Search

How the brain creates compositional cognitive maps that support both flexible and efficient planning remains poorly understood. Here, authors propose a biologically-realistic computational model addressing this question, which reproduces response fields across cells in the medial entorhinal cortex.

Making new decisions requires retrieving memories, but when this occurs is unclear. The authors show that people typically access memories before a choice unless there are many related memories to consider, in which case we wait until choice time.

Previous work has shown that learning models based on running averages of received rewards can account for sequential choices in value based decisions. Here the authors show that such choices can also be influenced by a process of sampling memories for individual past outcomes, and that the sampled memories are episodic in nature.

Mattar and Daw propose a normative theory predicting which memories should be accessed at each moment to optimize future decisions. This theory offers a simple explanation for numerous findings about hippocampal replay, bridging planning and learning.

We often look back and forth between options before deciding which one to choose, even if we have seen them both before. A new study suggests that people are biased to choose things they look at more, providing new insight into how the subjective values of options are constructed.

Use of a gambling task and a functional magnetic resonance imaging (fMRI) scanner shows that human subjects' choices can be characterized by a computationally well regarded strategy for addressing the explore/exploit dilemma.

Adaptive learning is difficult in noisy environments, yet people often succeed. Here, the authors show that humans do this by distinguishing between two easily confused types of noise—volatility and stochasticity—which require opposite adjustments to learning.

The authors present a feature-specific prediction error model that explains heterogeneity in dopaminergic signals within and across projection-defined populations. Model-derived predictions of dopamine activity align with empirical recordings.

Head-fixation system is widely used in neuroscience research but has limitations in application due to restraint. Here the authors developed a magnetic voluntary head-fixation system that allows stable rat hippocampal imaging during complex behaviors.

The authors employ the computational learning approach that is widely used in the striatum to examine the contributions of the amygdala, and find that these two structures have complementary roles in aversive learning.

Models of decision making have so far been unable to account for how humans’ choices can be flexible yet efficient. Here the authors present a linear reinforcement learning model which explains both flexibility, and rare limitations such as habits, as arising from efficient approximate computation

In a dynamic reinforcement learning task, human players adaptively employ a mix of decision strategies, including model-based learning but also the successor representation.

Estimating confidence in the decision making ability of others is important for cooperative behaviour. Here the authors combine computational modelling and fMRI to investigate how the brain supports this process.

Human learning depends on opposing effects of two noise factors: volatility and stochasticity. Here the authors present a model of learning that shows how and why joint estimation of these factors is important for understanding healthy and pathological learning.

Midbrain dopamine neurons have been implicated in two related but distinct processes: reward learning and action generation. By combining an operant learning task in mice with recordings from projection-defined dopamine neurons, the authors found that dopamine neurons targeting different parts of the striatum carry different information about rewards and chosen actions.

Hunter et al. find evidence that people with higher self-reported social anxiety deliberate more in a socially framed reinforcement learning task.

In some types of decision-making, people must accept or forego an option without knowing what prospects might later be available. Here, the authors reveal how a key bias– asymmetric learning from negative versus positive outcomes – emerges in this type of decision.

Changing one’s mind requires revising previous decisions in light of new evidence. The authors combine a psychophysical manipulation of post-decision evidence with fMRI to isolate neural mediators of changes of mind in human prefrontal cortex.

Animals exhibit a variety of different responses to threatening stimuli, including both innate and instrumental behaviours. LeDoux and Daw describe a new taxonomy of defensive behaviours and review what we know about the underlying neural circuits and computational processes.

When an animal is performing a cognitive task, individual neurons in the prefrontal cortex show a mixture of responses that is often difficult to decipher and interpret; here new computational methods to decode and extract rich sets of information from these neural responses are revealed and demonstrate how this mixed selectivity offers a computational advantage over specialized cells.

Individuals must compensate for their motor uncertainty—that is, the discrepancy between intended movement and actual. Here, the authors measured the subjective error representation used in planning reaching movements and found that, while the objective motor error was uni-modal, near-Gaussian, subjective distributions were typically multimodal. This suggests a flexible strategy for computing with uncertainty across many different sorts of problems.

The authors propose that deciding where to look and reach depends on how neurons in the posterior parietal cortex communicate with each other. They find that ‘dual-coherent’ neurons, which tend to fire spikes timed to neural activity within and across the banks of the intraparietal sulcus, predict look-reach choices before neurons without this property.

Although it has been widely hypothesized that decisions can be guided by mental simulation of their likely consequences, there has not been direct evidence linking prospection to choices. Here, using fMRI, the authors show that neural representation of future outcomes is related to the choices that participants make.

This study uses fMRI in humans to find that prediction errors about pain are encoded in the periaqueductal gray. Modeling inter-area connectivity suggests that the ventromedial prefrontal cortex and the putamen pass on a value-related signal to this midbrain structure, which then conveys predictor error signals to prefrontal regions that regulate behavior.

A new paper reports that dopaminergic neurons initially responded optimistically in rats given free choice between two rewards, as though the animal had chosen the better reward, even on trials when it failed to do so. These findings suggest that current computational theories of dopaminergic function may need to be revised.

What you choose depends on what information your brain considers and what it neglects when computing the value of actions. An early theory used this insight for a computational account of habits versus deliberation. It has ultimately helped uncover how choice in the brain goes beyond such simple dichotomies.

Two-photon calcium imaging of a large population of dopamine neurons in the ventral tegmental area of mice performing a virtual-reality navigation task reveals the organization principles of the dopamine system.

Dopaminergic neurons are thought to inform decisions by reporting errors in reward prediction. A new study reports dopaminergic responses as monkeys make choices, supporting one computational theory of appetitive learning.

Fung et al. show that participants’ trait anxiety is associated with earlier escape decisions when facing slowly approaching threats. Anxiety correlates with task-driven blood-oxygen-level-dependent activity in the cognitive fear circuits.

A revolution is underway in cognitive neuroscience, where tools and techniques from computer science and the tech industry are helping to extract more meaningful cognitive signals from noisy and increasingly large fMRI datasets. In this paper, the authors review the cutting edge of such computational analyses and discuss future opportunities and challenges.