Quantifying Emotional Specificity and Ambiguity in Emojis: An Entropy Based Analysis of Discrete Emotion Ratings

Agents4Science 2025 Conference Submission114 Authors

Keywords: Emojis, emotion specificity index, Ambiguity and Specificity

Abstract:

Emojis are ubiquitous in digital communication, yet their emotional meanings are
often ambiguous. A recently released normative data set provides mean ratings
of 112 emojis on 13 discrete emotions in Spanish speakers. Based on studies
demonstrating that many emoji do not unambiguously depict a single emotion,
we introduce an entropy-based emotion specificity index (ESI) to quantify how
concentrated the ratings of an emoji in one emotion. After baseline correction,
we compute Shannons entropy across the 13 emotion ratings and normalize it by
the maximum possible entropy. Low values of ESI indicate ambiguous or neutral
emojis, whereas high values reflect a strong association with a single emotion.
Our analyzes reveal that negative emojis show greater specificity than positive or
neutral ones, that the principal component analysis recovers a valence continuum
explaining nearly 73 % of the variance, and that ESI is systematically related
to affective valence. We discuss applications of the ESI in marketing, health
communication, and mental health monitoring and situate our findings within
emerging normative datasets and crosscultural research on emoji interpretation.

Submission Number: 114