Bayesian models: the structure of the world, uncertainty, behavior, and the brain

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry
• Decision Making

Background:

• Uncertainty arising from unreliable or incomplete information demonstrably impacts behavior in humans and animals.
• Recent research has formalized uncertainty, leading to various Bayesian models aimed at minimizing its behavioral effects.

Purpose of the Study:

• To analyze the differences and commonalities among various Bayesian approaches to modeling behavioral and neural data.
• To clarify the relationships between different Bayesian models used in cognitive science and neuroscience.
• To provide an overview of theories on how the brain represents uncertainty.

Main Methods:

• Utilized the framework of graphical models to systematically compare and contrast different Bayesian approaches.
• Reviewed existing behavioral and neural data linked to specific Bayesian models.
• Synthesized information on how these models can be interconnected.

Main Results:

• Identified key distinctions and shared features across diverse Bayesian models of behavior and neural activity.
• Established relationships between previously disparate Bayesian modeling techniques.
• Summarized empirical evidence supporting various Bayesian models.

Conclusions:

• Graphical models offer a unified perspective on Bayesian approaches to understanding uncertainty in behavior and the brain.
• This framework facilitates a clearer understanding of the relationships between different computational models of cognition.
• The review highlights the ongoing quest to elucidate the neural mechanisms underlying uncertainty representation.