performance
SessionCYB1: Cyber Deception
DescriptionOur previous work has showed that individuals differ in the extent to which they are able to detect/identify deepfake videos (videos in which the human subjects of the video has been fabricated or manipulated using artificial intelligence). Being unable to detect/identify these fabrications can leave one susceptible to being deceived or misinformed. It is therefore of great importance to develop techniques for reliably enhancing their ability. The present work is a report on the efficacy of three candidate techniques. Namely, interventions that: asked participants to focus on particular regions of the observed video subject's faces; gave participants feedback on the correctness of their judgements after each trial; and incentivized accurate judgements through a so-called "get out early" manipulation. None of the interventions yielded any significant effect on deepfake-detection performance.
Contributor
Doctoral Student
Event Type
Lecture
TimeTuesday, September 10th10:30am - 10:45am MST
LocationFLW Salon I
Cybersecurity