Exploration Disrupts Choice-Predictive Signals and Alters Dynamics in Prefrontal Cortex

Touchscreen response precision is sensitive to the explore/exploit tradeoff

The explore/exploit tradeoff is a fundamental property of choice selection during reward-guided decision making. In perceptual decision making, higher certainty decisions are more motorically precise, even when the decision does not require motor accuracy. However, while we can parametrically control uncertainty in perceptual tasks, we do not know what variables - if any - shape motor precision and reflect subjective certainty during reward-guided decision making. Touchscreens are increasingly used across species to measure choice, but provide no tactile feedback on whether an action is precise or not, and therefore provide a valuable opportunity to determine whether actions differ in precision due to explore/exploit state, reward, or individual variables. We find all three of these factors exert independent drives towards increased precision. During exploit states, successive touches to the same choice are closer together than those made in an explore state, consistent with exploit states reflecting higher certainty and/or motor stereotypy in responding. However, exploit decisions might be expected to be rewarded more frequently than explore decisions. We find that exploit choice precision is increased independently of a separate increase in precision due to immediate past reward, suggesting multiple mechanisms regulating choice precision. Finally, we see evidence that male mice in general are less precise in their interactions with the touchscreen than females, even when exploiting a choice. These results suggest that as exploit behavior emerges in reward-guided decision making, individuals become more motorically precise reflecting increased certainty, even when decision choice does not require additional motor accuracy, but this is influenced by individual differences and prior reward. These data uncover the hidden potential for touchscreen tasks in any species to uncover the latent neural states that unite cognition and movement.

Dynamic reinforcement learning reveals time-dependent shifts in strategy during reward learning

Different brain systems have been hypothesized to subserve multiple "experts" that compete to generate behavior. In reinforcement learning, two general processes, one model-free (MF) and one model-based (MB), are often modeled as a mixture of agents (MoA) and hypothesized to capture differences between automaticity vs. deliberation. However, shifts in strategy cannot be captured by a static MoA. To investigate such dynamics, we present the mixture-of-agents hidden Markov model (MoA-HMM), which simultaneously learns inferred action values from a set of agents and the temporal dynamics of underlying "hidden" states that capture shifts in agent contributions over time. Applying this model to a multi-step, reward-guided task in rats reveals a progression of within-session strategies: a shift from initial MB exploration to MB exploitation, and finally to reduced engagement. The inferred states predict changes in both response time and OFC neural encoding during the task, suggesting that these states are capturing real shifts in dynamics.

Distinct computational mechanisms of uncertainty processing explain opposing exploratory behaviors in anxiety and apathy

Decision-making in uncertain environments often leads to varied outcomes. Understanding how individuals interpret the causes of unexpected feedback is crucial for adaptive behavior and mental well-being. Uncertainty can be broadly categorized into two components: volatility and stochasticity. Volatility is about how quickly conditions change, impacting results. Stochasticity, on the other hand, refers to outcomes affected by random chance or "luck". Understanding these factors enables individuals to have more effective environmental analysis and strategy implementation (explore or exploit) for future decisions. This study investigates how anxiety and apathy, two prevalent affective states, influence the perceptions of uncertainty and exploratory behavior. Participants (N = 1001) completed a restless three-armed bandit task that was analyzed using latent state models. Anxious individuals perceived uncertainty as more volatile, leading to increased exploration and learning rates, especially after reward omission. Conversely, apathetic individuals viewed uncertainty as more stochastic, resulting in decreased exploration and learning rates. The perceived volatility-to-stochasticity ratio mediated the anxiety-exploration relationship post-adverse outcomes. Dimensionality reduction showed exploration and uncertainty estimation to be distinct but related latent factors shaping a manifold of adaptive behavior that is modulated by anxiety and apathy. These findings reveal distinct computational mechanisms for how anxiety and apathy influence decision-making, providing a framework for understanding cognitive and affective processes in neuropsychiatric disorders.

A Comparison of Rapid Rule-Learning Strategies in Humans and Monkeys

Interspecies comparisons are key to deriving an understanding of the behavioral and neural correlates of human cognition from animal models. We perform a detailed comparison of the strategies of female macaque monkeys to male and female humans on a variant of the Wisconsin Card Sorting Test (WCST), a widely studied and applied task that provides a multiattribute measure of cognitive function and depends on the frontal lobe. WCST performance requires the inference of a rule change given ambiguous feedback. We found that well-trained monkeys infer new rules three times more slowly than minimally instructed humans. Input-dependent hidden Markov model-generalized linear models were fit to their choices, revealing hidden states akin to feature-based attention in both species. Decision processes resembled a win-stay, lose-shift strategy with interspecies similarities as well as key differences. Monkeys and humans both test multiple rule hypotheses over a series of rule-search trials and perform inference-like computations to exclude candidate choice options. We quantitatively show that perseveration, random exploration, and poor sensitivity to negative feedback account for the slower task-switching performance in monkeys.

Humans forage for reward in reinforcement learning tasks

How do we make good decisions in uncertain environments? In psychology and neuroscience, the classic answer is that we calculate the value of each option and then compare the values to choose the most rewarding, modulo some exploratory noise. An ethologist, conversely, would argue that we commit to one option until its value drops below a threshold, at which point we start exploring other options. In order to determine which view better describes human decision-making, we developed a novel, foraging-inspired sequential decision-making model and used it to ask whether humans compare to threshold ("Forage") or compare alternatives ("Reinforcement-Learn" [RL]). We found that the foraging model was a better fit for participant behavior, better predicted the participants' tendency to repeat choices, and predicted the existence of held-out participants with a pattern of choice that was almost impossible under RL. Together, these results suggest that humans use foraging computations, rather than RL, even in classic reinforcement learning tasks.

