Inferring a Cognitive Architecture from Multitask Neuroimaging Data: A Data-Driven Test of the Common Model of...

Area of Science:

• Cognitive neuroscience
• Computational psychiatry
• Neuroimaging analysis

Background:

• Cognitive architectures offer theoretical blueprints for understanding the mind's structure and predicting brain activity.
• Existing models, like the Common Model of Cognition, successfully link cognitive architectures to functional MRI data but rely on top-down approaches.
• Top-down methods limit the exploration of alternative architectures, restricting comprehensive theory development.

Purpose of the Study:

• To develop and validate a novel bottom-up methodology for inferring cognitive architectures directly from neuroimaging data.
• To overcome the limitations of purely top-down approaches in cognitive architecture research.
• To integrate bottom-up findings with existing cognitive models to refine theories of cognition.

Main Methods:

• Applied Granger causality modeling to task-based functional MRI (fMRI) data.
• Inferred a network of causal connections between brain regions based on functional connectivity.
• Utilized a combined top-down and bottom-up modeling strategy.

Main Results:

• The inferred network of causal brain connections largely aligns with the Common Model of Cognition.
• The data-driven network suggests additional connections not present in the Common Model, particularly those related to episodic memory.
• Demonstrated the feasibility of a bottom-up approach to cognitive architecture inference.

Conclusions:

• A bottom-up method using Granger causality on fMRI data can effectively infer cognitive architectures.
• This approach reveals network structures that complement and extend established cognitive models.
• The findings support a hybrid modeling strategy for advancing computational theories of cognition and memory.