Opinion dynamics model based on the cognitive dissonance: An agent-based simulation

Highlights

• We develop an opinion dynamics model based on the cognitive dissonance.

• We propose a new mechanism for updating the connections.

• We design the simulation experiments with three initial opinion distributions.

• We compare the experimental results with different initial opinion distributions.

Abstract

In opinion dynamics, some agents may exhibit the cognitive dissonance behaviors owing to the contradictory belief, attitudes and opinions that they might confront with. This may greatly influence the evolutions of opinions and connections. To this end, this paper proposes an opinion dynamics model based on the cognitive dissonance (ODCD). In the ODCD model, with the consideration of bounded confidence effects, the methods for updating the opinions and network of agents are provided, respectively. Then, we design the simulation experiments with different initial opinion distributions to investigate the influences of bounded confidences and initial connection probabilities. Finally, we conduct a comparison analysis on the opinion dynamics with different initial opinion distributions.

Introduction

Opinion dynamics as a potent tool is used to investigate the dynamics of the public opinions in various fields [1], [2] such as, the emergence of extremism in the field of sociology [3], the emergence of new language in the field of linguistic [4] and the election in the field of politics [5]. As a result, the researches on the opinion dynamics have been gaining the favour for several decades.

The earliest study on opinion dynamics can be dated back to 1956 when French initiated the opinion dynamics model [6]. Following French's study, researchers with different backgrounds have proposed various models to analyze the evolution of opinions. For example, some studies extended one dimension opinion space into the multiple dimension opinion space [7], [8], [9]. Some studies assumed the opinion dynamics occurred in the specified network [10], [11], [12]. Some studies proposed the opinion dynamics model involved with the noise [13], [14], [15]. Some studies discussed the consensus reaching in opinion dynamics [16], [17], [18], [19], [20]. Some studies focused on the applications of opinion dynamics models [21], [22], [23], [24], [25].

Among all these studies, bounded confidence model is one of the most prevalent approaches. There exist two important bounded confidence models, which are independently proposed by Hegselmann and Klause (i.e., HK model) [26] and Deffuant, Weisbuch and others (i.e., DW model) [27], [28]. According to the HK and DW models, the researchers conducted some interesting research results. For example, the stability and convergence of bounded confidence model is studied [29], [30], the consensus threshold of the bounded confidence model [31], [32], the opinion dynamics model with group-based bounded confidence [33], the opinion dynamics model with different communicated number [34], and the noisy opinion dynamics [35].

Previous studies have made significant contributions on opinion dynamics. However, there are still some limitations as follows:

• In practical opinion dynamics, some agents may exhibit the cognitive dissonance behaviors. For example, some agents will reject the interactions of agents whose opinions have a larger difference from their own opinions, and build more connections with agents whose opinions are similar to their own opinions. Thus, the cognitive dissonance behaviors should be considered in opinion dynamics.

• The existing opinion dynamics models lack rationality in updating the connections. Currently, two main categories of updating the connections [36], [37], [38] have been proposed. In the first category, new connections are built between agents holding similar opinions. However it is harder for an agent to obtain the opinions of others when there are no connections between them. In the second category, new connections are randomly built. This may lead to the agents may build the connections with others whose opinions have a larger difference from their opinions.

• Simulation experiments with uniform distribution of initial opinions of agents are commonly conducted. However, in some practical opinion dynamics, there are always some agents expressing the extreme opinions in the initial time. More complexly, some agents will express the heterogeneous extreme opinions in the initial time.

In order to overcome the limitations mentioned above, this paper develops the opinion dynamics model based on the cognitive dissonance (ODCD). In the ODCD model, the methods for updating the opinions and network of agents are provided, respectively. Then, we design the simulation experiments with different initial opinion distributions to investigate the influences of bounded confidences and initial connection probabilities. Furthermore, a comparison analysis on the opinion dynamics with different initial opinion distributions is conducted.

The reminder of this study is arranged as follows: Section 2 introduces the preliminary regarding the opinion dynamics, the HK bounded confidence model and the cognitive dissonance theory. In Section 3, the opinion dynamics model based on the cognitive dissonance is proposed. In Section 4, simulation experiments with different initial opinion distributions are conducted. Finally, concluding remarks are included in Section 5.