Claustrum neurons projecting to the anterior cingulate restrict engagement during sleep and behavior

The claustrum has been linked to attention and sleep. We hypothesized that this reflects a shared function, determining responsiveness to stimuli, which spans the axis of engagement. To test this hypothesis, we recorded claustrum population dynamics from male mice during both sleep and an attentional task ('ENGAGE'). Heightened activity in claustrum neurons projecting to the anterior cingulate cortex (ACCp) corresponded to reduced sensory responsiveness during sleep. Similarly, in the ENGAGE task, heightened ACCp activity correlated with disengagement and behavioral lapses, while low ACCp activity correlated with hyper-engagement and impulsive errors. Chemogenetic elevation of ACCp activity reduced both awakenings during sleep and impulsive errors in the ENGAGE task. Furthermore, mice employing an exploration strategy in the task showed a stronger correlation between ACCp activity and performance compared to mice employing an exploitation strategy which reduced task complexity. Our results implicate ACCp claustrum neurons in restricting engagement during sleep and goal-directed behavior.

Impaired value-based decision-making in Parkinson's disease apathy

Apathy is a common and disabling complication of Parkinson's disease characterized by reduced goal-directed behaviour. Several studies have reported dysfunction within prefrontal cortical regions and projections from brainstem nuclei whose neuromodulators include dopamine, serotonin and noradrenaline. Work in animal and human neuroscience have confirmed contributions of these neuromodulators on aspects of motivated decision-making. Specifically, these neuromodulators have overlapping contributions to encoding the value of decisions, and influence whether to explore alternative courses of action or persist in an existing strategy to achieve a rewarding goal. Building upon this work, we hypothesized that apathy in Parkinson's disease should be associated with an impairment in value-based learning. Using a four-armed restless bandit reinforcement learning task, we studied decision-making in 75 volunteers; 53 patients with Parkinson's disease, with and without clinical apathy, and 22 age-matched healthy control subjects. Patients with apathy exhibited impaired ability to choose the highest value bandit. Task performance predicted an individual patient's apathy severity measured using the Lille Apathy Rating Scale (R = -0.46, P < 0.001). Computational modelling of the patient's choices confirmed the apathy group made decisions that were indifferent to the learnt value of the options, consistent with previous reports of reward insensitivity. Further analysis demonstrated a shift away from exploiting the highest value option and a reduction in perseveration, which also correlated with apathy scores (R = -0.5, P < 0.001). We went on to acquire functional MRI in 59 volunteers; a group of 19 patients with and 20 without apathy and 20 age-matched controls performing the Restless Bandit Task. Analysis of the functional MRI signal at the point of reward feedback confirmed diminished signal within ventromedial prefrontal cortex in Parkinson's disease, which was more marked in apathy, but not predictive of their individual apathy severity. Using a model-based categorization of choice type, decisions to explore lower value bandits in the apathy group activated prefrontal cortex to a similar degree to the age-matched controls. In contrast, Parkinson's patients without apathy demonstrated significantly increased activation across a distributed thalamo-cortical network. Enhanced activity in the thalamus predicted individual apathy severity across both patient groups and exhibited functional connectivity with dorsal anterior cingulate cortex and anterior insula. Given that task performance in patients without apathy was no different to the age-matched control subjects, we interpret the recruitment of this network as a possible compensatory mechanism, which compensates against symptomatic manifestation of apathy in Parkinson's disease.

Learning attentional templates for value-based decision-making

Attention filters sensory inputs to enhance task-relevant information. It is guided by an "attentional template" that represents the stimulus features that are currently relevant. To understand how the brain learns and uses templates, we trained monkeys to perform a visual search task that required them to repeatedly learn new attentional templates. Neural recordings found that templates were represented across the prefrontal and parietal cortex in a structured manner, such that perceptually neighboring templates had similar neural representations. When the task changed, a new attentional template was learned by incrementally shifting the template toward rewarded features. Finally, we found that attentional templates transformed stimulus features into a common value representation that allowed the same decision-making mechanisms to deploy attention, regardless of the identity of the template. Altogether, our results provide insight into the neural mechanisms by which the brain learns to control attention and how attention can be flexibly deployed across tasks.

Using Natural Scenes to Enhance our Understanding of the Cerebral Cortex's Role in Visual Search

Using natural scenes is an approach to studying the visual and eye movement systems approximating how these systems function in everyday life. This review examines the results from behavioral and neurophysiological studies using natural scene viewing in humans and monkeys. The use of natural scenes for the study of cerebral cortical activity is relatively new and presents challenges for data analysis. Methods and results from the use of natural scenes for the study of the visual and eye movement cortex are presented, with emphasis on new insights that this method provides enhancing what is known about these cortical regions from the use of conventional methods.

Reinforcement learning deficits exhibited by postnatal PCP-treated rats enable deep neural network classification

The ability to appropriately update the value of a given action is a critical component of flexible decision making. Several psychiatric disorders, including schizophrenia, are associated with impairments in flexible decision making that can be evaluated using the probabilistic reversal learning (PRL) task. The PRL task has been reverse-translated for use in rodents. Disrupting glutamate neurotransmission during early postnatal neurodevelopment in rodents has induced behavioral, cognitive, and neuropathophysiological abnormalities relevant to schizophrenia. Here, we tested the hypothesis that using the NMDA receptor antagonist phencyclidine (PCP) to disrupt postnatal glutamatergic transmission in rats would lead to impaired decision making in the PRL. Consistent with this hypothesis, compared to controls the postnatal PCP-treated rats completed fewer reversals and exhibited disruptions in reward and punishment sensitivity (i.e., win-stay and lose-shift responding, respectively). Moreover, computational analysis of behavior revealed that postnatal PCP-treatment resulted in a pronounced impairment in the learning rate throughout PRL testing. Finally, a deep neural network (DNN) trained on the rodent behavior could accurately predict the treatment group of subjects. These data demonstrate that disrupting early postnatal glutamatergic neurotransmission impairs flexible decision making and provides evidence that DNNs can be trained on behavioral datasets to accurately predict the treatment group of new subjects, highlighting the potential for DNNs to aid in the diagnosis of schizophrenia.

