Author Date

2024-03-14

Degree Name

Department

Computer Science

College

Physical and Mathematical Sciences

Defense Date

2024-03-06

Publication Date

2024-03-14

First Faculty Advisor

Christophe Giraud-Carrier

Second Faculty Advisor

Quinn Snell

First Faculty Reader

Carl Hanson

Honors Coordinator

Seth Holladay

Keywords

Deepfake, misinformation, deception, fake videos, generative AI

Abstract

This thesis is focused on deepfakes, a new term given to fake videos and images generated by deep learning algorithms and models. Deepfakes pose a considerable threat to society by raising the bar for quality in misinformation while also lowering the amount of skill and effort required. Deepfakes threaten to undermine democratic societies by swaying public opinion through misinformation. While many researchers are working hard to develop automated tools to combat deepfakes, this thesis used a 10-item IRB approved survey to examine whether two separate interventions could successfully improve an individual’s ability to recognize deepfakes. Demographic differences in recognizing deepfakes was also explored. The results of the survey found that while younger participants responded positively to interventions, older participants reacted adversely to interventions. Older participants also performed significantly worse at recognizing deepfakes.