Information Spread with Error Correction

TL;DR

This paper investigates how local error-correction can mitigate the spread of misinformation in networks, revealing that even minimal noise can significantly impair the accuracy of dispersed information in early stages.

Contribution

It introduces a simple model analyzing information spread with errors and local correction, highlighting limitations of error correction under noise.

Findings

Tiny noise levels can cause the average opinion to be nearly uncorrelated with the truth.

Local error-correction has limited effectiveness in early stages under significant noise.

Proposes methods for agents to gauge information stabilization.

Abstract

We study the process of information dispersal in a network with communication errors and local error-correction. Specifically we consider a simple model where a single bit of information initially known to a single source is dispersed through the network, and communication errors lead to differences in the agents' opinions on this information. Naturally, such errors can very quickly make the communication completely unreliable, and in this work we study to what extent this unreliability can be mitigated by local error-correction, where nodes periodically correct their opinion based on the opinion of (some subset of) their neighbors. We analyze how the error spreads in the "early stages" of information dispersal by monitoring the average opinion, i.e., the fraction of agents that have the correct information among all nodes that hold an opinion at a given time. Our main results show that even with significant effort in error-correction, tiny amounts of noise can lead the average opinion to be nearly uncorrelated with the truth in early stages. We also propose some local methods to help agents gauge when the information they have has stabilized.