Trustworthy Affective Computing
My research focuses on the deep integration of Artificial Intelligence and Psychological Behavioral Science, centered around the core theme of "Trustworthy Affective Computing." I have proposed and implemented a comprehensive psychological understanding system that spans from data collection to model construction, and from input security to output trustworthiness. This work advances emotion recognition from laboratory paradigms reliant on identity and explicit facial/speech signals toward de-identified, ethically-aware, psychologically-consistent, and practically-oriented general recognition frameworks.
Paradigm Shift in Emotion AI
From identity-dependent, explicit signal-based laboratory paradigms to de-identified, ethically-constrained, psychologically-grounded real-world recognition frameworks
The Three-Stage Trustworthy Framework
1
Input Authenticity & Security
Ensuring the genuineness of emotional input signals through robust verification mechanisms
2
Psychology-Guided Causal Reasoning
Transforming from black-box classification to psychologically-constrained causal understanding
3
Trustworthy Output & Ethical Control
Ensuring AI emotional feedback maintains ethical consistency and psychological credibility
Complete Trustworthy Loop: This three-in-one approach establishes a comprehensive framework from input source control through reasoning process modeling to output result correction and ethical control, providing theoretical support and technical pathways for sustainable development of AI emotion recognition.
Research Impact & Applications
🧠
Mental Health
Objective assessment tools for psychological well-being and emotional disorders through privacy-preserving monitoring
🤖
Human-Computer Interaction
Ethical and psychologically-aware interactive systems with trustworthy emotion understanding
🎓
Education
Adaptive learning systems responsive to student emotional states and engagement levels
🛡️
Public Safety
Reliable emotional state monitoring in security-critical environments while preserving individual privacy
Key Innovations
Data Foundation: Introduced identity-free micro-gesture and micro-movement video data systems, establishing the first research paradigm for micro-behavior analysis and long-term emotional motivation tracing that breaks through the limitation of focusing only on instantaneous emotions.
Technical Breakthrough: Systematically embedded psychological constraints in AI reasoning processes, achieving the leap from "black-box data classification" to "causal emotion understanding under psychological constraints," significantly enhancing model interpretability, controllability, and psychological consistency.
Ethical Framework: Pioneered the exploration of emotional "hallucinations" in large models, constructing correction mechanisms and ethical intervention modules based on causal chain constraints to ensure AI emotional feedback maintains ethical consistency and psychological credibility.