Document Type
Conference Proceeding
Publication Date
11-13-2024
Published In
CSCW Companion '24: Companion Publication Of The 2024 Conference On Computer-Supported Cooperative Work And Social Computing
Abstract
Scholars, politicians, and journalists have raised alarm over the potential for AI-generated photos, video, and audio—often referred to as deepfakes—to reduce trust in one another and our institutions. Despite these clarion calls, little empirical work exists on how deepfakes are being used to harm individuals outside of non-consensual intimate imagery (NCII). This research provides a preliminary analysis of 50 wide-ranging incidents of deepfake harm. We find that the most common types of harm are relational, systemic, financial, and emotional. Apart from AI-generated NCII, the most prevalent uses of deepfakes to cause harm were instances of mis- and disinformation, fraud, and misrepresentation of or stereotyping about marginalized groups (e.g., women and racial minorities). We concluded with recommendations for future work and discuss potential challenges in identifying, quantifying, and preventing harm caused by deepfakes both online and off.
Published By
Association for Computing Machinery
Conference
CSCW '24: The 27th ACM Conference On Computer-Supported Cooperative Work And Social Computing
Conference Dates
November 9-13, 2024
Conference Location
San José, Costa Rica
Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 International License.
Recommended Citation
Paulina Trifonova , '25 and Sukrit Venkatagiri.
(2024).
"Misinformation, Fraud, And Stereotyping: Towards A Typology Of Harm Caused By Deepfakes".
CSCW Companion '24: Companion Publication Of The 2024 Conference On Computer-Supported Cooperative Work And Social Computing.
533-538.
DOI: 10.1145/3678884.3685938
https://works.swarthmore.edu/fac-comp-sci/125
