Strategies for memory-based decision making: Modeling behavioral and neural signatures within a cognitive...

Area of Science:

• Cognitive Science
• Neuroscience
• Computational Modeling

Background:

• Humans often make inferences about real-world objects using memory.
• Decision-making involves integrating recognition memory with additional knowledge.

Purpose of the Study:

• To investigate how people use memory for object inferences.
• To compare recognition-based, knowledge-retrieval, and lexicographic memory strategies.
• To model behavioral and neural predictions using the ACT-R cognitive architecture.

Main Methods:

• Developed computational models of three memory strategies within the ACT-R architecture.
• Conducted a functional magnetic resonance imaging (fMRI) study where participants inferred city size.
• Compared model predictions to behavioral (response times) and neural (BOLD responses) data.

Main Results:

• Lexicographic memory strategies, which conditionally search knowledge, best explained observed data.
• These strategies accounted for joint patterns in response times and blood-oxygen-level-dependent (BOLD) responses.
• The findings suggest an adaptive interplay between recognition and additional knowledge in memory.

Conclusions:

• The lexicographic strategy provides a robust model for real-world object inferences.
• Implementing decision-making models within cognitive architectures yields valuable behavioral and neural predictions.
• Memory retrieval is an adaptive process balancing recognition and knowledge utilization.