This paper proposes a novel framework using real-time cognitive load metrics from multimodal physiological signals to detect and mitigate bias in human-AI collaboration, addressing the gap between human cognitive state monitoring and bias detection.
Key findings
Leverages EEG, pupillometry, and EDA for real-time cognitive load metrics.
Addresses the gap between human cognitive state monitoring and bias detection.
Enables proactive trust calibration and decision support in human-AI teams.
Establishes theoretical foundations and implementation guidelines for cognitive-aware AI systems.
Limitations & open questions
The proposed framework requires rigorous experimental validation.
Further research is needed to fully understand the causal relationship between cognitive load, trust calibration, and team performance.