Emergence of Hierarchical Emotion Organization in Large Language Models

2025-07-12

AI Summary

This study investigates how large language models (LLMs) intrinsically model user affective states to support ethical deployment in conversational agents. To address this, we introduce the psychological circumplex model of emotion into LLM representation analysis—integrating probabilistic dependency modeling, cross-group misclassification detection, and human-grounded behavioral experiments. Our analysis reveals that LLMs spontaneously develop a hierarchical emotion tree aligned with human psychological structure, with model scale positively correlating with emotional granularity and hierarchical depth. Critically, we identify systematic affective recognition biases against marginalized groups—particularly individuals from lower socioeconomic backgrounds. Beyond confirming the existence of human-like affective organization in LLMs, this work establishes the first cognitively grounded, theory-driven framework for interpretable emotion representation analysis in foundation models. It thus provides both theoretical foundations and empirical evidence for advancing fairness, reliability, and ethical accountability in affective AI systems.

Technology Category

• Cognitive Modeling & Cognitive Systems: Affective Computing
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

As large language models (LLMs) increasingly power conversational agents, understanding how they model users' emotional states is critical for ethical deployment. Inspired by emotion wheels -- a psychological framework that argues emotions organize hierarchically -- we analyze probabilistic dependencies between emotional states in model outputs. We find that LLMs naturally form hierarchical emotion trees that align with human psychological models, and larger models develop more complex hierarchies. We also uncover systematic biases in emotion recognition across socioeconomic personas, with compounding misclassifications for intersectional, underrepresented groups. Human studies reveal striking parallels, suggesting that LLMs internalize aspects of social perception. Beyond highlighting emergent emotional reasoning in LLMs, our results hint at the potential of using cognitively-grounded theories for developing better model evaluations.

Problem

• Understanding hierarchical emotion modeling in large language models
• Identifying biases in emotion recognition across socioeconomic personas
• Exploring cognitive theories for better model evaluations

Innovation

• Analyzing probabilistic dependencies in emotional states
• Discovering hierarchical emotion trees in LLMs
• Identifying biases in emotion recognition across personas