Dynamic rumor spreading of public opinion reversal on Weibo based on a two-stage SPNR model

doi:10.1016/j.physa.2020.125005

Physica A: Statistical Mechanics and its Applications

Volume 558, 15 November 2020, 125005

https://doi.org/10.1016/j.physa.2020.125005 Get rights and content

Highlights

•
Propose a two-stage SNPR model to simulate the spread of public opinion.
•
Model considers the hysteresis of official news and people’s swing mentality based on epidemic models.
•
Simulate the spread and reversal of rumors based on SNPR.
•
Represent real situation of news breakout.

Abstract

Exploring the dynamic mechanism of rumor reversal in several public emergency events can support managers to control and guild the spread of public opinion. In this study a two-stage rumor model, namely, susceptible, positive, negative, and recovered state (SPNR) is built to analyze rumor spread and reversal of rumors regarding emergencies on Weibo. The model considers the hysteresis of official news and public swing mentality based on epidemic models. To validate the proposed model, we compare the simulated data and the spread curve of rumor reversal on the basis of the real event “Chongqing bus plunging into the river”. The result shows the proposed model well simulates this real situation of news breakout, confirming the change of rumor-infected rate and probability of rumor disseminators transforming into positive public opinion disseminators, and the time of official statement for truth affecting rumor propagation in varying degrees. To effectively control and dispel rumors, proper guidance of official statement or authority release for truth in the public event is necessary.

Introduction

Mobile Internet penetrates human daily lives as a convenient platform for acquiring and releasing information. People can acquire news about an event immediately after it transpires. Weibo, one of most important social media in China, attracts multiple netizens with its efficient sharing of text, pictures, voice, videos, and emojis. Weibo is apparently the main channel for social media news dissemination. The flexibility of sharing and forwarding blogs considerably accelerates the speed and scale of information fusion. However, the disseminated information is not always trustworthy. With the increasing speed and space of information spread, distinguishing the verity and faithfulness of facts [1] is difficult. This nature of social media provides a fertile ground for rumormongers to post and spread rumors that may result in major chaos and unpredictable reactions from involved individuals [2]. Understanding the dynamics of information spread can support the government to effectively publish actual information [3].

The release of official news regarding an incident tends to lag given the burstiness, uncertainty, and limited information. By comparison, a rumor is a story or statement whose truthfulness has not been verified [2], and thus always appears and spreads first on the Internet. Wrong, biased, and uncertain rumors often mislead the public or exert an adverse effect on social and public order. The rapid propagation of rumors through social media is a main factor in undermining social stability.

A real example of rumor reversal in a public event is the “the switch of the answer sheet in the college entrance examination” in Henan, China. Four parents questioned whether the answer sheets of test-takers were switched, which incited online public outrage. Although the rumor was debunked and reversed within several days, its spread in the early stages of the event exerted an unfortunate impact on the Department of Education and affected social stability. Such an adverse effect may have been avoided or minimized if the government has taken effective measures to control the rumor spread and guide the rumor reversal in the early stages. This example, including many similar real-life public events, shows that the government can pay more attention to the rumor spread and reversal.

This study explores the spread dynamics and reversal path of rumors in Sina Weibo regarding accidental events. We propose a new model that considers the impact of the hysteresis of official news and the “Follow the crowd” phenomenon on rumor development. The main contributions can be summarized as follows.

•
We propose a new model called two-stage SPNR (S: Susceptible infected, $P_{1}$ : positive infected (people who do not know about the truth), $P_{2}$ : positive infected (people who know about the truth) N: negative infected, and R: Removal states) model to explore the development process of breaking news. Our model optimizes the SPNR model by considering the delay characteristics of official news released by authorities. To best simulate and predict the rumor spread and reversal, we divide this process into two stages.
•
To better consider the effect of the official truth release on rumor spread, we classify the positive disseminators in the model into rational and truth disseminators based on their knowledge of the truth.
•
We add the new transformation of rational (positive disseminators who do not know the truth) disseminators into rumor disseminators to consider the “follow the crowd” phenomenon. The probability of this transformation is positively correlated with the number of rumor disseminators, which is also in accordance with reality.
•
The experimental results show that our proposed model can effectively simulate rumor spread and reversal, and helps in the understanding of rumor dynamics and then guiding its reversal.

Most current rumor propagation models are derived from the classic epidemic model [4]. A key to our understanding of epidemic theory are dynamic models [5], where the grouping of individuals in the same state can be regarded as a stock. The differential of stocks over time is the changing rate of stocks caused by state transition, which can be called a flow. Therefore, we create the model with system dynamics equations. Simultaneously, we use the classic simulation software, AnyLogic, to build and simulate the proposed two-stage SPNR model using real-life scenarios.

