Detecting and Measuring Online Emotional and Moral Reactions to Offline Events

Core Concepts

The rich and dynamic information environment of social media provides opportunities to learn about social phenomena in a timely manner. This work presents a method for systematically detecting and measuring emotional and moral reactions to offline events using change point detection on the time series of collective affect, and further explaining these reactions using a transformer-based topic model.

Abstract

The authors present a pipeline to detect, measure, and explain online emotional and moral reactions to offline events. They first perform emotion and morality detection from text using state-of-the-art transformer-based models. They then construct the time series of the aggregate affect on a daily basis and perform change point detection to identify significant reactions. To explain the detected reactions, they apply topic modeling to the tweets around the change points.
The authors demonstrate the utility of their method on a corpus of tweets from a large US metropolitan area between January and August 2020, a period with significant social, political, and cultural changes. They successfully detect reactions to major events like the COVID-19 pandemic and the Black Lives Matter protests, as well as smaller events like earthquakes and sports games.
The authors quantify the short-term and long-term changes in emotions and moral sentiments around these events. For example, the COVID-19 outbreak triggered increases in negative emotions like anger, disgust, and fear, as well as more moral sentiments related to care and harm. The BLM protests were associated with complex reactions, including increases in negative emotions and moral concerns about fairness and betrayal.
The authors also show the importance of disaggregating topics when studying specific issues. For COVID-19, they find that aggregating emotions from all tweets can give misleading impressions, as positive emotions were mostly expressed in tweets about leisure activities rather than directly COVID-related topics.

Stats

The COVID-19 pandemic was associated with a short-term increase in anger, disgust, fear, and sadness, as well as a long-term increase in these negative emotions.
The COVID-19 pandemic was associated with a short-term and long-term decrease in positive emotions like joy and love.
The COVID-19 pandemic was associated with a short-term and long-term increase in moral sentiments related to care and harm.
The BLM protests were associated with a short-term and long-term increase in negative emotions like anger and disgust, and a decrease in positive emotions.
The BLM protests were associated with a short-term and long-term increase in moral sentiments related to fairness, cheating, loyalty, betrayal, and subversion.

Quotes

"The rich and dynamic information environment of social media provides researchers, policy makers, and entrepreneurs with opportunities to learn about social phenomena in a timely manner."
"We demonstrate that our method is able to disaggregate topics to measure population's emotional and moral reactions. This capability allows for better monitoring of population's reactions during crises using online data."
"Our results suggest that studying the collective emotional reactions on social media can provide valuable insights into understanding people's opinions and responses to timely socio-political events, and aid policy makers in crafting messages that align with the values and concerns of the population."

Key Insights Distilled From

Measuring Online Emotional Reactions to Events

by Siyi Guo,Zih... at arxiv.org 04-01-2024

https://arxiv.org/pdf/2307.10245.pdf

Deeper Inquiries

How can the insights from analyzing online emotional and moral reactions be used to inform policy decisions and crisis response strategies?

Analyzing online emotional and moral reactions can provide valuable insights for informing policy decisions and crisis response strategies. By understanding the collective affect of a population in response to events, policymakers can gauge public sentiment, identify areas of concern, and tailor their messaging to align with the values and emotions of the population. For example, during a crisis such as the COVID-19 pandemic, monitoring online emotional reactions can help policymakers assess the effectiveness of their communication strategies, identify areas of high emotional distress or misinformation, and adjust their response accordingly.
Moreover, by detecting and measuring emotional and moral reactions to specific events, policymakers can anticipate public reactions to policy changes, social initiatives, or crisis management strategies. This proactive approach allows for more targeted and effective interventions, as policymakers can address concerns, alleviate fears, and promote positive emotions within the population. Additionally, understanding the moral sentiments expressed online can help policymakers align their policies with the ethical values and principles of the community, fostering trust and cooperation.
In essence, insights from analyzing online emotional and moral reactions can serve as a real-time barometer of public sentiment, guiding policymakers in making informed decisions, crafting effective communication strategies, and implementing crisis response measures that resonate with the emotional and moral landscape of the population.

What are the potential limitations or biases in using social media data to measure population-level emotions and morality, and how can these be addressed?

Using social media data to measure population-level emotions and morality comes with several potential limitations and biases that need to be addressed to ensure the accuracy and reliability of the insights derived from such analyses. Some of these limitations include:

Selection Bias: Social media users may not be representative of the entire population, leading to biases in the data. Users who are more active on social media platforms may have different emotional expressions and moral sentiments compared to those who are less active or not present on social media.

Self-Selection Bias: People choose what to share on social media, which may not always reflect their true emotions or moral values. Users may selectively post content that aligns with a certain image they want to portray, leading to skewed data.

Algorithmic Bias: The algorithms used to analyze social media data may introduce biases based on the training data, model design, or data preprocessing techniques. These biases can impact the accuracy of emotion and morality detection.

Contextual Bias: Social media posts lack context, making it challenging to interpret the true meaning behind the expressions of emotions and moral sentiments. Without context, misinterpretations and misattributions can occur.

To address these limitations and biases, researchers and analysts can:

Implement robust sampling techniques to ensure the representativeness of the social media data.
Validate the emotion and morality detection models using diverse datasets and cross-validation techniques.
Incorporate context-aware analysis to better understand the nuances of emotional and moral expressions in social media posts.
Combine social media data with other sources of information, such as surveys or interviews, to triangulate findings and enhance the validity of the results.
By acknowledging and mitigating these limitations and biases, researchers can improve the accuracy and reliability of using social media data to measure population-level emotions and morality.

What other types of offline events or phenomena could be studied using a similar approach to understand their psychological and social impact on the public?

The approach of analyzing online emotional and moral reactions to understand the psychological and social impact of events can be applied to a wide range of offline events and phenomena. Some examples include:

Natural Disasters: Studying social media data to analyze emotional reactions and moral sentiments following natural disasters such as hurricanes, earthquakes, or wildfires can provide insights into the psychological impact on affected communities, resilience strategies, and recovery efforts.

Political Events: Analyzing online reactions to political events such as elections, policy changes, or diplomatic crises can help in understanding public sentiment, political polarization, and the effectiveness of government communication strategies.

Cultural Events: Studying emotional and moral reactions to cultural events like art exhibitions, music festivals, or heritage celebrations can shed light on the cultural significance, community engagement, and social cohesion fostered by such events.

Health Crises: Examining online emotional and moral responses to health crises beyond the COVID-19 pandemic, such as outbreaks of infectious diseases, mental health awareness campaigns, or vaccination drives, can provide insights into public perceptions, healthcare behaviors, and community resilience.

Social Movements: Analyzing social media data to understand emotional and moral reactions to social movements, protests, or advocacy campaigns can offer insights into public support, activism strategies, and the impact of social change initiatives on society.

By applying a similar approach to a diverse range of offline events and phenomena, researchers can gain a deeper understanding of the psychological and social dynamics at play, inform evidence-based interventions, and contribute to the development of more effective crisis response strategies and policy decisions.