A low dimensional manifold of human exploratory behavior reveals opposing roles for apathy and anxiety

Exploration-exploitation decision-making is a feature of daily life that is altered in a number of neuropsychiatric conditions. Humans display a range of exploration and exploitation behaviors, which can be affected by apathy and anxiety. It remains unknown how factors underlying decision-making generate the spectrum of observed exploration-exploitation behavior and how they relate to states of anxiety and apathy. Here, we report a latent structure underlying sequential exploration and exploitation decisions that explains variation in anxiety and apathy. 1001 participants in a gender-balanced sample completed a three-armed restless bandit task along with psychiatric symptom surveys. Using dimensionality reduction methods, we found that decision sequences reduced to a low-dimensional manifold. The axes of this manifold explained individual differences in the balance between states of exploration and exploitation and the stability of those states, as determined by a statistical mechanics model of decision-making. Position along the balance axis was correlated with opposing symptoms of behavioral apathy and anxiety, while position along the stability axis correlated with the level of emotional apathy. This result resolves a paradox over how these symptoms can be correlated in samples but have opposite effects on behavior. Furthermore, this work provides a basis for using behavioral manifolds to reveal relationships between behavioral dynamics and affective states, with important implications for behavioral measurement approaches to neuropsychiatric conditions.

Pupil size predicts the onset of exploration in brain and behavior

In uncertain environments, intelligent decision-makers exploit actions that have been rewarding in the past, but also explore actions that could be even better. Several neuromodulatory systems are implicated in exploration, based, in part, on work linking exploration to pupil size-a peripheral correlate of neuromodulatory tone and index of arousal. However, pupil size could instead track variables that make exploration more likely, like volatility or reward, without directly predicting either exploration or its neural bases. Here, we simultaneously measured pupil size, exploration, and neural population activity in the prefrontal cortex while two rhesus macaques explored and exploited in a dynamic environment. We found that pupil size under constant luminance specifically predicted the onset of exploration, beyond what could be explained by reward history. Pupil size also predicted disorganized patterns of prefrontal neural activity at both the single neuron and population levels, even within periods of exploitation. Ultimately, our results support a model in which pupil-linked mechanisms promote the onset of exploration via driving the prefrontal cortex through a critical tipping point where prefrontal control dynamics become disorganized and exploratory decisions are possible.

Neural responses in macaque prefrontal cortex are linked to strategic exploration

Humans have been shown to strategically explore. They can identify situations in which gathering information about distant and uncertain options is beneficial for the future. Because primates rely on scarce resources when they forage, they are also thought to strategically explore, but whether they use the same strategies as humans and the neural bases of strategic exploration in monkeys are largely unknown. We designed a sequential choice task to investigate whether monkeys mobilize strategic exploration based on whether information can improve subsequent choice, but also to ask the novel question about whether monkeys adjust their exploratory choices based on the contingency between choice and information, by sometimes providing the counterfactual feedback about the unchosen option. We show that monkeys decreased their reliance on expected value when exploration could be beneficial, but this was not mediated by changes in the effect of uncertainty on choices. We found strategic exploratory signals in anterior and mid-cingulate cortex (ACC/MCC) and dorsolateral prefrontal cortex (dlPFC). This network was most active when a low value option was chosen, which suggests a role in counteracting expected value signals, when exploration away from value should to be considered. Such strategic exploration was abolished when the counterfactual feedback was available. Learning from counterfactual outcome was associated with the recruitment of a different circuit centered on the medial orbitofrontal cortex (OFC), where we showed that monkeys represent chosen and unchosen reward prediction errors. Overall, our study shows how ACC/MCC-dlPFC and OFC circuits together could support exploitation of available information to the fullest and drive behavior towards finding more information through exploration when it is beneficial.

Humans adaptively resolve the explore-exploit dilemma under cognitive constraints: Evidence from a multi-armed bandit task

When navigating uncertain worlds, humans must balance exploring new options versus exploiting known rewards. Longer horizons and spatially structured option values encourage humans to explore, but the impact of real-world cognitive constraints such as environment size and memory demands on explore-exploit decisions is unclear. In the present study, humans chose between options varying in uncertainty during a multi-armed bandit task with varying environment size and memory demands. Regression and cognitive computational models of choice behavior showed that with a lower cognitive load, humans are more exploratory than a simulated value-maximizing learner, but under cognitive constraints, they adaptively scale down exploration to maintain exploitation. Thus, while humans are curious, cognitive constraints force people to decrease their strategic exploration in a resource-rational-like manner to focus on harvesting known rewards.

Tonic activity in lateral habenula neurons acts as a neutral valence brake on reward-seeking behavior

Survival requires both the ability to persistently pursue goals and the ability to determine when it is time to stop, an adaptive balance of perseverance and disengagement. Neural activity in the lateral habenula (LHb) has been linked to negative valence, but its role in regulating the balance between engaged reward seeking and disengaged behavioral states remains unclear. Here, we show that LHb neural activity is tonically elevated during minutes-long periods of disengagement from reward-seeking behavior, both when due to repeated reward omission (negative valence) and when sufficient reward has been consumed (positive valence). Furthermore, we show that LHb inhibition extends ongoing reward-seeking behavioral states but does not prompt task re-engagement. We find no evidence for similar tonic activity changes in ventral tegmental area dopamine neurons. Our findings support a framework in which tonic activity in LHb neurons suppresses engagement in reward-seeking behavior in response to both negatively and positively valenced factors.

