Detecting and Improving Human Cognitive State in Real-Time Using Data-Driven Adaptive Systems: A Systematic Review

Area of Science:

• Cognitive Science
• Human-Computer Interaction
• Neuroscience

Background:

• Cognitive state changes (attention, workload, alertness) impact task performance and safety.
• Real-time detection of these states offers potential for system improvement and injury risk reduction.

Purpose of the Study:

• To review existing research on sensing physiological signals, classifying cognitive states, and implementing automated interventions.
• To identify trends, limitations, and future research directions in cognitive state adaptive systems.

Main Methods:

• Systematic review of 27 studies developing models for cognitive state sensing and intervention.
• Analysis of sensor types, targeted cognitive states, classification accuracy, intervention strategies, and system latency.
• Categorization of applications including driving, education, rehabilitation, and human-robot collaboration.

Main Results:

• Electroencephalography (EEG) was the predominant sensor (70%), focusing on attention (56%) and mental workload (26%).
• High within-subject classification accuracy (81.85-95.81%) was achieved in laboratory settings.
• Common interventions included neurofeedback and task difficulty adjustment; automation adjustment was less frequent. Systems were reactive, not predictive, with limited reported latency data.

Conclusions:

• Current cognitive state adaptive systems are largely reactive and lack predictive capabilities.
• Future research should focus on detecting multiple interacting cognitive states and developing predictive models for cognitive trajectories.
• Establishing robust, real-time, predictive cognitive state adaptive systems requires further investigation into sensor technology, algorithms, and intervention strategies.