Cyber violence caused by the disclosure of route information during the COVID-19 pandemic

The study addresses cyber violence (CV) associated with government disclosure of infected individuals’ travel route information during COVID-19, termed route information disclosure-caused cyber violence (RIDCCV). While route disclosure is widely used for epidemic control (e.g., China, Japan), it can heighten risks of privacy leakage and online harassment as disclosures often include demographic details and movement histories and are disseminated via official sites and social media. Prior research shows CV is prevalent globally and is facilitated by social media; CV increased during COVID-19 amid isolation and anxiety. Yet, limited research has examined the specific link between route disclosure and CV. The authors argue that necessary public health disclosures must be balanced against privacy rights and social harms. They analyze 13 RIDCCV cases and conduct in-depth social media content and network analyses of six cases to investigate: RQ1: What were the social media contents of the six RIDCCV cases? RQ2: Were there differences among the six cases regarding social media contents? RQ3: Who plays a vital role in disseminating information on social media?

The literature indicates CV encompasses online behaviors that incite or enact violence and can produce harms comparable to offline violence, including fear and suicidality. Social media intensifies CV due to anonymity, scale, and reduced moral sensitivity; gendered CV is notably prevalent. During COVID-19, studies report increased cyberbullying linked to isolation and psychological strain, with CV mediating depression and coping. Information disclosure by authorities can unintentionally prompt discrimination and abuse toward infected individuals due to perceived threat. Prior work in South Korea highlighted privacy risks in public disclosures (including route data) but did not analyze CV outcomes. Rumor propagation, uncertainty, and issue salience on social media further fuel CV, with threat/effectiveness messaging influencing diffusion. Agenda-setting by mainstream and influential we-media shapes public attention and attitudes.

Case identification and coding: The authors monitored social media for RIDCCV incidents in China (Nov 2020–Nov 2021), identifying 13 representative cases. Three trained doctoral researchers reconstructed each case via web searches and coded six features: (a) violation of epidemic regulations, (b) first locally confirmed case status, (c) personal information leakage, (d) nicknaming, (e) rumors, (f) moral criticism.

Six cases with the largest online discussion volume were selected for in-depth analysis and renamed RCase 1, 2, 3, 4, 5, 6 (original Cases 1, 2, 4, 7, 9, 13). They split into two groups: violators of regulations (RCase 2, 5, 6) vs non-violators (RCase 1, 3, 4).

Data collection: Platform: Weibo. Tool: Octopus web crawler. For each RCase, specific keyword queries and time windows captured original posts and associated comments:

• RCase1: “girl” AND “COVID-19” AND “Pixian County,” 2020-12-07 to 2020-12-16; 63,719 posts (final clean: 62,888)
• RCase2: “old lady Yin” AND “COVID-19” AND “Shenyang,” 2020-12-23 to 2021-01-18; 47,913 (45,538)
• RCase3: “Wuhan” AND “COVID-19,” 2021-08-05 to 2021-08-11; 13,840 (12,558)
• RCase4: “Harbin” AND “COVID-19” AND “LARP Games,” 2021-09-21 to 2021-09-29; 17,037 (15,817)
• RCase5: “Xi’an” AND “COVID-19,” 2021-10-17 to 2021-10-24; 21,896 (21,073)
• RCase6: “Chengdu” AND “COVID-19,” 2021-11-05 to 2021-11-12; 13,114 (12,291) Collected fields: username, timestamp, content. Data cleaning used SPSS 25.0; ads/invalid records removed.

Text mining:

• Sentiment analysis: Bi-LSTM (forward+backward LSTM). Word segmentation: Jieba with Sougou Pinyin lexicon. Embedding: Word2Vec. Hyperparameters: hidden size=300; learning rate=0.01. Training data: 120,000 Weibo posts (50% positive/50% negative). Manual validation: 5,000 posts per RCase checked by three students; average accuracy=89.77%.
• Topic clustering: LDA with α=0.1, β=0.01. Number of topics chosen via trial-and-error. Preprocessing: Jieba segmentation; TF-IDF weighting.

Social network analysis: Built weighted, directed user commenting networks from posts and comments. Importance measured via weighted PageRank (damping=0.85). Users classified using Weibo industrial category into: mainstream/official media, influential we-media (>300,000 followers), we-media, and normal users, following prior classification schemes.

Across 13 RIDCCV cases (Table summaries):

• 11/13 victims were the first locally confirmed case of a wave.
• 11/13 involved personal information leakage.
• Rumors and moral condemnation frequently accompanied route disclosures.