Uncertainty modulates visual maps during noninstrumental information demand

Animals are intrinsically motivated to obtain information independently of instrumental incentives. This motivation depends on two factors: a desire to resolve uncertainty by gathering accurate information and a desire to obtain positively-valenced observations, which predict favorable rather than unfavorable outcomes. To understand the neural mechanisms, we recorded parietal cortical activity implicated in prioritizing stimuli for spatial attention and gaze, in a task in which monkeys were free (but not trained) to obtain information about probabilistic non-contingent rewards. We show that valence and uncertainty independently modulated parietal neuronal activity, and uncertainty but not reward-related enhancement consistently correlated with behavioral sensitivity. The findings suggest uncertainty-driven and valence-driven information demand depend on partially distinct pathways, with the former being consistently related to parietal responses and the latter depending on additional mechanisms implemented in downstream structures.

Sequential delay and probability discounting tasks in mice reveal anchoring effects partially attributable to decision noise

Delay discounting and probability discounting decision making tasks in rodent models have high translational potential. However, it is unclear whether the discounted value of the large reward option is the main contributor to variability in animals' choices in either task, which may limit translation to humans. Male and female mice underwent sessions of delay and probability discounting in sequence to assess how choice behavior adapts over experience with each task. To control for "anchoring" (persistent choices based on the initial delay or probability), mice experienced "Worsening" schedules where the large reward was offered under initially favorable conditions that became less favorable during testing, followed by "Improving" schedules where the large reward was offered under initially unfavorable conditions that improved over a session. During delay discounting, both male and female mice showed elimination of anchoring effects over training. In probability discounting, both sexes of mice continued to show some anchoring even after months of training. One possibility is that "noisy", exploratory choices could contribute to these persistent anchoring effects, rather than constant fluctuations in value discounting. We fit choice behavior in individual animals using models that included both a value-based discounting parameter and a decision noise parameter that captured variability in choices deviating from value maximization. Changes in anchoring behavior over time were tracked by changes in both the value and decision noise parameters in delay discounting, but by the decision noise parameter in probability discounting. Exploratory decision making was also reflected in choice response times that tracked the degree of conflict caused by both uncertainty and temporal cost, but was not linked with differences in locomotor activity reflecting chamber exploration. Thus, variable discounting behavior in mice can result from changes in exploration of the decision options rather than changes in reward valuation.

Long-range cortical synchronization supports abrupt visual learning

Visual plasticity declines sharply after the critical period, yet we easily learn to recognize new faces and places, even as adults. Such learning is often characterized by a "moment of insight," an abrupt and dramatic improvement in recognition. The mechanisms that support abrupt learning are unknown, but one hypothesis is that they involve changes in synchronization between brain regions. To test this hypothesis, we used a behavioral task in which non-human primates rapidly learned to recognize novel images and to associate them with specific responses. Simultaneous recordings from inferotemporal and prefrontal cortices revealed a transient synchronization of neural activity between these areas that peaked around the moment of insight. Synchronization was strongest between inferotemporal sites that encoded images and reward-sensitive prefrontal sites. Moreover, its magnitude intensified gradually over image exposures, suggesting that abrupt learning is the culmination of a search for informative signals within a circuit linking sensory information to task demands.

Mice exhibit stochastic and efficient action switching during probabilistic decision making

In probabilistic and nonstationary environments, individuals must use internal and external cues to flexibly make decisions that lead to desirable outcomes. To gain insight into the process by which animals choose between actions, we trained mice in a task with time-varying reward probabilities. In our implementation of such a two-armed bandit task, thirsty mice use information about recent action and action–outcome histories to choose between two ports that deliver water probabilistically. Here we comprehensively modeled choice behavior in this task, including the trial-to-trial changes in port selection, i.e., action switching behavior. We find that mouse behavior is, at times, deterministic and, at others, apparently stochastic. The behavior deviates from that of a theoretically optimal agent performing Bayesian inference in a hidden Markov model (HMM). We formulate a set of models based on logistic regression, reinforcement learning, and sticky Bayesian inference that we demonstrate are mathematically equivalent and that accurately describe mouse behavior. The switching behavior of mice in the task is captured in each model by a stochastic action policy, a history-dependent representation of action value, and a tendency to repeat actions despite incoming evidence. The models parsimoniously capture behavior across different environmental conditionals by varying the stickiness parameter, and like the mice, they achieve nearly maximal reward rates. These results indicate that mouse behavior reaches near-maximal performance with reduced action switching and can be described by a set of equivalent models with a small number of relatively fixed parameters.

Lower Levels of Directed Exploration and Reflective Thinking Are Associated With Greater Anxiety and Depression

Anxiety and depression are often associated with strong beliefs that entering specific situations will lead to aversive outcomes - even when these situations are objectively safe and avoiding them reduces well-being. A possible mechanism underlying this maladaptive avoidance behavior is a failure to reflect on: (1) appropriate levels of uncertainty about the situation, and (2) how this uncertainty could be reduced by seeking further information (i.e., exploration). To test this hypothesis, we asked a community sample of 416 individuals to complete measures of reflective cognition, exploration, and symptoms of anxiety and depression. Consistent with our hypotheses, we found significant associations between each of these measures in expected directions (i.e., positive relationships between reflective cognition and strategic information-seeking behavior or "directed exploration", and negative relationships between these measures and anxiety/depression symptoms). Further analyses suggested that the relationship between directed exploration and depression/anxiety was due in part to an ambiguity aversion promoting exploration in conditions where information-seeking was not beneficial (as opposed to only being due to under-exploration when more information would aid future choices). In contrast, reflectiveness was associated with greater exploration in appropriate settings and separately accounted for differences in reaction times, decision noise, and choice accuracy in expected directions. These results shed light on the mechanisms underlying information-seeking behavior and how they may contribute to symptoms of emotional disorders. They also highlight the potential clinical relevance of individual differences in reflectiveness and exploration and should motivate future research on their possible contributions to vulnerability and/or maintenance of affective disorders.

