Abstract:The study of the brain's representations of uncertainty is a central topic in neuroscience. Unlike most quantities of which the neural representation is studied, uncertainty is a property of an observer's beliefs about the world, which poses specific methodological challenges. We analyze how the literature on the neural representations of uncertainty addresses those challenges and distinguish between "code-driven" and "correlational" approaches. Code-driven approaches make assumptions about the neural code for representing world states and the associated uncertainty. By contrast, correlational approaches search for relationships between uncertainty and neural activity without constraints on the neural representation of the world state that this uncertainty accompanies. To compare these two approaches, we apply several criteria for neural representations: sensitivity, specificity, invariance, functionality. Our analysis reveals that the two approaches lead to different, but complementary findings, shaping new research questions and guiding future experiments.

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is to understand how the brain represents uncertainty. Unlike most studied neural representational quantities, uncertainty is an attribute of an observer's belief about the world, rather than a direct state of the world. This poses specific challenges to research methods. The paper analyzes how research on neural uncertainty representation in the existing literature addresses these challenges and distinguishes between two research methods: "code - driven" and "correlational". 1. **Defining Uncertainty**: - Uncertainty describes an observer's representation of the world state \( s \). Due to the ambiguity and randomness of the input information \( I \), there is uncertainty in the observer's estimate of \( s \). - Mathematically, uncertainty can be described by the posterior probability distribution \( p(s | I_{\text{observed}}) \), where \( I_{\text{observed}} \) is the observed input. If the uncertainty is zero, then \( s \) can be completely determined by a specific \( I_{\text{observed}} \); otherwise, there are multiple possible values of \( s \). 2. **Research Methods**: - **Correlational Approach**: This method does not assume a specific form of the neural code for \( s \), but instead looks for parts related to brain activity \( r \) through proxy variables \( u(I) \) or \( u(b) \) of the input or behavior. For example, in fMRI studies, researchers may test the relationship between the activity of each voxel and \( u(I) \). - **Code - Driven Approach**: This method assumes a specific form of the neural code for \( s \) and its associated uncertainty \( u \). Based on these assumptions, researchers can read \( u \) from the neural activity \( r \), and verify the validity of the model by comparing the uncertainty \( u(r) \) derived from the neural activity with the uncertainty \( u(I) \) or \( u(b) \) derived from the input or behavior. 3. **Evaluation Criteria**: - The paper proposes four general evaluation criteria to evaluate claims about neural representation of uncertainty: sensitivity, specificity, invariance, and functionality. - **Sensitivity**: Neural activity \( r \) is sensitive to feature \( x \) if the change in \( r \) is related to the change in \( x \). - **Specificity**: \( r \) specifically represents \( x \) rather than other features \( y \), if in the case of controlling \( y \), the change in \( r \) is still related to the change in \( x \). - **Invariance**: The representation of \( x \) by \( r \) is not affected by the change in \( y \) in the case of controlling \( x \). - **Functionality**: The representation of \( x \) by \( r \) is functional if it can cause a behavioral response \( b \), such as reporting a perceived value \( x \) or making a decision based on the uncertainty of the information source. Through these methods and criteria, the paper aims to provide a framework for future research to better understand and evaluate how the brain represents uncertainty.

Studying the neural representations of uncertainty

Studying the neural representations of uncertainty

Representation of visual uncertainty through neural gain variability

Joint representation of working memory and uncertainty in human cortex

Neural Representation of Valenced and Generic Probability and Uncertainty

Brain networks under uncertainty: A coordinate-based meta-analysis of brain imaging studies

Spatiotemporal neural dynamics of object recognition under uncertainty in humans

Neural Representations of Sensory Uncertainty and Confidence Are Associated with Perceptual Curiosity

Approaching Neural Network Uncertainty Realism

Task-dependent Uncertainty Modulation of Perceptual Decisions in the Human Brain.

A Survey of Uncertainty in Deep Neural Networks

Symmetrical uncertainty based neural selection in motor cortex brain-machine interfaces

Uncertainty Quantification in Working Memory via Moment Neural Networks

On the Uncertainty Principle of Neural Networks

Temporally organized representations of reward and risk in the human brain

Automatic and Fast Encoding of Representational Uncertainty Underlies the Distortion of Relative Frequency

Modeling correlated noise is necessary to decode uncertainty

The Neural Correlates of Novelty and Variability in Human Decision-Making under an Active Inference Framework

A neural signature for the subjective experience of threat anticipation under uncertainty

Different types of uncertainty in multisensory perceptual decision making

Uncertainty deconstructed: conceptual analysis and state-of-the-art review of the ERP correlates of risk and ambiguity in decision-making