Six-case content analysis (sentiment; topics):

• Negative sentiment dominated in all RCases: RCase1 72.55%; RCase2 77.66%; RCase3 62.54%; RCase4 85.66%; RCase5 84.53%; RCase6 73.33%.
• RCase1 (non-violator): high focus on bar visits; topics included condemnation for visiting crowded places, condemnation of CV and rumor/privacy leaks, fear of outbreaks, and holiday travel worries; positive topics praised medical staff, compliance, ending pandemic, and reasonable disclosure.
• RCase2 (violator): focus on quarantine breach leading to infections; negative topics condemned the woman and feared outbreaks; positives praised medical workers/government and hoped for quick end.
• RCase3 (non-violator): keywords included playboy, privacy, epidemiological survey; negatives condemned CV, leaks, rumors, disease spread; positives praised medical workers/government and hoped for end.
• RCase4 (non-violator): focus on LARP games; negatives included holiday travel worries, condemnation of CV/leaks, and fear; positives hoped for end, praised medical staff, and event-related humor.
• RCase5 (violator): focus on Xi’an/Jiayuguan and ‘abnormal’ nucleic acid; negatives condemned the couple’s violations and CV, worried about disease, accused governance/leaks; positives included hope for end, calls to improve control, acceptance of couple’s apology after clarification.
• RCase6 (violator): negatives condemned the person (calls for conviction), holiday travel worries, and panic; positives hoped for end, praised government measures, and saluted medical workers.

Group differences (violator vs non-violator):

• Violator cases drew stronger condemnation of individuals (sometimes calls for punishment). Non-violator cases saw more condemnation of CV, rumor-spreading, and privacy leaks.
• Over time (RCase1 vs RCase2 comparisons), discourse shifted from personal attacks to denouncing doxxing/rumors and to calls for better disclosure management.

Dissemination roles (PageRank top-10 composition):

• Dominance of mainstream media and influential we-media among top nodes:
  • RCase1: mainstream 50%, influential we-media 30%
  • RCase2: mainstream 70%, influential we-media 30%
  • RCase3: mainstream 10%, influential we-media 70%
  • RCase4: mainstream 50%, influential we-media 50%
  • RCase5: mainstream 80%, influential we-media 20%
  • RCase6: mainstream 80%, influential we-media 20%

Overall: Disclosing route information increases CV risk, especially when individuals breach prevention rules. First locally confirmed cases are particularly vulnerable. Rumors and moral judgments are common triggers.

The findings answer the research questions by showing that social media discourse around RIDCCV is largely negative, with themes of blame, fear, and policy support, but also contains resistance to CV where no intentional harm is perceived. Differences between violator and non-violator cases highlight that perceived intentional wrongdoing drives moral condemnation and can escalate to CV, whereas non-violator cases elicit defenses against doxxing and rumor-mongering. Temporal analyses indicate discourse can evolve from early moralizing/personal attacks to critiques of privacy breaches and calls for better disclosure governance. Network analysis underscores mainstream and influential we-media as agenda-setters and key disseminators, shaping public opinion and potentially mitigating rumor spread if used responsibly. These results suggest a formation mechanism for RIDCCV in which route disclosures trigger moral judgment (especially amid pandemic anxiety), rumors, and group polarization, amplifying online attacks. Balancing the public health benefits of route disclosure with privacy protection and responsible media amplification is crucial to reduce CV risk, especially for first locally confirmed cases.

This study is among the first to systematically examine cyber violence arising from governmental route information disclosures during COVID-19. Using 13 cases (with six analyzed via text and network mining), it identifies common risk features (first locally confirmed status, personal information leakage) and demonstrates that mainstream and influential we-media dominate information diffusion. It shows that perceived intentional violations catalyze CV, while non-violator cases evoke resistance to CV and privacy breaches. The paper proposes governance recommendations: (1) promote effective, fact-based publicity through mainstream media and influential we-media to guide public opinion and curb rumors; (2) optimize route information collection/disclosure systems (e.g., “mention routes, not people”), strengthen training, and enhance privacy protection; (3) address public anxiety with psychological support and improved cyberspace governance. These insights can inform crisis communication and digital governance beyond pandemics. Future work should expand case coverage across periods, incorporate patient characteristics, and examine inter-agency coordination and crisis stages for CV governance.

• The analysis did not stratify by patient characteristics (e.g., gender, age, occupation) that may influence CV.
• The six focal cases cover a specific period; public focus likely shifts across pandemic stages, affecting CV dynamics.
• Additional factors such as types of government departments involved and crisis stages were not examined; building and evaluating interdepartmental, collaborative CV governance—especially in early crisis phases—requires further study.