Reinforcement-learning in fronto-striatal circuits

We review the current state of knowledge on the computational and neural mechanisms of reinforcement-learning with a particular focus on fronto-striatal circuits. We divide the literature in this area into five broad research themes: the target of the learning-whether it be learning about the value of stimuli or about the value of actions; the nature and complexity of the algorithm used to drive the learning and inference process; how learned values get converted into choices and associated actions; the nature of state representations, and of other cognitive machinery that support the implementation of various reinforcement-learning operations. An emerging fifth area focuses on how the brain allocates or arbitrates control over different reinforcement-learning sub-systems or "experts". We will outline what is known about the role of the prefrontal cortex and striatum in implementing each of these functions. We then conclude by arguing that it will be necessary to build bridges from algorithmic level descriptions of computational reinforcement-learning to implementational level models to better understand how reinforcement-learning emerges from multiple distributed neural networks in the brain.

Interactions between ventrolateral prefrontal and anterior cingulate cortex during learning and behavioural change

Hypotheses and beliefs guide credit assignment - the process of determining which previous events or actions caused an outcome. Adaptive hypothesis formation and testing are crucial in uncertain and changing environments in which associations and meanings are volatile. Despite primates' abilities to form and test hypotheses, establishing what is causally responsible for the occurrence of particular outcomes remains a fundamental challenge for credit assignment and learning. Hypotheses about what surprises are due to stochasticity inherent in an environment as opposed to real, systematic changes are necessary for identifying the environment's predictive features, but are often hard to test. We review evidence that two highly interconnected frontal cortical regions, anterior cingulate cortex and ventrolateral prefrontal area 47/12o, provide a biological substrate for linking two crucial components of hypothesis-formation and testing: the control of information seeking and credit assignment. Neuroimaging, targeted disruptions, and neurophysiological studies link an anterior cingulate - 47/12o circuit to generation of exploratory behaviour, non-instrumental information seeking, and interpretation of subsequent feedback in the service of credit assignment. Our observations support the idea that information seeking and credit assignment are linked at the level of neural circuits and explain why this circuit is important for ensuring behaviour is flexible and adaptive.

The population doctrine in cognitive neuroscience

A major shift is happening within neurophysiology: a population doctrine is drawing level with the single-neuron doctrine that has long dominated the field. Population-level ideas have so far had their greatest impact in motor neuroscience, but they hold great promise for resolving open questions in cognition as well. Here, we codify the population doctrine and survey recent work that leverages this view to specifically probe cognition. Our discussion is organized around five core concepts that provide a foundation for population-level thinking: (1) state spaces, (2) manifolds, (3) coding dimensions, (4) subspaces, and (5) dynamics. The work we review illustrates the progress and promise that population-level thinking holds for cognitive neuroscience-for delivering new insight into attention, working memory, decision-making, executive function, learning, and reward processing.

Adaptive learning is structure learning in time

People use information flexibly. They often combine multiple sources of relevant information over time in order to inform decisions with little or no interference from intervening irrelevant sources. They adjust the degree to which they use new information over time rationally in accordance with environmental statistics and their own uncertainty. They can even use information gained in one situation to solve a problem in a very different one. Learning flexibly rests on the ability to infer the context at a given time, and therefore knowing which pieces of information to combine and which to separate. We review the psychological and neural mechanisms behind adaptive learning and structure learning to outline how people pool together relevant information, demarcate contexts, prevent interference between information collected in different contexts, and transfer information from one context to another. By examining all of these processes through the lens of optimal inference we bridge concepts from multiple fields to provide a unified multi-system view of how the brain exploits structure in time to optimize learning.

The modulation of brain network integration and arousal during exploration

There is growing interest in how neuromodulators shape brain networks. Recent neuroimaging studies provide evidence that brainstem arousal systems, such as the locus coeruleus-norepinephrine system (LC-NE), influence functional connectivity and brain network topology, suggesting they have a role in flexibly reconfiguring brain networks in order to adapt behavior and cognition to environmental demands. To date, however, the relationship between brainstem arousal systems and functional connectivity has not been assessed within the context of a task with an established relationship between arousal and behavior, with most prior studies relying on incidental variations in arousal or pharmacological manipulation and static brain networks constructed over long periods of time. These factors have likely contributed to a heterogeneity of effects across studies. To address these issues, we took advantage of the association between LC-NE-linked arousal and exploration to probe the relationships between exploratory choice, arousal—as measured indirectly via pupil diameter—and brain network dynamics. Exploration in a bandit task was associated with a shift toward fewer, more weakly connected modules that were more segregated in terms of connectivity and topology but more integrated with respect to the diversity of cognitive systems represented in each module. Functional connectivity strength decreased, and changes in connectivity were correlated with changes in pupil diameter, in line with the hypothesis that brainstem arousal systems influence the dynamic reorganization of brain networks. More broadly, we argue that carefully aligning dynamic network analyses with task designs can increase the temporal resolution at which behaviorally- and cognitively-relevant modulations can be identified, and offer these results as a proof of concept of this approach.

