Distributed Fake News Detection
Abstract
World Economy Forum’s report names Digital Wildfires as one of the major challenges threatening the global economy. These digital wildfires and misinformation campaigns account for a variety of societal, economical, educational, and even health-related adverse impacts. One way to address these challenges is to use experts for reviewing the available new pieces of information. In addition to being a costly solution, such a solution is not scalable, as for example, 500 million tweets are shared daily on Twitter. On the flip side, the growth of the internet has brought a large number of individuals who show an interest in a particular subject-context. For example, the popular website Reddit that hosts forums on a diverse set of topics, has 1.2 million subreddits-forums and some of these subreddits have millions of subscribers. In particular, just the News subreddit has about 25 million subscribers. The main idea of this proposal is to provide a mechanism that utilizes these massive, distributed, but unreliable, users to provide a reliable assessment of news-information. This project aims at an extensive research program to propose and study distributed fake news detection using inexpert fact-checkers. In a nutshell, our novel approach utilizes the availability of such agents to evaluate the validity of each statement arriving from a news stream, and dynamically learn about each agent’s reliability. To do so, we propose a research program consisting of four thrusts that start from distributed fact-checking under simple model assumptions and progressively become more realistic and complex. In Thrust I, we study distributed fake news detection for non-strategic and non-interacting (inexpert) agents. In Thrust II, we provide alternative approaches to this problem and identify fundamental limits in addressing this problem. In Thrust III, we study this problem in the context of non-strategic but interacting (inexpert) agents, where agents interact with each other over a network and influence their neighbors’ opinions. Finally, in Thrust IV, we propose to study distributed fact-checking in a network of interacting and strategic agents, where either a subset of the agents strategically mislabel the news or the network designer strategically varies the network to prevent the spread of fake news within the network. For our study, we utilize, generalize, and invent a variety of tools in probability theory (in particular, the martingale and stochastic approximation theory), stability of nonlinear dynamics, information theory, statistics, graphical models, network science, and game theory.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 29, 2024
- Source ID
- FA95502310057
Entities
People
- Behrouz Touri
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California, San Diego