Abstract
Various discrete emotions are expressed on Twitter which mirror the course of the 2014 Ebola outbreak. This study demonstrates that geographic distance from locations where Ebola cases are confirmed significantly influences the expression of discrete emotions, providing a nuanced view of public sentiment during a global health crisis.
Introduction
Public health crises are as much behavioral as they are biological. The effectiveness of safety protocols and public health interventions is heavily influenced by the emotional state of the population. Traditionally, research has treated "negative emotion" as a monolithic category or focused on static snapshots of sentiment. This paper addresses the need for a granular, dynamic understanding of public sentiment by investigating how discrete emotions—fear, anger, sadness, and positive sentiment—co-evolve over the course of an outbreak and how geographic proximity to the threat alters these emotional profiles.
Methodology
The researchers utilized a hybrid approach combining manual content analysis with large-scale data processing. A stratified random sample of 21,000 original tweets from 30 countries throughout 2014 was manually coded into categories: Positive, Neutral, and Negative (further decomposed into Fear, Anger, and Sadness). This human-coded "gold standard" sample was then used to categorize an additional 424,700 associated retweets, resulting in a total corpus of 435,700 tweets.
To analyze spatial dynamics, countries were categorized into three tiers:
- Confirmed: Countries with confirmed Ebola cases.
- Adjacent: Countries geographically adjacent to confirmed cases.
- Remote: Geographically distant countries.
Results and Discussion
Overall Sentiment and the Role of Information
The analysis revealed that the majority of discourse was informational rather than emotional. Approximately 62.5% of tweets were classified as "neutral," consisting primarily of shared news headlines. This underscores Twitter's role as a primary news wire during health crises. Of the remaining sentiment-bearing tweets, 27.9% were negative and 9.6% were positive. Within the negative category, fear was the dominant emotion (61.3%), followed by anger (34.2%) and sadness (4.5%).
Temporal Dynamics of Emotion
Emotions were not static but fluctuated significantly in response to the progression of the outbreak. Different emotions reached their peaks at different stages of the crisis.
As shown in Figure 1, sadness peaked early in July, while anger exhibited two distinct peaks in August and December. Fear reached a massive singular peak in October 2014, coinciding with the height of global concern. Interestingly, as the crisis progressed and survival stories emerged, positive emotions like encouragement and relief began to blend with the prevailing negative sentiments.
The Proximity Effect
Geographic distance played a critical role in shaping emotional responses. Fear and anger were significantly more prevalent in countries with confirmed cases (18.1% fear, 10.5% anger) compared to remote countries (12.4% fear, 6.0% anger). Conversely, sadness was slightly higher in remote countries (1.5%) than in confirmed-case countries (1.2%), suggesting a "distant sympathy" effect where those further from the immediate danger felt more sorrow for the victims. Positive sentiment was also higher in adjacent (11.1%) and remote (10.5%) regions compared to confirmed-case countries (9.1%).
Conclusion and Future Directions
This study highlights the "blended" nature of public emotional response during an epidemic. While fear is the primary driver, other emotions like anger and sadness follow distinct temporal and spatial patterns. A limitation of this study is its heavy reliance on U.S. data (52.6%), which may skew the findings toward Western perspectives. Furthermore, while the study maps emotional shifts, it does not directly link these emotions to specific health behaviors. Future research should investigate how these mixed emotional states influence cognitive and behavioral outcomes, such as adherence to safety protocols, to help public health officials design more effective communication strategies.
Summary updated