Dichotomous dopaminergic and noradrenergic neural states mediate distinct aspects of exploitative behavioral states

The balance between exploiting known actions and exploring alternatives is critical for survival and hypothesized to rely on shifts in neuromodulation. We developed a behavioral paradigm to capture exploitative and exploratory states and imaged calcium dynamics in genetically identified dopaminergic and noradrenergic neurons. During exploitative states, characterized by motivated repetition of the same action choice, dopamine neurons in SNc encoding movement vigor showed sustained elevation of basal activity that lasted many seconds. This sustained activity emerged from longer positive responses, which accumulated during exploitative action-reward bouts, and hysteretic dynamics. Conversely, noradrenergic neurons in LC showed sustained inhibition of basal activity due to the accumulation of longer negative responses in LC. Chemogenetic manipulation of these sustained dynamics revealed that dopaminergic activity mediates action drive, whereas noradrenergic activity modulates choice diversity. These data uncover the emergence of sustained neural states in dopaminergic and noradrenergic networks that mediate dissociable aspects of exploitative bouts.

The anterior cingulate cortex directs exploration of alternative strategies

The ability to adjust one's behavioral strategy in complex environments is at the core of cognition. Doing so efficiently requires monitoring the reliability of the ongoing strategy and, when appropriate, switching away from it to evaluate alternatives. Studies in humans and non-human primates have uncovered signals in the anterior cingulate cortex (ACC) that reflect the pressure to switch away from the ongoing strategy, whereas other ACC signals relate to the pursuit of alternatives. However, whether these signals underlie computations that actually underpin strategy switching or merely reflect tracking of related variables remains unclear. Here we provide causal evidence that the rodent ACC actively arbitrates between persisting with the ongoing behavioral strategy and temporarily switching away to re-evaluate alternatives. Furthermore, by individually perturbing distinct output pathways, we establish that the two associated computations-determining whether to switch strategy and committing to the pursuit of a specific alternative-are segregated in the ACC microcircuitry.

Balancing exploration and exploitation with information and randomization

Explore-exploit decisions require us to trade off the benefits of exploring unknown options to learn more about them, with exploiting known options, for immediate reward. Such decisions are ubiquitous in nature, but from a computational perspective, they are notoriously hard. There is therefore much interest in how humans and animals make these decisions and recently there has been an explosion of research in this area. Here we provide a biased and incomplete snapshot of this field focusing on the major finding that many organisms use two distinct strategies to solve the explore-exploit dilemma: a bias for information ('directed exploration') and the randomization of choice ('random exploration'). We review evidence for the existence of these strategies, their computational properties, their neural implementations, as well as how directed and random exploration vary over the lifespan. We conclude by highlighting open questions in this field that are ripe to both explore and exploit.

The dynamics of explore-exploit decisions reveal a signal-to-noise mechanism for random exploration

Growing evidence suggests that behavioral variability plays a critical role in how humans manage the tradeoff between exploration and exploitation. In these decisions a little variability can help us to overcome the desire to exploit known rewards by encouraging us to randomly explore something else. Here we investigate how such 'random exploration' could be controlled using a drift-diffusion model of the explore-exploit choice. In this model, variability is controlled by either the signal-to-noise ratio with which reward is encoded (the 'drift rate'), or the amount of information required before a decision is made (the 'threshold'). By fitting this model to behavior, we find that while, statistically, both drift and threshold change when people randomly explore, numerically, the change in drift rate has by far the largest effect. This suggests that random exploration is primarily driven by changes in the signal-to-noise ratio with which reward information is represented in the brain.

Rules warp feature encoding in decision-making circuits

We have the capacity to follow arbitrary stimulus-response rules, meaning simple policies that guide our behavior. Rule identity is broadly encoded across decision-making circuits, but there are less data on how rules shape the computations that lead to choices. One idea is that rules could simplify these computations. When we follow a rule, there is no need to encode or compute information that is irrelevant to the current rule, which could reduce the metabolic or energetic demands of decision-making. However, it is not clear if the brain can actually take advantage of this computational simplicity. To test this idea, we recorded from neurons in 3 regions linked to decision-making, the orbitofrontal cortex (OFC), ventral striatum (VS), and dorsal striatum (DS), while macaques performed a rule-based decision-making task. Rule-based decisions were identified via modeling rules as the latent causes of decisions. This left us with a set of physically identical choices that maximized reward and information, but could not be explained by simple stimulus-response rules. Contrasting rule-based choices with these residual choices revealed that following rules (1) decreased the energetic cost of decision-making; and (2) expanded rule-relevant coding dimensions and compressed rule-irrelevant ones. Together, these results suggest that we use rules, in part, because they reduce the costs of decision-making through a distributed representational warping in decision-making circuits.

Reward uncertainty asymmetrically affects information transmission within the monkey fronto-parietal network

A central hypothesis in research on executive function is that controlled information processing is costly and is allocated according to the behavioral benefits it brings. However, while computational theories predict that the benefits of new information depend on prior uncertainty, the cellular effects of uncertainty on the executive network are incompletely understood. Using simultaneous recordings in monkeys, we describe several mechanisms by which the fronto-parietal network reacts to uncertainty. We show that the variance of expected rewards, independently of the value of the rewards, was encoded in single neuron and population spiking activity and local field potential (LFP) oscillations, and, importantly, asymmetrically affected fronto-parietal information transmission (measured through the coherence between spikes and LFPs). Higher uncertainty selectively enhanced information transmission from the parietal to the frontal lobe and suppressed it in the opposite direction, consistent with Bayesian principles that prioritize sensory information according to a decision maker’s prior uncertainty.

