Accessible Machine Learning for Misinformation and Influence Operation Analysis
Abstract
Approved for Public Release Advancements in the field of Artificial Intelligence (AI) and Machine Learning (ML) have allowed for rap,id improvement of capabilities within nearly all industries. Employment of these technologies is undoubtedly a core component of the, modernization of the United States Department of Defense s (DoD) and Intelligence Community (IC) capability set. However, there are, unique challenges that the DoD faces in regards to the implementation of traditional AI/ML paradigms -- the impact and consequences, that the standard problems of data sensitivity, cost, and time-to-deploy are multiplied when applied within the context of DoD/IC m,issions. -This is particularly salient in the case of analyzing influence operations and misinformation campaigns. Over the past sev,eral years, the United States has witnessed the grave effects of both such phenomena, whether it be misinformation surrounding COVID,-19, or targeted influence operations seeking to delegitimize the democratic process. The USN, and the DoD more broadly, need tools, that are rapidly deployable and improvable, as well as accessible to ensure both operational efficacy and compliance with the DoD s, Ethical AI Principles. We propose the use of programmatic labeling and weakly-supervised ML as part of this toolkit to combat the g,rowing threat that influence operations and misinformation pose to the United States and its national security interests.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 16, 2022
- Source ID
- N000142212426
Entities
People
- Christopher Ré
Organizations
- Office of Naval Research
- Stanford University
- United States Navy