Learning Automata-based Misinformation Mitigation via Hawkes Processes

Area of Science:

• Computer Science
• Social Network Analysis
• Information Science

Background:

• Social media misinformation poses a significant challenge due to complex information spread.
• Multivariate Hawkes Processes (MHP) are vital for modeling social network dynamics and evaluating mitigation strategies.

Purpose of the Study:

• To propose a novel, lightweight, intervention-based misinformation mitigation framework.
• To control Multivariate Hawkes Processes (MHP) using decentralized Learning Automata (LA).

Main Methods:

• Developed a framework where each Learning Automaton (LA) controls a user's involvement in mitigation.
• Employed a joint random walk over the state space for LA interaction with MHP.
• Evaluated the approach on three Twitter datasets, including a new COVID-19 dataset.

Main Results:

• Demonstrated fast convergence of the proposed framework.
• Showcased increased exposure to valid information.
• Confirmed effectiveness across various network structures, including those with central misinformation sources.

Conclusions:

• The decentralized LA-controlled MHP framework offers an effective solution for social media misinformation mitigation.
• The approach facilitates distributed deployment due to its decentralized nature.
• Results indicate improved information quality and faster mitigation policy evaluation.