A characterization of the neural representation of confidence during probabilistic learning.

Learning in a stochastic and changing environment is a difficult task. Models of learning typically postulate that observations that deviate from the learned predictions are surprising and used to update those predictions. Bayesian accounts further posit the existence of a confidence-weighting mechanism: learning should be modulated by the confidence level that accompanies those predictions. However, the neural bases of this confidence are much less known than the ones of surprise. Here, we used a dynamic probability learning task and high-field MRI to identify putative cortical regions involved in the representation of confidence about predictions during human learning. We devised a stringent test based on the conjunction of four criteria. We localized several regions in parietal and frontal cortices whose activity is sensitive to the confidence of an ideal observer, specifically so with respect to potential confounds (surprise and predictability), and in a way that is invariant to which item is predicted. We also tested for functionality in two ways. First, we localized regions whose activity patterns at the subject level showed an effect of both confidence and surprise in qualitative agreement with the confidence-weighting principle. Second, we found neural representations of ideal confidence that also accounted for subjective confidence. Taken together, those results identify a set of cortical regions potentially implicated in the confidence-weighting of learning.

A theory of memory for binary sequences: Evidence for a mental compression algorithm in humans.

Working memory capacity can be improved by recoding the memorized information in a condensed form. Here, we tested the theory that human adults encode binary sequences of stimuli in memory using an abstract internal language and a recursive compression algorithm. The theory predicts that the psychological complexity of a given sequence should be proportional to the length of its shortest description in the proposed language, which can capture any nested pattern of repetitions and alternations using a limited number of instructions. Five experiments examine the capacity of the theory to predict human adults' memory for a variety of auditory and visual sequences. We probed memory using a sequence violation paradigm in which participants attempted to detect occasional violations in an otherwise fixed sequence. Both subjective complexity ratings and objective violation detection performance were well predicted by our theoretical measure of complexity, which simply reflects a weighted sum of the number of elementary instructions and digits in the shortest formula that captures the sequence in our language. While a simpler transition probability model, when tested as a single predictor in the statistical analyses, accounted for significant variance in the data, the goodness-of-fit with the data significantly improved when the language-based complexity measure was included in the statistical model, while the variance explained by the transition probability model largely decreased. Model comparison also showed that shortest description length in a recursive language provides a better fit than six alternative previously proposed models of sequence encoding. The data support the hypothesis that, beyond the extraction of statistical knowledge, human sequence coding relies on an internal compression using language-like nested structures.

Brain dynamics for confidence-weighted learning.

Learning in a changing, uncertain environment is a difficult problem. A popular solution is to predict future observations and then use surprising outcomes to update those predictions. However, humans also have a sense of confidence that characterizes the precision of their predictions. Bayesian models use a confidence-weighting principle to regulate learning: for a given surprise, the update is smaller when the confidence about the prediction was higher. Prior behavioral evidence indicates that human learning adheres to this confidence-weighting principle. Here, we explored the human brain dynamics sub-tending the confidence-weighting of learning using magneto-encephalography (MEG). During our volatile probability learning task, subjects' confidence reports conformed with Bayesian inference. MEG revealed several stimulus-evoked brain responses whose amplitude reflected surprise, and some of them were further shaped by confidence: surprise amplified the stimulus-evoked response whereas confidence dampened it. Confidence about predictions also modulated several aspects of the brain state: pupil-linked arousal and beta-range (15-30 Hz) oscillations. The brain state in turn modulated specific stimulus-evoked surprise responses following the confidence-weighting principle. Our results thus indicate that there exist, in the human brain, signals reflecting surprise that are dampened by confidence in a way that is appropriate for learning according to Bayesian inference. They also suggest a mechanism for confidence-weighted learning: confidence about predictions would modulate intrinsic properties of the brain state to amplify or dampen surprise responses evoked by discrepant observations.

Biases and Variability from Costly Bayesian Inference.

