Loading...

Media is loading
 

Paper Number

2101

Paper Type

Short

Abstract

The rapid proliferation of deepfake images in the context of the upcoming US general elections has generated significant anxiety, given the potential of these photorealistic fakes to undermine the democratic process. Most mitigation measures have focused on the computational detection of anomalies in these images' endogenous features, such as faces, eyes, and hair. However, many such algorithms suffer from low accuracy rates or high input costs. In this study, we adopted a human-led approach to help the general public discover exogenous cues from deepfake images. We developed instruments of inquiry based on the four distinct conceptualizations of the theories of Incongruity of Humor and adopted the Delphi method to provide a validated and ranked list of detectable exogenous cues. The results from the pilot study suggest the utility of exogenous cues of the types Atypical and Juxtaposition in deepfake detection.

Comments

05-SocImpact

Share

COinS
 
Dec 15th, 12:00 AM

Unmasking the Phantom: Discovering Exogenous Cues to Detect Political Deepfake Images

The rapid proliferation of deepfake images in the context of the upcoming US general elections has generated significant anxiety, given the potential of these photorealistic fakes to undermine the democratic process. Most mitigation measures have focused on the computational detection of anomalies in these images' endogenous features, such as faces, eyes, and hair. However, many such algorithms suffer from low accuracy rates or high input costs. In this study, we adopted a human-led approach to help the general public discover exogenous cues from deepfake images. We developed instruments of inquiry based on the four distinct conceptualizations of the theories of Incongruity of Humor and adopted the Delphi method to provide a validated and ranked list of detectable exogenous cues. The results from the pilot study suggest the utility of exogenous cues of the types Atypical and Juxtaposition in deepfake detection.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.