Affective computing is an interdisciplinary research field focused on enabling machines to recognize, interpret, and generate human emotions, thus enriching human-computer interaction. Advances in multimodal analysis—leveraging visual, auditory, textual, and physiological signals—have opened new possibilities in emotion-aware systems. Recent developments in Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) have further enhanced the capacity to model and respond to emotions with unprecedented nuance. These technologies are increasingly applied in domains such as virtual assistants, storytelling, immersive environments, social robotics, and healthcare. However, the integration of affective intelligence into AI systems raises crucial challenges related to explainability, privacy, bias mitigation, and fairness. This special issue is linked to the 4th edition of the Automatic Affect Analysis and Synthesis (3AS) Workshop, which brings together researchers and practitioners to advance affective computing while addressing these pressing ethical and technical challenges.

The proposed Research Topic aims to address the dual challenges of Affective Understanding and Affective Generation in AI systems. Affective Understanding involves developing models capable of accurately detecting and interpreting human emotions from multimodal inputs while ensuring transparency, privacy, and fairness. This requires robust methodologies for emotion recognition in diverse contexts and the application of explainable AI techniques to increase system trustworthiness.

Affective Generation focuses on creating emotionally resonant content or responses that can adapt to human affective states. The rise of LLMs and MLLMs offers new opportunities for generating sophisticated, context-aware emotional outputs in applications ranging from conversational agents to immersive entertainment. The Research Topic seeks to foster contributions that tackle these challenges by advancing algorithms, data collection, methodologies, evaluation metrics, and application frameworks. By uniting theoretical and practical perspectives, it aims to accelerate the development of equitable, privacy-conscious, and human-centered affective systems.

This Research Topic welcomes original research, surveys, and application papers covering, but not limited to:

- Emotion recognition from large-scale multimodal datasets;
- Sentiment analysis with deep learning techniques;
- Affective computing in generative AI applications;
- Use of LLMs and MLLMs for affective computing;
- Bias mitigation and fairness in emotion-aware AI;
- Privacy-enhancing techniques for emotional data processing;
- Explainable AI in affective analysis;
- Emotion-aware reinforcement learning;
- Applications in robotics, human-robot interaction, virtual reality, assistive technologies, and entertainment.

We encourage submissions that present novel datasets, benchmarks, algorithms, and system evaluations. Manuscripts may address theoretical foundations, technical innovations, or applied solutions, with an emphasis on reproducibility, ethical considerations, and interdisciplinary relevance. Please note that submitting a manuscript summary is not mandatory for the submission of a full manuscript; it is merely a recommendation by Frontiers.

Are you interested in contributing a manuscript to this Research Topic?

Advances in Automatic Affect Analysis and Synthesis