When humans infer underlying probabilities from stochastic observations, they exhibit biases and variability that cannot be explained on the basis of sound, Bayesian manipulations of probability. This is especially salient when beliefs are updated as a function of sequential observations. We introduce a theoretical framework in which biases and variability emerge from a trade-off between Bayesian inference and the cognitive cost of carrying out probabilistic computations. We consider two forms of the cost: a precision cost and an unpredictability cost; these penalize beliefs that are less entropic and less deterministic, respectively. We apply our framework to the case of a Bernoulli variable: the bias of a coin is inferred from a sequence of coin flips. Theoretical predictions are qualitatively different depending on the form of the cost. A precision cost induces overestimation of small probabilities, on average, and a limited memory of past observations, and, consequently, a fluctuating bias. An unpredictability cost induces underestimation of small probabilities and a fixed bias that remains appreciable even for nearly unbiased observations. The case of a fair (equiprobable) coin, however, is singular, with non-trivial and slow fluctuations in the inferred bias. The proposed framework of costly Bayesian inference illustrates the richness of a 'resource-rational' (or 'bounded-rational') picture of seemingly irrational human cognition.

Confidence resets reveal hierarchical adaptive learning in humans.

Hierarchical processing is pervasive in the brain, but its computational significance for learning under uncertainty is disputed. On the one hand, hierarchical models provide an optimal framework and are becoming increasingly popular to study cognition. On the other hand, non-hierarchical (flat) models remain influential and can learn efficiently, even in uncertain and changing environments. Here, we show that previously proposed hallmarks of hierarchical learning, which relied on reports of learned quantities or choices in simple experiments, are insufficient to categorically distinguish hierarchical from flat models. Instead, we present a novel test which leverages a more complex task, whose hierarchical structure allows generalization between different statistics tracked in parallel. We use reports of confidence to quantitatively and qualitatively arbitrate between the two accounts of learning. Our results support the hierarchical learning framework, and demonstrate how confidence can be a useful metric in learning theory.

Brain networks for confidence weighting and hierarchical inference during probabilistic learning.

Learning is difficult when the world fluctuates randomly and ceaselessly. Classical learning algorithms, such as the delta rule with constant learning rate, are not optimal. Mathematically, the optimal learning rule requires weighting prior knowledge and incoming evidence according to their respective reliabilities. This "confidence weighting" implies the maintenance of an accurate estimate of the reliability of what has been learned. Here, using fMRI and an ideal-observer analysis, we demonstrate that the brain's learning algorithm relies on confidence weighting. While in the fMRI scanner, human adults attempted to learn the transition probabilities underlying an auditory or visual sequence, and reported their confidence in those estimates. They knew that these transition probabilities could change simultaneously at unpredicted moments, and therefore that the learning problem was inherently hierarchical. Subjective confidence reports tightly followed the predictions derived from the ideal observer. In particular, subjects managed to attach distinct levels of confidence to each learned transition probability, as required by Bayes-optimal inference. Distinct brain areas tracked the likelihood of new observations given current predictions, and the confidence in those predictions. Both signals were combined in the right inferior frontal gyrus, where they operated in agreement with the confidence-weighting model. This brain region also presented signatures of a hierarchical process that disentangles distinct sources of uncertainty. Together, our results provide evidence that the sense of confidence is an essential ingredient of probabilistic learning in the human brain, and that the right inferior frontal gyrus hosts a confidence-based statistical learning algorithm for auditory and visual sequences.

Human Inferences about Sequences: A Minimal Transition Probability Model.

The brain constantly infers the causes of the inputs it receives and uses these inferences to generate statistical expectations about future observations. Experimental evidence for these expectations and their violations include explicit reports, sequential effects on reaction times, and mismatch or surprise signals recorded in electrophysiology and functional MRI. Here, we explore the hypothesis that the brain acts as a near-optimal inference device that constantly attempts to infer the time-varying matrix of transition probabilities between the stimuli it receives, even when those stimuli are in fact fully unpredictable. This parsimonious Bayesian model, with a single free parameter, accounts for a broad range of findings on surprise signals, sequential effects and the perception of randomness. Notably, it explains the pervasive asymmetry between repetitions and alternations encountered in those studies. Our analysis suggests that a neural machinery for inferring transition probabilities lies at the core of human sequence knowledge.

The Sense of Confidence during Probabilistic Learning: A Normative Account.

