江智浩

My research develops theoretical foundations for Human-Cyber-Physical Systems (HCPS) in safety-critical domains. To ensure safe and effective decision-making, two requirements must be met: (1) explicit mechanism-to-observation and decision-to-outcome causality; (2) real-time understanding of human operators' cognitive states and preferences.

I propose a dual-digital-twin framework resolving these challenges: the Environment Digital Twin enables counterfactual analysis for causality establishment, while the Cognitive Digital Twin infers decision-makers’ cognitive parameters and predicts their responses to decision supports. This architecture enables AI systems to complement—rather than replace—human experts in medical and automotive applications, achieving safe and effective human-machine collaboration. Crucially, decisions guided by these twins ensure decision-makers operate with the correct contextual understanding, while the AI-generated support becomes inherently more intuitive, easier to accept, and ultimately more effective in guiding the human towards optimal choices.