Affective Design: The Influence of Facebook Reactions on the Emotional Expression of the 114th US Congress

Title of the Paper

Affective Design: The Influence of Facebook Reactions on the Emotional Expression of the 114th US Congress

Bibliographic Information

• Subject Area: Social Computing and Political Communication
• Keywords: Social Media, Text Analysis, Functional Design, Natural Experiment, Political Communication, Negative Emotions, User Engagement

Research Background and Problem

• Problem and Challenges: Political communication on social media can exacerbate polarization due to emotional intensification, hindering constructive discussions. The introduction of affective design features may further influence the expressive style of political figures, but this issue remains underexplored.
• Significance: Emotional expression may enhance user engagement, but negative emotions (e.g., anger, disgust) can lead to adverse outcomes, such as increased polarization, weakened rational dialogue, and the spread of misinformation, posing potential threats to democratic systems.
• Research Motivation and Related Work:
  • Modern political communication often employs emotional appeals, such as historical advertisements mobilizing voters through fear.
  • Previous studies have explored the impact of emotional features on overall user behavior on social media but have rarely examined how these features alter the behavior of content creators (e.g., politicians).
  • This paper fills a research gap by investigating how Facebook Reactions' design feature influences the emotional expression of members of the 114th US Congress.

Solution

• Proposed Method or Solution:
  • This study leverages Facebook's introduction of the Reactions feature in 2016 as a natural (quasi-) experiment to analyze its impact on the emotional expression of members of the 114th US Congress.
  • The experiment employs the Difference-in-Differences (DID) method and incorporates fixed-effects regression models to analyze changes in emotional characteristics before and after the feature's introduction.
• Innovations:
  • Focuses on the impact of affective interaction design on content creators rather than limiting the analysis to audience reactions.
  • Compares data from Facebook and Instagram, using Instagram (unaffected by the feature change) as a control group to eliminate confounding background factors.
• Implementation Steps and Key Techniques:
  1. Data Collection: Extracted post data from members of the 114th Congress on Facebook and Instagram, divided into pre- and post-feature introduction groups, encompassing 172,000 posts.
  2. Data Preprocessing: Removed non-text content, performed tokenization, stopword removal, lemmatization, and other standardization processes.
  3. Sentiment Analysis: Used the dictionary-based NRCLex tool to calculate the frequency of positive and negative emotion words in posts.
  4. Model Analysis:
     • Applied the Difference-in-Differences model to evaluate the impact of Facebook's feature on emotional expression.
     • Conducted linear regression analysis to predict the influence of user interactions (e.g., likes, comments, shares) on changes in emotional expression.
  5. Control Analysis: Compared Instagram data to validate the unique impact of Facebook's feature.

Research Findings

• Specific Findings:
  1. Overall Discovery: The introduction of Facebook Reactions significantly increased the use of negative emotions in posts by members of Congress, while having no significant effect on positive emotional expression.
  2. Mechanism Analysis: A significant positive correlation was found between user engagement (e.g., likes and comments) and the expression of negative emotions. Increased user engagement over time predicted a rise in negative emotions in subsequent periods.
  3. Quantitative Results: For every 10,000 additional likes, there was an average increase of 0.09 negative emotion words. Similarly, increases in reactions such as "angry" and "sad" were also associated with growth in negative emotions.
• Advantages: Compared to the Instagram platform, which was unaffected by Facebook Reactions, these changes were more pronounced, confirming the feature's impact on behavior.
• Experiment or Evaluation Results:
  • The Difference-in-Differences model showed that, after controlling for other variables (e.g., post length, post type), Facebook's Reactions feature significantly increased the expression of negative emotions (DID coefficient = 0.08, p < 0.001).
  • In linear regression analysis, negative emotional expression was primarily driven by negative user interactions (e.g., "angry" and "sad"), while the influence of positive interactions was relatively weaker.
• Limitations and Future Directions:
  1. The data is limited to Facebook and Instagram platforms from 2015-2016, and further validation is needed to determine if the findings apply to other platforms or contexts.
  2. The analysis of user emotional responses relies primarily on the NRC dictionary; future studies could incorporate more advanced sentiment analysis tools (e.g., BERT or GPT).
  3. The study did not deeply explore the potential impact of factors such as gender and party affiliation on emotional expression, which may exhibit variations in different contexts.
  4. Broader research is needed to examine the applicability of affective design features in local politics and other domains.

Conclusion

This paper provides important insights into how the affective design of social media influences political communication, revealing the potential of design features to drive negative emotional expression. The study highlights the significant impact of platform design on user behavior and broader patterns of social interaction, offering guidance for optimizing future platform designs and communication strategies.