Community Fact-Checks Trigger Moral Outrage in Replies to Misleading Posts on Social Media

Research Background and Issues

• Identified Challenges and Issues:

  • Misinformation spreads rapidly on social media due to a lack of editorial oversight, potentially misleading the public and posing serious threats to social stability and critical areas such as democratic elections.
  • Traditional expert-based fact-checking, while reliable, requires significant resources and has limited coverage, making it inadequate to address the vast amount of misinformation. Additionally, users often exhibit low trust in expert fact-checking.
  • Community fact-checking (e.g., "Community Notes" on Platform X) leverages collective efforts, but existing research primarily focuses on its behavioral impact on misinformation dissemination, without delving into its effects on users' emotional and moral responses.

• Significance of the Issue:

  • Understanding the specific emotional impacts of community fact-checking on users can help improve the information exchange environment, balancing free expression on social media with misinformation control. This understanding is also crucial for enhancing trust in fact-checking technologies.

• Research Motivation and Related Work:

  • Existing studies indicate that Community Notes effectively reduce the spread of false content but have yet to clarify the causal relationship between these notes and users' moral outrage or other emotional responses.
  • Emotions (e.g., anger and disgust) play a critical role in driving collective actions and condemning misinformation, but they may also escalate polarization and conflict on social media.

Solution

• Methods and Solutions:

  • This paper adopts a causal analysis framework to explore the impact of displaying "Community Notes" on emotional and moral outrage expressions in social media replies.
  • The data integrates Community Notes, misleading posts, and over 2.2 million direct replies from Platform X.
  • Natural language processing (NLP) models are used to analyze emotional and moral responses, combined with Regression Discontinuity Design (RDD) to evaluate the results.

• Innovations:

  • For the first time, the emotional driving effects of community fact-checking are analyzed from a fine-grained emotional perspective.
  • Negative emotions (e.g., anger, disgust) and heightened moral outrage triggered by Community Notes are quantitatively measured.
  • Differences in emotional responses to political versus non-political misleading content are explored, providing insights for designing fact-checking systems in complex information environments.

• Implementation Steps and Techniques:

  1. Data Collection: Using Platform X's API and public Community Notes data, a time-series dataset of misleading posts and their direct replies is constructed.
  2. Emotion Analysis: Employing state-of-the-art NLP models, such as the highly persuasive Twitter-roBERTa, for emotion classification (positive/negative) and extraction of Ekman's six basic emotions (e.g., anger, disgust).
  3. Causal Inference: Using RDD to identify emotional changes before and after Community Notes are displayed, measuring the actual impact of fact-checking.
  4. Sensitivity Analysis: Further dissecting the differences in emotional responses to political and non-political misleading information.

Research Findings

• Specific Findings:

  • Community Notes significantly altered the emotional tone of replies: negative emotions increased by 7.3%, anger rose by 13.2%, and disgust by 4.7%. Additionally, moral outrage increased by 16%.
  • These emotional driving effects were more pronounced in misleading posts related to political topics, particularly in terms of disgust.
  • The stronger the emotional tone of the original post, the more likely replies exhibited emotional congruence, indicating a significant emotional contagion effect among users.

• Advantages Over Existing Approaches:

  • The fine-grained emotional analysis provides more specific evidence for information control, revealing patterns and logic in users' emotional shifts.
  • Broader impact validation demonstrates that community fact-checking not only reduces misinformation dissemination but also triggers strong emotional reactions among users, potentially reinforcing social accountability for content creators.

• Experimental or Evaluation Results:

  • Across multiple experimental models and time windows, the study confirmed the causal link and robustness of emotional response trends.
  • Validation of moral outrage targets showed that highly emotional replies were primarily directed at misleading posts and their authors, rather than random discussions or platform moderation.
  • Even when comparing different types of misleading content (e.g., political vs. non-political), the effects of Community Notes remained consistent.

• Limitations and Future Directions:

  • The data spans only the first four months after the launch of "Community Notes," suggesting future research should extend to longer periods to study long-term impacts.
  • The study focuses solely on English content; future research could explore cross-cultural and multilingual analyses to capture variations in emotional expression across cultural contexts.
  • The current study does not explain the specific impact of moral outrage on content creators' deletion behavior, requiring further quantitative experiments.

Conclusion

The study demonstrates that community fact-checking profoundly influences the conversational environment on social media from an emotional perspective. Future interventions should aim for a balanced design to address potential dual effects, such as enhancing public scrutiny while avoiding unnecessary polarization. Additionally, broader applicability should be tested in multilingual and long-term scenarios.