Learning in a stochastic environment consists of estimating a model from a limited amount of noisy data, and is therefore inherently uncertain. However, many classical models reduce the learning process to the updating of parameter estimates and neglect the fact that learning is also frequently accompanied by a variable "feeling of knowing" or confidence. The characteristics and the origin of these subjective confidence estimates thus remain largely unknown. Here we investigate whether, during learning, humans not only infer a model of their environment, but also derive an accurate sense of confidence from their inferences. In our experiment, humans estimated the transition probabilities between two visual or auditory stimuli in a changing environment, and reported their mean estimate and their confidence in this report. To formalize the link between both kinds of estimate and assess their accuracy in comparison to a normative reference, we derive the optimal inference strategy for our task. Our results indicate that subjects accurately track the likelihood that their inferences are correct. Learning and estimating confidence in what has been learned appear to be two intimately related abilities, suggesting that they arise from a single inference process. We show that human performance matches several properties of the optimal probabilistic inference. In particular, subjective confidence is impacted by environmental uncertainty, both at the first level (uncertainty in stimulus occurrence given the inferred stochastic characteristics) and at the second level (uncertainty due to unexpected changes in these stochastic characteristics). Confidence also increases appropriately with the number of observations within stable periods. Our results support the idea that humans possess a quantitative sense of confidence in their inferences about abstract non-sensory parameters of the environment. This ability cannot be reduced to simple heuristics, it seems instead a core property of the learning process.

Event-Related Potential, Time-frequency, and Functional Connectivity Facets of Local and Global Auditory Novelty Processing: An Intracranial Study in Humans.

Auditory novelty detection has been associated with different cognitive processes. Bekinschtein et al. (2009) developed an experimental paradigm to dissociate these processes, using local and global novelty, which were associated, respectively, with automatic versus strategic perceptual processing. They have mostly been studied using event-related potentials (ERPs), but local spiking activity as indexed by gamma (60-120 Hz) power and interactions between brain regions as indexed by modulations in beta-band (13-25 Hz) power and functional connectivity have not been explored. We thus recorded 9 epileptic patients with intracranial electrodes to compare the precise dynamics of the responses to local and global novelty. Local novelty triggered an early response observed as an intracranial mismatch negativity (MMN) contemporary with a strong power increase in the gamma band and an increase in connectivity in the beta band. Importantly, all these responses were strictly confined to the temporal auditory cortex. In contrast, global novelty gave rise to a late ERP response distributed across brain areas, contemporary with a sustained power decrease in the beta band (13-25 Hz) and an increase in connectivity in the alpha band (8-13 Hz) within the frontal lobe. We discuss these multi-facet signatures in terms of conscious access to perceptual information.

Neurocomputational account of how the human brain decides when to have a break.

No pain, no gain: cost-benefit trade-off has been formalized in classical decision theory to account for how we choose whether to engage effort. However, how the brain decides when to have breaks in the course of effort production remains poorly understood. We propose that decisions to cease and resume work are triggered by a cost evidence accumulation signal reaching upper and lower bounds, respectively. We developed a task in which participants are free to exert a physical effort knowing that their payoff would be proportional to their effort duration. Functional MRI and magnetoencephalography recordings conjointly revealed that the theoretical cost evidence accumulation signal was expressed in proprioceptive regions (bilateral posterior insula). Furthermore, the slopes and bounds of the accumulation process were adapted to the difficulty of the task and the money at stake. Cost evidence accumulation might therefore provide a dynamical mechanistic account of how the human brain maximizes benefits while preventing exhaustion.

Better get back to work: a role for motor beta desynchronization in incentive motivation.

Much research has been devoted to characterizing brain representations of reward and movement. However, the mechanisms allowing expected rewards to influence motor commands remain poorly understood. Unraveling such mechanisms is crucial to providing explanations of how behavior can be driven by goals, hence accounting for apathy cases in clinics. Here, we propose that the reduction of motor beta synchrony (MBS) before movement onset could participate in this incentive motivation process. To test this hypothesis, we recorded brain activity using magnetoencenphalography (MEG) while human participants were exerting physical effort to win monetary incentives. Knowing that the payoff was proportional to the time spent above a target force, subjects spontaneously took breaks when exhausted and resumed effort production when repleted. Behavioral data indicated that the rest periods were shorter when higher incentives were at stake. MEG data showed that the amplitude of MBS reduction correlated to both incentive level and rest duration. Moreover, the time of effort initiation could be predicted by MBS reduction measured at the beginning of rest periods. Incentive effects on MBS reduction and rest duration were also correlated across subjects. Finally, Bayesian comparison between possible causal models suggested that MBS reduction mediates the impact of incentive level on rest duration. We conclude that MBS reduction could represent a neural mechanism that speeds the initiation of effort production when the effort is more rewarded.

How the brain decides when to work and when to rest: dissociation of implicit-reactive from explicit-predictive computational processes.

