Impact of Dimensionality of User Interaction Subspace on Multi-dimensional Opinion Formation

This paper proposes a multi-dimensional opinion formation model for online social networks. Conventional multidimensional opinion formation models do not take into account the "curse of dimensionality" triggered by higher-dimensional state spaces. In addition, conventional models are static in that topics are predetermined and do not change. On the other hand, the proposed model avoids these problems by using word distribution representation when constructing a subspace of user interactions that corresponds to the topic at that point in time. We evaluate the impact of the dimensionality of the interaction subspace on opinion formation through simulations, and show that the concentration of discussion on a particular topic encourages opinion formation.