Bahareh Taghizadeh and Nicholas Foley et al. show that individual neuronal responses, population spiking activity, and local field potential oscillations encode the variance of expected rewards independent of their value. They also demonstrate that reward uncertainty asymmetrically affects neuronal transmission within the monkey fronto-parietal network.

Effects of Amygdala Lesions on Object-Based Versus Action-Based Learning in Macaques

The neural systems that underlie reinforcement learning (RL) allow animals to adapt to changes in their environment. In the present study, we examined the hypothesis that the amygdala would have a preferential role in learning the values of visual objects. We compared a group of monkeys (Macaca mulatta) with amygdala lesions to a group of unoperated controls on a two-armed bandit reversal learning task. The task had two conditions. In the What condition, the animals had to learn to select a visual object, independent of its location. And in the Where condition, the animals had to learn to saccade to a location, independent of the object at the location. In both conditions choice-outcome mappings reversed in the middle of the block. We found that monkeys with amygdala lesions had learning deficits in both conditions. Monkeys with amygdala lesions did not have deficits in learning to reverse choice-outcome mappings. Rather, amygdala lesions caused the monkeys to become overly sensitive to negative feedback which impaired their ability to consistently select the more highly valued action or object. These results imply that the amygdala is generally necessary for RL.

Heuristics and optimal solutions to the breadth-depth dilemma

In multialternative risky choice, we are often faced with the opportunity to allocate our limited information-gathering capacity between several options before receiving feedback. In such cases, we face a natural trade-off between breadth-spreading our capacity across many options-and depth-gaining more information about a smaller number of options. Despite its broad relevance to daily life, including in many naturalistic foraging situations, the optimal strategy in the breadth-depth trade-off has not been delineated. Here, we formalize the breadth-depth dilemma through a finite-sample capacity model. We find that, if capacity is small (∼10 samples), it is optimal to draw one sample per alternative, favoring breadth. However, for larger capacities, a sharp transition is observed, and it becomes best to deeply sample a very small fraction of alternatives, which roughly decreases with the square root of capacity. Thus, ignoring most options, even when capacity is large enough to shallowly sample all of them, is a signature of optimal behavior. Our results also provide a rich casuistic for metareasoning in multialternative decisions with bounded capacity using close-to-optimal heuristics.

Dopaminergic modulation of the exploration/exploitation trade-off in human decision-making

Involvement of dopamine in regulating exploration during decision-making has long been hypothesized, but direct causal evidence in humans is still lacking. Here, we use a combination of computational modeling, pharmacological intervention and functional magnetic resonance imaging to address this issue. Thirty-one healthy male participants performed a restless four-armed bandit task in a within-subjects design under three drug conditions: 150 mg of the dopamine precursor L-dopa, 2 mg of the D2 receptor antagonist haloperidol, and placebo. Choices were best explained by an extension of an established Bayesian learning model accounting for perseveration, directed exploration and random exploration. Modeling revealed attenuated directed exploration under L-dopa, while neural signatures of exploration, exploitation and prediction error were unaffected. Instead, L-dopa attenuated neural representations of overall uncertainty in insula and dorsal anterior cingulate cortex. Our results highlight the computational role of these regions in exploration and suggest that dopamine modulates how this circuit tracks accumulating uncertainty during decision-making.

Prefrontal Cortex Predicts State Switches during Reversal Learning

Reinforcement learning allows organisms to predict future outcomes and to update their beliefs about value in the world. The dorsal-lateral prefrontal cortex (dlPFC) integrates information carried by reward circuits, which can be used to infer the current state of the world under uncertainty. Here, we explored the dlPFC computations related to updating current beliefs during stochastic reversal learning. We recorded the activity of populations up to 1,000 neurons, simultaneously, in two male macaques while they executed a two-armed bandit reversal learning task. Behavioral analyses using a Bayesian framework showed that animals inferred reversals and switched their choice preference rapidly, rather than slowly updating choice values, consistent with state inference. Furthermore, dlPFC neural populations accurately encoded choice preference switches. These results suggest that prefrontal neurons dynamically encode decisions associated with Bayesian subjective values, highlighting the role of the PFC in representing a belief about the current state of the world.

The Transition from Evaluation to Selection Involves Neural Subspace Reorganization in Core Reward Regions

Economic choice proceeds from evaluation, in which we contemplate options, to selection, in which we weigh options and choose one. These stages must be differentiated so that decision makers do not proceed to selection before evaluation is complete. We examined responses of neurons in two core reward regions, orbitofrontal (OFC) and ventromedial prefrontal cortex (vmPFC), during two-option choice with asynchronous offer presentation. Our data suggest that neurons selective during the first (presumed evaluation) and second (presumed comparison and selection) offer epochs come from a single pool. Stage transition is accompanied by a shift toward orthogonality in the low-dimensional population response manifold. Nonetheless, the relative position of each option in driving responses in the population subspace is preserved. The orthogonalization we observe supports the hypothesis that the transition from evaluation to selection leads to reorganization of response subspace and suggests a mechanism by which value-related signals are prevented from prematurely driving choice.

Prefrontal attentional saccades explore space rhythmically

Recent studies suggest that attention samples space rhythmically through oscillatory interactions in the frontoparietal network. How these attentional fluctuations coincide with spatial exploration/displacement and exploitation/selection by a dynamic attentional spotlight under top-down control is unclear. Here, we show a direct contribution of prefrontal attention selection mechanisms to a continuous space exploration. Specifically, we provide a direct high spatio-temporal resolution prefrontal population decoding of the covert attentional spotlight. We show that it continuously explores space at a 7-12 Hz rhythm. Sensory encoding and behavioral reports are increased at a specific optimal phase w/ to this rhythm. We propose that this prefrontal neuronal rhythm reflects an alpha-clocked sampling of the visual environment in the absence of eye movements. These attentional explorations are highly flexible, how they spatially unfold depending both on within-trial and across-task contingencies. These results are discussed in the context of exploration-exploitation strategies and prefrontal top-down attentional control.

