Affective-CARA: A Knowledge Graph Driven Framework for Culturally Adaptive Emotional Intelligence in HCI

2025-06-17

AI Summary

Current human-computer interaction (HCI) systems lack culturally adaptive emotional responses and often exhibit cultural bias. Method: We propose the first culture-adaptive affective intelligence framework for HCI, comprising: (1) a cultural affective knowledge graph built upon StereoKG, integrating the VAD emotion model with cross-cultural consistency validation; (2) a culture-aware reward strategy and sensitive narrative generation paradigm; and (3) a culture-aware response mediator that jointly orchestrates knowledge retrieval, reinforcement learning, and multi-source historical affective state fusion. Results: Evaluations on AffectNet and two other major benchmarks show a cultural semantic density of 9.32/10, a 61% reduction in cultural representation bias (KL divergence reduced to 0.28), and significant improvements in affective alignment, cultural appropriateness, and narrative generation quality.

Technology Category

• Cognitive Modeling & Cognitive Systems: Affective Computing
• Humans and AI: Emotional Intelligence
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• 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: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Culturally adaptive emotional responses remain a critical challenge in affective computing. This paper introduces Affective-CARA, an agentic framework designed to enhance user-agent interactions by integrating a Cultural Emotion Knowledge Graph (derived from StereoKG) with Valence, Arousal, and Dominance annotations, culture-specific data, and cross-cultural checks to minimize bias. A Gradient-Based Reward Policy Optimization mechanism further refines responses according to cultural alignment, affective appropriateness, and iterative user feedback. A Cultural-Aware Response Mediator coordinates knowledge retrieval, reinforcement learning updates, and historical data fusion. By merging real-time user input with past emotional states and cultural insights, Affective-CARA delivers narratives that are deeply personalized and sensitive to diverse cultural norms. Evaluations on AffectNet, SEMAINE DB, and MERD confirm that the framework consistently outperforms baseline models in sentiment alignment, cultural adaptation, and narrative quality. Affective-CARA achieved a Cultural Semantic Density of 9.32 out of 10 and lowered cultural representation bias by 61% (KL-Divergence: 0.28), demonstrating robust performance in generating ethical, adaptive responses. These findings suggest the potential for more inclusive and empathetic interactions, making Affective-CARA an avenue for fostering culturally grounded user experiences across domains such as cross-cultural communication, mental health support, and education.

Problem

• Enhancing culturally adaptive emotional responses in affective computing
• Minimizing bias in user-agent interactions with cultural knowledge
• Improving sentiment alignment and narrative quality across cultures

Innovation

• Cultural Emotion Knowledge Graph integration
• Gradient-Based Reward Policy Optimization mechanism
• Cultural-Aware Response Mediator coordination