The Language of Influence: Sentiment, Emotion, and Hate Speech in State Sponsored Influence Operations

2025-05-12

AI Summary

This study addresses how state-sponsored influence operations (SIOs) exploit sentiment, emotion, and hate speech to manipulate public opinion and disseminate disinformation. It presents the first cross-national comparative analysis of official SIO activities on Twitter by China, Iran, and Russia. Leveraging 1.5 million tweets, the methodology integrates fine-grained sentiment analysis, Plutchik’s eight-dimensional emotion taxonomy, hate/toxicity detection, and statistical significance testing. Results reveal distinct discursive strategies: Russia predominantly employs negative emotional arousal; Iran hybridizes positive and negative emotions to exacerbate polarization; and China emphasizes positive narrative construction. The study identifies interpretable “discursive fingerprints” with an F1-score of 0.89 for country-level classification. These findings provide the first linguistically grounded, cross-cultural empirical basis for automated SIO attribution and intervention.

Technology Category

• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Application Domains: Humanities & Computational Social Science
• Knowledge Representation and Reasoning: Preferences

Application Category

• Web Mining and Content Analysis: Sentiment analysis and opinion mining
• Social Networks and Social Media: Media and governance, opinion dynamics, filter bubbles, polarization
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

State-sponsored influence operations (SIOs) on social media have become an instrumental tool for manipulating public opinion and spreading unverified information. This study analyzes the sentiment, emotion, and abusive speech in tweets circulated by influence campaigns originating from three distinct countries: China, Iran, and Russia. We examined 1.5 million tweets to uncover patterns in the content of the influence operations using the dataset provided by Twitter. Our findings reveal distinct patterns of sentiment, emotion, and abusive nature in different SIOs. Our experimental result shows that Russian influence operations predominantly employ negative sentiment and toxic language to polarize audiences, Iranian operations balance negative and positive tones to provoke hostility while fostering support, and Chinese campaigns focus on positive messaging to promote favorable narratives.

Problem

• Analyzing sentiment in state-sponsored social media influence operations
• Identifying emotion patterns in tweets from China, Iran, Russia campaigns
• Detecting abusive language differences between geopolitical influence strategies

Innovation

• Analyzed sentiment, emotion, hate speech in tweets
• Examined 1.5 million tweets from three countries
• Revealed distinct patterns in influence operations