A pervasive case of cost-benefit problem is how to allocate effort over time, i.e. deciding when to work and when to rest. An economic decision perspective would suggest that duration of effort is determined beforehand, depending on expected costs and benefits. However, the literature on exercise performance emphasizes that decisions are made on the fly, depending on physiological variables. Here, we propose and validate a general model of effort allocation that integrates these two views. In this model, a single variable, termed cost evidence, accumulates during effort and dissipates during rest, triggering effort cessation and resumption when reaching bounds. We assumed that such a basic mechanism could explain implicit adaptation, whereas the latent parameters (slopes and bounds) could be amenable to explicit anticipation. A series of behavioral experiments manipulating effort duration and difficulty was conducted in a total of 121 healthy humans to dissociate implicit-reactive from explicit-predictive computations. Results show 1) that effort and rest durations are adapted on the fly to variations in cost-evidence level, 2) that the cost-evidence fluctuations driving the behavior do not match explicit ratings of exhaustion, and 3) that actual difficulty impacts effort duration whereas expected difficulty impacts rest duration. Taken together, our findings suggest that cost evidence is implicitly monitored online, with an accumulation rate proportional to actual task difficulty. In contrast, cost-evidence bounds and dissipation rate might be adjusted in anticipation, depending on explicit task difficulty.

Two Determinants of Dynamic Adaptive Learning for Magnitudes and Probabilities.

Humans face a dynamic world that requires them to constantly update their knowledge. Each observation should influence their knowledge to a varying degree depending on whether it arises from a stochastic fluctuation or an environmental change. Thus, humans should dynamically adapt their learning rate based on each observation. Although crucial for characterizing the learning process, these dynamic adjustments have only been investigated empirically in magnitude learning. Another important type of learning is probability learning. The latter differs from the former in that individual observations are much less informative and a single one is insufficient to distinguish environmental changes from stochasticity. Do humans dynamically adapt their learning rate for probabilities? What determinants drive their dynamic adjustments in magnitude and probability learning? To answer these questions, we measured the subjects' learning rate dynamics directly through real-time continuous reports during magnitude and probability learning. We found that subjects dynamically adapt their learning rate in both types of learning. After a change point, they increase their learning rate suddenly for magnitudes and prolongedly for probabilities. Their dynamics are driven differentially by two determinants: change-point probability, the main determinant for magnitudes, and prior uncertainty, the main determinant for probabilities. These results are fully in line with normative theory, both qualitatively and quantitatively. Overall, our findings demonstrate a remarkable human ability for dynamic adaptive learning under uncertainty, and guide studies of the neural mechanisms of learning, highlighting different determinants for magnitudes and probabilities.

Resource-rational account of sequential effects in human prediction.

An abundant literature reports on 'sequential effects' observed when humans make predictions on the basis of stochastic sequences of stimuli. Such sequential effects represent departures from an optimal, Bayesian process. A prominent explanation posits that humans are adapted to changing environments, and erroneously assume non-stationarity of the environment, even if the latter is static. As a result, their predictions fluctuate over time. We propose a different explanation in which sub-optimal and fluctuating predictions result from cognitive constraints (or costs), under which humans however behave rationally. We devise a framework of costly inference, in which we develop two classes of models that differ by the nature of the constraints at play: in one case the precision of beliefs comes at a cost, resulting in an exponential forgetting of past observations, while in the other beliefs with high predictive power are favored. To compare model predictions to human behavior, we carry out a prediction task that uses binary random stimuli, with probabilities ranging from 0.05 to 0.95. Although in this task the environment is static and the Bayesian belief converges, subjects' predictions fluctuate and are biased toward the recent stimulus history. Both classes of models capture this 'attractive effect', but they depart in their characterization of higher-order effects. Only the precision-cost model reproduces a 'repulsive effect', observed in the data, in which predictions are biased away from stimuli presented in more distant trials. Our experimental results reveal systematic modulations in sequential effects, which our theoretical approach accounts for in terms of rationality under cognitive constraints.

Multiple and subject-specific roles of uncertainty in reward-guided decision-making.

