Find the Real: A Study of Individuals’ Ability to Differentiate Between Authentic Human Faces and Artificial-Intelligence Generated Faces

Abstract

As advances in Artificial-Intelligence (AI) generated technology accelerate, much of the general population is unaware when they may come across AI-generated content, including AI-generated images of humans. In attempting to see if the general population can ascertain the real vs. fake as society encounters an increasing amount of deep fake imagery, this study was conducted using AI human portraits generated at thispersondoesnotexist.com [1] using SyleGan2 (a generative adversarial network developed by Nvidia) [2]. The deep fake portraits were then juxtaposed with authentic portraits of humans from the Flickr Faces HQ Dataset (FFHQ) of Creative Commons licenses and public domain images [3]. Survey respondents were asked to select which face they believed to be an authentic human face. This is an important study in the realm of Human-Computer Interaction as AI resources are utilized in a variety of forms, including the spread of information, and can be linked to disinformation with the implementation of deep fakes.