Emoji Reactions on Telegram Often Reflect Social Approval Over Emotional Resonance

2025-08-08

AI Summary

This study investigates the core pragmatic function of emoji reactions on Telegram—whether they primarily signal social affiliation or emotional resonance. Method: Analyzing 650,000 messages and their corresponding emoji reactions, we employed a multimodal annotation framework integrating lexicon-based sentiment analysis, large language models (LLMs), and manual coding to infer message sentiment polarity, emotion categories, rhetorical strategies, and speech acts. Contribution/Results: We find pervasive affective mismatch: positively valenced reactions (e.g., 👍, ❤️) dominate across all sentiment classes and exhibit no significant correlation with original message sentiment; a robust positive bias persists cross-contextually. These results systematically challenge the conventional assumption that emoji reactions directly reflect emotional feedback, instead confirming their primary role as markers of social approval. Moreover, we identify, for the first time, specific rhetorical and pragmatic strategies that significantly predict reaction frequency—offering theoretical refinements to digital pragmatics and methodological insights for social media sentiment analysis.

Technology Category

• Natural Language Processing: Discourse, Pragmatics & Argument Mining
• Multiagent Systems: Agent Communication
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Web Mining and Content Analysis: Sentiment analysis and opinion mining
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience

Abstract

Emoji reactions are a frequently used feature of messaging platforms. Prior work mainly interpreted emojis as indicators of emotional resonance or user sentiment. However, emoji reactions may instead reflect broader social dynamics. Here, we investigate the communicative function of emoji reactions on Telegram by analyzing the relationship between the emotional and rhetorical content of messages and the emoji reactions they receive. We collect and analyze over 650k Telegram messages that received at least one emoji reaction. We annotate each message with sentiment, emotion, persuasion strategy, and speech act labels, and infer the sentiment and emotion of emoji reactions using both lexicons and large languages. We find a systematic mismatch between message sentiment and reaction sentiment, with positive reactions dominating even when the message is neutral or negative. We show that this pattern remains consistent across rhetorical strategies and emotional tones, suggesting that emoji reactions may signal a degree of social approval rather than reflecting emotional resonance. Finally, we shed light on the communicative strategies that predict greater emoji engagement. These findings have methodological implications for sentiment analysis, as interpreting emoji reactions as direct proxies for emotional response may be misleading.

Problem

• Emoji reactions indicate social approval, not emotional resonance.
• Mismatch between message sentiment and reaction sentiment exists.
• Emoji engagement linked to communicative strategies, not emotions.

Innovation

• Analyze Telegram messages with emoji reactions
• Annotate messages with sentiment and emotion labels
• Use lexicons and large languages for sentiment inference