Dynamic Representation of the Subjective Value of Information.

To improve future decisions, people should seek information based on the value of information (VOI), which depends on the current evidence and the reward structure of the upcoming decision. When additional evidence is supplied, people should update the VOI to adjust subsequent information seeking, but the neurocognitive mechanisms of this updating process remain unknown. We used a modified beads task to examine how the VOI is represented and updated in the human brain of both sexes. We theoretically derived, and empirically verified, a normative prediction that the VOI depends on decision evidence and is biased by reward asymmetry. Using fMRI, we found that the subjective VOI is represented in the right dorsolateral prefrontal cortex (DLPFC). Critically, this VOI representation was updated when additional evidence was supplied, showing that the DLPFC dynamically tracks the up-to-date VOI over time. These results provide new insights into how humans adaptively seek information in the service of decision-making.SIGNIFICANCE STATEMENT For adaptive decision-making, people should seek information based on what they currently know and the extent to which additional information could improve the decision outcome, formalized as the VOI. Doing so requires dynamic updating of VOI according to outcome values and newly arriving evidence. We formalize these principles using a normative model and show that information seeking in people adheres to them. Using fMRI, we show that the underlying subjective VOI is represented in the dorsolateral prefrontal cortex and, critically, that it is updated in real time according to newly arriving evidence. Our results reveal the computational and neural dynamics through which evidence and values are combined to inform constantly evolving information-seeking decisions.

Embo: a Python package for empirical data analysis using the Information Bottleneck.

We present embo, a Python package to analyze empirical data using the Information Bottleneck (IB) method and its variants, such as the Deterministic Information Bottleneck (DIB). Given two random variables X and Y, the IB finds the stochastic mapping M of X that encodes the most information about Y, subject to a constraint on the information that M is allowed to retain about X. Despite the popularity of the IB, an accessible implementation of the reference algorithm oriented towards ease of use on empirical data was missing. Embo is optimized for the common case of discrete, low-dimensional data. Embo is fast, provides a standard data-processing pipeline, offers a parallel implementation of key computational steps, and includes reasonable defaults for the method parameters. Embo is broadly applicable to different problem domains, as it can be employed with any dataset consisting in joint observations of two discrete variables. It is available from the Python Package Index (PyPI), Zenodo and GitLab.

Reward and punishment reversal-learning in major depressive disorder.

Depression has been associated with impaired reward and punishment processing, but the specific nature of these deficits is still widely debated. We analyzed reinforcement-based decision making in individuals with major depressive disorder (MDD) to identify the specific decision mechanisms contributing to poorer performance. Individuals with MDD (n = 64) and matched healthy controls (n = 64) performed a probabilistic reversal-learning task in which they used feedback to identify which of two stimuli had the highest probability of reward (reward condition) or lowest probability of punishment (punishment condition). Learning differences were characterized using a hierarchical Bayesian reinforcement learning model. Depressed individuals made fewer optimal choices and adjusted more slowly to reversals in both the reward and punishment conditions. Computational modeling revealed that depressed individuals showed lower learning-rates and, to a lesser extent, lower value sensitivity in both the reward and punishment conditions. Learning-rates also predicted depression more accurately than simple performance metrics. These results demonstrate that depression is characterized by a hyposensitivity to positive outcomes, but not a hypersensitivity to negative outcomes. Additionally, we demonstrate that computational modeling provides a more precise characterization of the dynamics contributing to these learning deficits, offering stronger insights into the mechanistic processes affected by depression. (PsycInfo Database Record (c) 2020 APA, all rights reserved).