This paper is organized as follows. Section 2 presents a review or related literature. Section 3 describes the two-stage SPNR model, which considers the hysteresis of official news and “follow the crowd” phenomenon. Section 4 shows the model implementation using AnyLogic and simulation of opinion process in Weibo. Then the experimental results are compared with the real development process released by

Section 5 discusses the experimental results and its implications, including its theoretical and practical significance. Section 6 concludes the paper, identifies research limitations, and highlights directions for future research.

Section snippets

Rumor spreading

Online social networks play an important role in spreading information on a large scale. Information diffusion occurs when information flows from one individual or community to another [6], and relevant models are widely used in research to understand this phenomenon. Related models can be divided into two categories, namely, explanatory and predictive. The former includes epidemics and influence models, whereas the latter includes independent cascade, linear threshold, game theory [6], SIS [7]

Proposed model: Two-stage SPNR model

Given that personality traits affect false rumor beliefs [27] and reposting negative information on microblogs [28], similar to the existing SPNR model [26], we divide rumor infection into two opposite states, positive-infected and negative-infected, according to whether people spread positive opinions or rumors, respectively. The difference is the addition of a secondary classification of positive public opinion disseminators, categorizing them as rational and truth reviewers on the basis of

Data description

The incident of the Chongqing bus plunging into a river is a typical case study of rumor reversal. At approximately 10 am on October 28, 2018, a bus plunged into a river in Wanzhou, Chongqing and a red car crashed on the scene. Shortly after the accident, several media, even mainstream, issued online reports that the female driver of the red car retrograded, causing the bus to plunge into the river. The news widely spread on Weibo and was forwarded by numerous Internet celebrity and the media.

Discussion

Analysis of the parameters affecting rumor spread lead to the following conclusions.

First, the low rumor-infected rate results in strong inhibition of rumor spread and the small maximum number of rumor disseminators. Apart from not believing rumors, not spreading rumors is equally important. Therefore, netizens not believing and not spreading rumors can better suppress its spread and maintain social stability.

Second, high probability of rumor disseminators transforming into truth disseminators

Conclusions and future research

With the Internet development, the information disseminated in online social networks is increasing more than in offline channels. Several characteristics of social media can facilitate posting of unverified information and the rapid diffusion among netizens across the world. In this study, we propose the two-stage SPNR based on the original model and consider the hysteresis of official news and the “follow the crowd” to examine the rumor spread and reversal about emergencies on Weibo. The

CRediT authorship contribution statement

Guoyin Jiang: Conceptualization, Methodology, Writing - review & editing. Saipeng Li: Programing, Writing - original draft. Minglei Li: Validation, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

This work was partially supported by a grant from the National Natural Science Foundation of China (No. 71671060)

References (36)

AlkhodairS.A. et al.
Detecting breaking news rumors of emerging topics in social media
Inf. Process. Manage.
(2020)
LanY. et al.
A statistical model of the impact of online rumors on the information quantity of online public opinion
Physica A
(2020)
ZhaoL. et al.
SIR rumor spreading model in the new media age
Physica A
(2013)
ZhaoL. et al.
SIHR rumor spreading model in social networks
Phys. A
(2012)
ZhaoL. et al.
Rumor spreading model considering forgetting and remembering mechanisms in inhomogeneous networks
Physica A
(2013)
WangJ. et al.
SIRaRu rumor spreading model in complex networks
Phys. A
(2014)
ZanY. et al.
SICR rumor spreading model in complex networks: Counterattack and self-resistance
Phys. A
(2014)
WangJ. et al.
2SI2R rumor spreading model in homogeneous networks
Phys. A
(2014)
HosniA.I.E. et al.
Minimizing rumor influence in multiplex online social networks based on human individual and social behaviors
Inform. Sci.
(2020)
YaoY. et al.
Stability analysis of an SDILR model based on rumor recurrence on social media
Phys. A
(2019)

PalA. et al.

How do users respond to online rumor rebuttals?

Comput. Hum. Behav.

(2020)

HuangC. et al.

Modeling of agent-based complex network under cyber-violence

Phys. A

(2016)

GiornoV. et al.

Rumor spreading models with random denials

Physica A

(2016)

YinC. et al.

Reposting negative information on microblogs: Do personality traits matter?

Inf. Process. Manage.

(2020)

ZhaoZ.-j. et al.

