Decoding a Cognitive Performance State From Behavioral Data in the Presence of Auditory Stimuli

Area of Science:

• Cognitive Neuroscience
• Computational Psychology
• Statistical Modeling

Background:

• Cognitive performance state is an unobserved variable crucial for understanding cognitive functions.
• Accurate decoding requires informative observation vectors and adaptive models.
• Traditional methods may overestimate performance, especially with fast incorrect responses.

Purpose of the Study:

• To decode the hidden cognitive performance state using a Bayesian state-space approach.
• To introduce and evaluate a marked point process (MPP) framework for performance decoding.
• To compare MPP-based decoding with autoregressive (AR) and adaptive heteroskedasticity (AR-ARCH) models.

Main Methods:

• Utilized a Bayesian state-space framework to decode cognitive performance.
• Applied a marked point process (MPP) to model observations, including correct/incorrect responses and reaction times.
• Compared MPP with AR and AR-ARCH models using simulated and n-back experimental data.

Main Results:

• The Bayesian state-space approach effectively decodes cognitive performance.
• MPP-based and ARCH-based performance estimations outperformed AR-based estimations at the individual level.
• The ARCH-based performance decoder showed superior performance on aggregated data analysis.

Conclusions:

• The Bayesian state-space approach offers a promising method for decoding cognitive performance.
• The MPP framework provides an effective way to model behavioral observations for performance decoding.
• Developed decoders have potential applications in educational settings and smart workplaces for performance monitoring and feedback control.