The formation and use of hierarchical cognitive maps in the brain: A neural network model

Many researchers have tried to model how environmental knowledge is learned by the brain and used in the form of cognitive maps. However, previous work was limited in various important ways: there was little consensus on how these cognitive maps were formed and represented, the planning mechanism was inherently limited to performing relatively simple tasks, and there was little consideration of how these mechanisms would scale up. This paper makes several significant advances. Firstly, the planning mechanism used by the majority of previous work propagates a decaying signal through the network to create a gradient that points towards the goal. However, this decaying signal limited the scale and complexity of tasks that can be solved in this manner. Here we propose several ways in which a network can can self-organize a novel planning mechanism that does not require decaying activity. We also extend this model with a hierarchical planning mechanism: a layer of cells that identify frequently-used sequences of actions and reuse them to significantly increase the efficiency of planning. We speculate that our results may explain the apparent ability of humans and animals to perform model-based planning on both small and large scales without a noticeable loss of efficiency.

A Basal Forebrain-Cingulate Circuit in Macaques Decides It Is Time to Act

The medial frontal cortex has been linked to voluntary action, but an explanation of why decisions to act emerge at particular points in time has been lacking. We show that, in macaques, decisions about whether and when to act are predicted by a set of features defining the animal’s current and past context; for example, respectively, cues indicating the current average rate of reward and recent previous voluntary action decisions. We show that activity in two brain areas—the anterior cingulate cortex and basal forebrain—tracks these contextual factors and mediates their effects on behavior in distinct ways. We use focused transcranial ultrasound to selectively and effectively stimulate deep in the brain, even as deep as the basal forebrain, and demonstrate that alteration of activity in the two areas changes decisions about when to act.

•

Likelihood and timing of voluntary action in macaques can be partially predicted

•

Recent experience and present context influence when voluntary action occurs

•

A basal forebrain-cingulate circuit mediated effects of these factors on behavior

•

Stimulation of this circuit by ultrasound changed decisions about when to act

Tonic exploration governs both flexibility and lapses

In many cognitive tasks, lapses (spontaneous errors) are tacitly dismissed as the result of nuisance processes like sensorimotor noise, fatigue, or disengagement. However, some lapses could also be caused by exploratory noise: randomness in behavior that facilitates learning in changing environments. If so, then strategic processes would need only up-regulate (rather than generate) exploration to adapt to a changing environment. This view predicts that more frequent lapses should be associated with greater flexibility because these behaviors share a common cause. Here, we report that when rhesus macaques performed a set-shifting task, lapse rates were negatively correlated with perseverative error frequency across sessions, consistent with a common basis in exploration. The results could not be explained by local failures to learn. Furthermore, chronic exposure to cocaine, which is known to impair cognitive flexibility, did increase perseverative errors, but, surprisingly, also improved overall set-shifting task performance by reducing lapse rates. We reconcile these results with a state-switching model in which cocaine decreases exploration by deepening attractor basins corresponding to rule states. These results support the idea that exploratory noise contributes to lapses, affecting rule-based decision-making even when it has no strategic value, and suggest that one key mechanism for regulating exploration may be the depth of rule states.

Subcortical Substrates of Explore-Exploit Decisions in Primates

The explore-exploit dilemma refers to the challenge of deciding when to forego immediate rewards and explore new opportunities that could lead to greater rewards in the future. While motivational neural circuits facilitate learning based on past choices and outcomes, it is unclear whether they also support computations relevant for deciding when to explore. We recorded neural activity in the amygdala and ventral striatum of rhesus macaques as they solved a task that required them to balance novelty-driven exploration with exploitation of what they had already learned. Using a partially observable Markov decision process (POMDP) model to quantify explore-exploit trade-offs, we identified that the ventral striatum and amygdala differ in how they represent the immediate value of exploitative choices and the future value of exploratory choices. These findings show that subcortical motivational circuits are important in guiding explore-exploit decisions.

Stable Representations of Decision Variables for Flexible Behavior

Decisions occur in dynamic environments. In the framework of reinforcement learning, the probability of performing an action is influenced by decision variables. Discrepancies between predicted and obtained rewards (reward prediction errors) update these variables, but they are otherwise stable between decisions. Although reward prediction errors have been mapped to midbrain dopamine neurons, it is unclear how the brain represents decision variables themselves. We trained mice on a dynamic foraging task in which they chose between alternatives that delivered reward with changing probabilities. Neurons in the medial prefrontal cortex, including projections to the dorsomedial striatum, maintained persistent firing rate changes over long timescales. These changes stably represented relative action values (to bias choices) and total action values (to bias response times) with slow decay. In contrast, decision variables were weakly represented in the anterolateral motor cortex, a region necessary for generating choices. Thus, we define a stable neural mechanism to drive flexible behavior.

Both a Gauge and a Filter: Cognitive Modulations of Pupil Size

Over 50 years of research have established that cognitive processes influence pupil size. This has led to the widespread use of pupil size as a peripheral measure of cortical processing in psychology and neuroscience. However, the function of cortical control over the pupil remains poorly understood. Why does visual attention change the pupil light reflex? Why do mental effort and surprise cause pupil dilation? Here, we consider these functional questions as we review and synthesize two literatures on cognitive effects on the pupil: how cognition affects pupil light response and how cognition affects pupil size under constant luminance. We propose that cognition may have co-opted control of the pupil in order to filter incoming visual information to optimize it for particular goals. This could complement other cortical mechanisms through which cognition shapes visual perception.