An analysis of rumor propagation based on propagation force

Physica A

(2016)

XiaL.-L. et al.

Rumor spreading model considering hesitating mechanism in complex social networks

Physica A

(2015)

NekoveeM. et al.

Theory of rumour spreading in complex social networks

Physica A

(2007)

JiangG. et al.

Clicking position and user posting behavior in online review systems: A data-driven agent-based modeling approach

Inform. Sci.

(2020)

Cited by (53)

Research on cross-platform dissemination of rational and irrational information based on a two-layer network
2024, Heliyon
Numerous social media platforms have evolved into fertile grounds for the proliferation of irrational information, expanding the avenues of information dissemination. This paper initially utilized the Weibo and Bilibili social platforms as exemplars and devised a cross-platform two-layer network SEIaIbR-FXYaYbZ dissemination model grounded in classical infectious disease models. Subsequently, this research computed the model equilibrium point, basic reproduction number, and information entropy through dynamic equations. Finally, the model equations were fitted to real cases to determine optimal parameter solutions and conduct simulation analysis. The simulation results reveal that: (i) information entropy values on both platforms are low, with irrational information predominantly influencing public opinion; (ii) concerning various types of information, the augmentation of rational information results in a reduction of irrational information, while the quantity of rational information remains largely unaffected by changes in the quantity of irrational information; (iii) examining different platforms for information dissemination, alterations in the circulation rate and quantity of rational information on the Weibo platform impact the quantity of rational and irrational information on the Bilibili platform, while those changes on the Bilibili platform exert minimal influence on public opinion information on the Weibo platform. The results and corresponding strategies obtained from this study on the cross-platform dissemination of rational and irrational information on Weibo and Bilibili can provide a reference for relevant departments to guide the rational development of online information and enhance the effective management of public opinion in social media platforms.
A novel group multi-criteria sorting approach integrating social network analysis for ability assessment of health rumor-refutation accounts
2024, Expert Systems with Applications
Blooming social media platforms provide breeding ground for health rumors. Despite the establishment of accounts by numerous organizations to counter health rumors, the effectiveness of these endeavors exhibits considerable variability. Thus, there exists a pressing need to refine the framework and operation of rumor-refutation accounts. Aiming at enhancing the proficiency of accounts in refuting health rumors on social media platforms and exploring the factors affecting it, this paper proposes a novel group multi-criteria sorting approach integrating social network analysis (SNA) to classify accounts’ health rumor-refutation ability. To commence, an evaluation indicator system for accounts’ health rumor-refutation ability is established using SNA. Subsequently, the indicator values are computed, incorporating methods such as triangular fuzzy number (TFN), a lite bert (ALBERT) pre-trained language model, and PageRank. Furthermore, hesitant fuzzy linguistic term set (HFLTS) and triangular intuitionistic fuzzy number (TIFN) are used to determine the expert weights and indicator weights. After that, on the basis of original best worst method-sort (BWM-Sort), classification boundaries are discovered creatively using optimal clustering (OC), and minimum discrimination information (MDI) is adopted as the objective function for priority assignment. Consequently, an OC-MDI-BWM-Sort method is newly proposed which offers distinct advantages in computational efficiency, information integration, decision-making objectivity, and result effectiveness. Lastly, regarding to four cases of widely circulated rumors, health rumor-refutation ability of 35 accounts on Weibo platform is classified using the proposed method. The findings underscore that merely 8.57% of accounts exhibit stable and good health rumor-refutation ability, while up to 28.57% and 80.00% display poor and inconsistent ability in certain instances. Tailored to accounts with excellent or good, satisfactory or fair, and poor health rumor-refutation ability, respectively, managerial suggestions are provided regarding information expression standards, account operator proficiency, and account cooperation, and all accounts are advised to watch audience behavior.
Evolution of online public opinions on major accidents: Implications for post-accident response based on social media network
2024, Expert Systems with Applications
Based on the incredible breadth and speed of information distribution within social media network, and continuous development of natural language processing techniques for sentiment and topic analyses, it is potential to access online public opinions on major accidents. An integrated framework including four-stage evolution model, lexicon-based sentiment analysis, and latent Dirichlet allocation (LDA)-based topic extraction was developed based upon social media data analysis. The explosion of Xiangshui eco-chemical industrial zone (EXEIZ) was taken as a case for investigating the evolution of online public opinions. The large accident domain (LAD)-based sentiment dictionary was established for the lexicon-based approach for sentiment determination. Effective methods (i.e., combination of short microblogs and comments into a new text, and the hybrid approach integrating perplexity with principal component analysis) were deployed for overcoming two typical shortcomings (i.e., inappropriate for short text dataset, and sensitive to the number of topics within a corpus) of the LDA-based approach for topic analysis and extraction. According to the four-stage evolution model, customized strategies and suggestions were provided for guiding and controlling online public opinions on major accidents, in order to decrease their negative impacts on the society. This enabled conducting targeted communication efforts among different stakeholders for the reduction of negative sentiments such as indignation, sadness, and threat, and the avoidance of social media crises. Although a major accident poses tangible threats to the public, it may improve their awareness for preventing from these threats. In case of appropriate measures in time, the focus tends to steer toward effective and prosocial solutions. It is helpful for sustaining and re-establishing the image of authorities, enterprises or individuals that are closely associated with the major accident.
Online attention dynamics: The triangle framework of theory, big data and simulations
2023, Expert Systems with Applications
In computational social science era, the theory, big data and simulations should be well combined and deeply integrated, and this work serves as a typical example. Online collective actions have great impacts to societies, and we focus on the life cycle pattern of coexisting multiple online cases. We propose the analysis framework of “Theory-Big Data-simulation” (TDS). The theory includes two essential action rules of netizens. We combined macro-level analysis (big data analysis) and micro-level behavior (agent-based modeling), to surpport the theory. For the big data, macro-level life cycle pattern (persistent) is caused by micro-level individual actions. For macro-level verification, basic assumption is that netizens are allocating limited attentions to limited coexisting online events, which has been supported by our big data from the Weibo platform (N=197,985). For the micro-level, we use agent-based modeling to verify the behavioral rules of Netizens. We set two agent categories (Netizens & Hots), which behave and interact autonomously. Based on multiple-round simulations, we obtained optimal solution outcomes. Besides, we have implemented mutual validation between big data and agent-based modeling, which has not been considered before. Three big data sets have been applied to check both validity and robustness of our solutions. For the dataset A, we use 89 online cases of the Shaanxi Province in China. For the dataset B, we used 147 online cases of the whole China in 2021. For the dataset C, we used 138 online cases under specific topic (the pandemic). It suggests that three optimal solutions have both validity and robustness. As three big data sets have different levels, cases, topics and durations, the model’s generality can be well supported. Hence, general behavior pattern of the Netizens can be revealed, which is of great significance for investigating human behavior online. Moreover, our work contains rich managerial implications, which helps to improve social simulation and prediction. Also, our TDS analysis framework can support or inspire other researches.
Dynamic modeling and simulation of rumor propagation based on the double refutation mechanism
2023, Information Sciences
Rumors have been widespread with the development of self-media, which has brought new challenges to government departments in dealing with rumor management. Considering the combined effect of the external refutation of media reports and the internal refutation of the counter individuals, the susceptible-infective-counter-media-remover (SICMR) rumor propagation model with a double refutation mechanism is proposed. The two thresholds for rumor propagation are obtained by mathematical analysis of the model. Based on the Jordan matrix theory and Routh-Hurwitz judgment, two rumor-free equilibriums (RFE) and four rumor-spread equilibriums (RSE) are obtained, and the stability of the equilibria is proved. The theoretical results and the influence of key parameters are verified by numerical simulation, and the validity of the model is verified through comparative experiments. Using the simulation data for function fitting, the results showed that the initial values of the media reports and the counter individuals are exponentially related to the peak of rumor propagation, respectively. Finally, based on two real rumor datasets of Twitter, the least square method is used to fit the model parameters. Through the comparison experiment with other similar works, the R-squared of the SICMR is 0.8811 and 0.9469 on two real datasets, respectively, which is the best result.
Simulating rumor spreading and rebuttal strategy with rebuttal forgetting: An agent-based modeling approach
2023, Physica A: Statistical Mechanics and its Applications
Rumor is a typical social psychological phenomenon, and forgetting is a universal psychological trait. Exploring the role of forgetting will improve the understanding of rumor spreading and its control. Based on the classical SIR epidemic framework, the conceptual model of rumor spreading and rebuttal strategy is proposed with consideration of forgetting, and the mathematical model at the individual level is constructed from a competitive innovation diffusion perspective. A Java-based Agent-based Modeling approach is adopted to conduct comprehensive simulation experiments. Results indicate that both rebuttal strategy and rebuttal forgetting would affect the process and the result of rumor spreading and control, and complex interactions exist among different factors. Although the influence of control strategy is dominant in most cases, more active control might be not optimal in extreme cases because of the moderation of forgetting, the nonlinearity of influence increases the control challenge.

View all citing articles on Scopus

View full text