A multi-dimensional opinion formation model for online social networks

Abstract

In this paper, we propose a multi-dimensional opinion formation model for online social networks. Previous studies have not taken into account the “curse of dimensionality” that comes from the higher-dimensional state space used in multi-dimensional opinion formation; the dimensionality is high as opinions with various types of preferences must be handled simultaneously. Conventionally, the interactions between users are limited to static models in which the topics that users discuss are predetermined and do not change. Our proposed method can avoid the above problems with previous studies by introducing low-dimensional subspaces for interacting user pairs by using word distributed representation. We propose a dynamic model that allows users to discuss various topics at each interaction. Simulations of the proposed model show that discussion focused on a particular topic encourages opinion formation.