Abstract:Nowadays, social media is the ground for political debate and exchange of opinions. There is a significant amount of research that suggests that social media are highly polarized. A phenomenon that is commonly observed is the echo chamber structure, where users are organized in polarized communities and form connections only with similar-minded individuals, limiting themselves to consume specific content. In this paper we explore a way to decrease the polarization of networks with two echo chambers. Particularly, we observe that if some users adopt a moderate opinion about a topic, the polarization of the network decreases. Based on this observation, we propose an efficient algorithm to identify a good set of K users, such that if they adopt a moderate stance around a topic, the polarization is minimized. Our algorithm employs a Graph Neural Network and thus it can handle large graphs more effectively than other approaches
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: how to reduce the polarization in social networks with two echo chambers. Specifically, the author proposes an effective algorithm based on Graph Neural Network (GNN) to identify a group of users (K users). When these users take a neutral attitude towards a certain topic, the polarization degree of the entire network can be minimized.
### Problem Background
In today's society, social media has become the main platform for political debates and opinion exchanges. Research shows that social media is highly polarized, forming the so - called "echo chamber" structure, that is, users only establish connections with like - minded people and only consume specific types of content. This phenomenon makes users' views more extreme and reduces their openness to different views.
### Specific Problem Description
The author focuses on how to reduce network polarization by changing the opinions of certain users. They base on the following observation: if some users take a neutral attitude towards a certain topic, then the polarization degree of the network will be reduced. Therefore, the author proposes an optimization problem: given a social network and the initial opinions of its users, how to select K users so that after these users take a neutral attitude, the polarization degree of the network is minimized.
### Mathematical Model
According to Friedkin and Johnsen's opinion evolution model, each node \( i \) has two opinion values:
- Internal opinion \( s_i \), representing the user's inherent belief.
- External opinion \( z_i \), representing the opinion expressed by the user in the social environment.
The update rule for the external opinion \( z_i \) is:
\[ z_i=\frac{s_i + \sum_{j \in N(i)} w_{ij}z_j}{1 + \sum_{j \in N(i)} w_{ij}} \tag{1} \]
where \( N(i) \) represents the neighbor set of node \( i \), and \( w_{ij} \) is the weight of edge \((i, j)\).
The polarization index \( \pi(z) \) is defined as:
\[ \pi(z)=\frac{\|z\|_2^2}{|V|} \tag{3} \]
where \( \|z\|_2^2 \) is the squared norm of the opinion vector \( z \), and \( |V| \) is the number of nodes.
### Solution
In order to find the optimal K users, the author proposes an algorithm based on GNN. The core idea of this algorithm is to use GNN to predict the reduction in network polarization after setting the external opinion of a certain user to zero. The specific steps are as follows:
1. **Offline part**: Train a GNN model so that it can predict the reduction in network polarization (Gain) after setting the external opinion of a certain node to zero. This can be achieved through supervised learning, using synthetic data sets for training.
2. **Online part**: In practical applications, use the trained GNN model to gradually select K users so that after these users take a neutral attitude, the polarization degree of the network is minimized.
### Experimental Results
The experimental results show that the GNN - based algorithm performs equivalently to the traditional Greedy algorithm (GreedyExt) in reducing network polarization, but has a significant improvement in computational efficiency. For large - scale networks, the GNN algorithm can find a set of users close to the optimal solution in a shorter time.
### Summary
This paper provides an effective method to reduce the polarization in social networks with two echo chambers by introducing the GNN model. This method not only has equivalent performance to traditional methods, but also has obvious advantages in computational efficiency, and is suitable for the analysis and optimization of large - scale networks.