Decision-making in noisy, changing, and partially observable environments entails a basic tradeoff between immediate reward and longer-term information gain, known as the exploration-exploitation dilemma. Computationally, an effective way to balance this tradeoff is by leveraging uncertainty to guide exploration. Yet, in humans, empirical findings are mixed, from suggesting uncertainty-seeking to indifference and avoidance. In a novel bandit task that better captures uncertainty-driven behavior, we find multiple roles for uncertainty in human choices. First, stable and psychologically meaningful individual differences in uncertainty preferences actually range from seeking to avoidance, which can manifest as null group-level effects. Second, uncertainty modulates the use of basic decision heuristics that imperfectly exploit immediate rewards: a repetition bias and win-stay-lose-shift heuristic. These heuristics interact with uncertainty, favoring heuristic choices under higher uncertainty. These results, highlighting the rich and varied structure of reward-based choice, are a step to understanding its functional basis and dysfunction in psychopathology.

Brainstem fMRI signaling of surprise across different types of deviant stimuli.

Detection of deviant stimuli is crucial to orient and adapt our behavior. Previous work shows that deviant stimuli elicit phasic activation of the locus coeruleus (LC), which releases noradrenaline and controls central arousal. However, it is unclear whether the detection of behaviorally relevant deviant stimuli selectively triggers LC responses or other neuromodulatory systems (dopamine, serotonin, and acetylcholine). We combine human functional MRI (fMRI) recordings optimized for brainstem imaging with pupillometry to perform a mapping of deviant-related responses in subcortical structures. Participants have to detect deviant items in a "local-global" paradigm that distinguishes between deviance based on the stimulus probability and the sequence structure. fMRI responses to deviant stimuli are distributed in many cortical areas. Both types of deviance elicit responses in the pupil, LC, and other neuromodulatory systems. Our results reveal that the detection of task-relevant deviant items recruits the same multiple subcortical systems across computationally different types of deviance.

Studying the neural representations of uncertainty.

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 and functionality. Our analysis reveals that the two approaches lead to different but complementary findings, shaping new research questions and guiding future experiments.

Rational arbitration between statistics and rules in human sequence processing.

Detecting and learning temporal regularities is essential to accurately predict the future. A long-standing debate in cognitive science concerns the existence in humans of a dissociation between two systems, one for handling statistical regularities governing the probabilities of individual items and their transitions, and another for handling deterministic rules. Here, to address this issue, we used finger tracking to continuously monitor the online build-up of evidence, confidence, false alarms and changes-of-mind during sequence processing. All these aspects of behaviour conformed tightly to a hierarchical Bayesian inference model with distinct hypothesis spaces for statistics and rules, yet linked by a single probabilistic currency. Alternative models based either on a single statistical mechanism or on two non-commensurable systems were rejected. Our results indicate that a hierarchical Bayesian inference mechanism, capable of operating over distinct hypothesis spaces for statistics and rules, underlies the human capability for sequence processing.

Gated recurrence enables simple and accurate sequence prediction in stochastic, changing, and structured environments.

From decision making to perception to language, predicting what is coming next is crucial. It is also challenging in stochastic, changing, and structured environments; yet the brain makes accurate predictions in many situations. What computational architecture could enable this feat? Bayesian inference makes optimal predictions but is prohibitively difficult to compute. Here, we show that a specific recurrent neural network architecture enables simple and accurate solutions in several environments. This architecture relies on three mechanisms: gating, lateral connections, and recurrent weight training. Like the optimal solution and the human brain, such networks develop internal representations of their changing environment (including estimates of the environment's latent variables and the precision of these estimates), leverage multiple levels of latent structure, and adapt their effective learning rate to changes without changing their connection weights. Being ubiquitous in the brain, gated recurrence could therefore serve as a generic building block to predict in real-life environments.

Bifurcation in brain dynamics reveals a signature of conscious processing independent of report.

An outstanding challenge for consciousness research is to characterize the neural signature of conscious access independently of any decisional processes. Here we present a model-based approach that uses inter-trial variability to identify the brain dynamics associated with stimulus processing. We demonstrate that, even in the absence of any task or behavior, the electroencephalographic response to auditory stimuli shows bifurcation dynamics around 250-300 milliseconds post-stimulus. Namely, the same stimulus gives rise to late sustained activity on some trials, and not on others. This late neural activity is predictive of task-related reports, and also of reports of conscious contents that are randomly sampled during task-free listening. Source localization further suggests that task-free conscious access recruits the same neural networks as those associated with explicit report, except for frontal executive components. Studying brain dynamics through variability could thus play a key role for identifying the core signatures of conscious access, independent of report.