“Stunning and creepy…I gotta learn this”: Exploring Concepts of Information Ethics in Deepfake Creation Tutorials on YouTube

Abstract

Deepfakes combine artificial intelligence with machine deep learning to produce manipulated images and are made to appear hyperrealistic to viewers (Westerlund, 2019), complicating, if not, preventing viewers from separating fact from fiction and eroding trust in video footage or images from reputable mainstream news sources. Previous literature has articulated the dangers of deepfakes and their dissemination. This ongoing project explores the creation of deepfakes by seeking to understand who makes deepfakes through analyzing deepfake tutorials on YouTube.  A thematic analysis of the content and production features of YouTube deepfake creation videos provides a critical perspective of how and why individuals instruct others to do so.  Findings of 73 YouTube videos indicate higher proportions non-western participation in deepfake creation. While still ongoing, these early findings impact how LIS professionals frame deepfakes around mis/disinformation and suggest that in order to overcome deepfakes we must understand them in relation to larger socio-technical discourses.