Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Social Convos: Capturing Agendas and Emotions on Social Media

Ankita Bhaumik, Ning Sa, Gregorios Katsios, Tomek Strzalkowski

TL;DR

This paper tackles detecting deliberate influence operations in social media by focusing on convo-based groups and two key indicators: agenda control and emotional language. It introduces a top-down workflow that uses instruction-tuned large language models to extract agendas, entities, promoted actions, and emotions from focused convo subsets, coupled with network analysis of top influencers. The authors validate the approach on the 2022 French presidential Twitter dataset, analyzing #frexit and #covid_19 convos to reveal influencer networks and the distribution of agendas and emotions. The work demonstrates that dense cross-influencer retweet networks and coherent agenda-emotion signals can signal targeted operations, while also acknowledging limitations and ethical considerations for real-world deployment.

Abstract

Social media platforms are popular tools for disseminating targeted information during major public events like elections or pandemics. Systematic analysis of the message traffic can provide valuable insights into prevailing opinions and social dynamics among different segments of the population. We are specifically interested in influence spread, and in particular whether more deliberate influence operations can be detected. However, filtering out the essential messages with telltale influence indicators from the extensive and often chaotic social media traffic is a major challenge. In this paper we present a novel approach to extract influence indicators from messages circulating among groups of users discussing particular topics. We build upon the concept of a convo to identify influential authors who are actively promoting some particular agenda around that topic within the group. We focus on two influence indicators: the (control of) agenda and the use of emotional language.

Social Convos: Capturing Agendas and Emotions on Social Media

TL;DR

This paper tackles detecting deliberate influence operations in social media by focusing on convo-based groups and two key indicators: agenda control and emotional language. It introduces a top-down workflow that uses instruction-tuned large language models to extract agendas, entities, promoted actions, and emotions from focused convo subsets, coupled with network analysis of top influencers. The authors validate the approach on the 2022 French presidential Twitter dataset, analyzing #frexit and #covid_19 convos to reveal influencer networks and the distribution of agendas and emotions. The work demonstrates that dense cross-influencer retweet networks and coherent agenda-emotion signals can signal targeted operations, while also acknowledging limitations and ethical considerations for real-world deployment.

Abstract

Social media platforms are popular tools for disseminating targeted information during major public events like elections or pandemics. Systematic analysis of the message traffic can provide valuable insights into prevailing opinions and social dynamics among different segments of the population. We are specifically interested in influence spread, and in particular whether more deliberate influence operations can be detected. However, filtering out the essential messages with telltale influence indicators from the extensive and often chaotic social media traffic is a major challenge. In this paper we present a novel approach to extract influence indicators from messages circulating among groups of users discussing particular topics. We build upon the concept of a convo to identify influential authors who are actively promoting some particular agenda around that topic within the group. We focus on two influence indicators: the (control of) agenda and the use of emotional language.
Paper Structure (20 sections, 3 figures, 4 tables)

This paper contains 20 sections, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Identification of Convos
  4. Identification of the Top Influencers
  5. Agenda and Emotion Detection
  6. Experiments
  7. Dataset
  8. Implementation
  9. Results
  10. #frexit convo (Fig \ref{['fig:frexit_clusters']}):
  11. #covid_19 convo (Fig. \ref{['fig:covid_clusters']}):
  12. Analysis and Discussion
  13. Related Work
  14. Agenda and Emotion Detection
  15. Conclusion
  16. ...and 5 more sections

Figures (3)

  • Figure 1: Overview of our approach to extract and characterize convos.
  • Figure 2: Influencer Networks among top influencers in each convos. Each node (red point) represents an influencer, with her index marked below the node. The arrows indicate the direction of the connections and the numbers above show the weight of the connections. A balloon on top of a node means a self-RT link.
  • Figure 3: Different clusters in the convos about #Frexit and #covid_19. They are characterized by the indicators of top entities, promoted actions, and emotions. The influencers are the top 10 retweeted authors in that convo. The expanded network of the influencers are illustrated in Fig. \ref{['fig:influencer_frexit']} and \ref{['fig:influencer_covid']}. Entities are grey blobs and promoted actions are the grey edge labels. Emotion labels over the edges are colored as negative/positive ones. The model also generates an agenda summary, given in Table \ref{['tab:agenda-summary']}.