Decoding neural activity to assess individual latent state in ecologically valid contexts.

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Isolating cognitive processes in naturalistic settings is challenging.
• Laboratory studies of neural activity may not generalize to real-world behavior.

Purpose of the Study:

• To validate laboratory-based neural activity findings in ecologically valid contexts.
• To develop methods for inferring latent cognitive states from neural data during complex tasks.

Main Methods:

• Utilized domain generalization techniques from brain-computer interface research.
• Trained models on laboratory data and applied them to a driving simulator task.
• Decoded tonic neural activity patterns related to latent states.

Main Results:

• Neural activity patterns from laboratory settings were successfully applied to a driving simulator.
• Changes in neural activity correlated with behavioral performance, mirroring laboratory observations.
• The approach allowed for estimation of latent states during complex tasks.

Conclusions:

• Domain generalization provides a valid methodology for studying neural activity in ecologically relevant scenarios.
• This approach bridges the gap between controlled laboratory experiments and real-world cognitive processes.
• Findings support the ecological validity of traditional experimental designs.