Abstract:Prior expectations can be used to improve perceptual judgments about ambiguous stimuli. However, little is known about if and how these improvements are maintained in dynamic environments in which the quality of appropriate priors changes from one stimulus to the next. Using a sound-localization task, we show that changes in stimulus predictability lead to arousal-mediated adjustments in the magnitude of prior-driven biases that optimize perceptual judgments about each stimulus. These adjustments depend on tas… Show more
“…This noradrenergic activity changes neurons' membrane potential (McGinley et al, 2015) and its slow fluctuations (Reimer et al, 2014), promoting the selectivity of sensory processing, akin to the neural gain model. Similarly, activation of the locus-coeruleus (and increased pupil size) promotes feature selectivity in the sensory domain (Krishnamurthy et al, 2017;Rodenkirch et al, 2019). Changes in pupil size are also related to changes in the processing of sensory information and performance during perceptual decision making tasks (de Gee et al, 2014(de Gee et al, , 2017, with different effects of phasic and tonic pupil size (van Kempen et al, 2019) as in the present results.…”
Learning in a changing and 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 this confidence to regulate learning: for a given surprise, the update is smaller when confidence is higher. We explored the human brain dynamics sub-tending such a confidenceweighting using magneto-encephalography. During our volatile probability learning task, subjects' confidence reports conformed with Bayesian inference. Several stimulus-evoked brain responses reflected surprise, and some of them were indeed further modulated by confidence. Confidence about predictions also modulated pupil-linked arousal and beta-range (15-30 Hz) oscillations, which in turn modulated specific stimulus-evoked surprise responses. Our results suggest thus that confidence about predictions modulates intrinsic properties of the brain state to amplify or dampen surprise responses evoked by discrepant observations. Meyniel et al., 2015b), the weight of evidence (Rohe et al., 2019), the precision of predictions (Iglesias et al., 2013;Mathys et al., 2014;Vossel et al., 2014) discussed at the end of this article.Here, we propose to use optimal Bayesian models as a benchmark to formalize, at a computational level, the learning process. In particular, we formalize the notion of discrepancy between
“…This noradrenergic activity changes neurons' membrane potential (McGinley et al, 2015) and its slow fluctuations (Reimer et al, 2014), promoting the selectivity of sensory processing, akin to the neural gain model. Similarly, activation of the locus-coeruleus (and increased pupil size) promotes feature selectivity in the sensory domain (Krishnamurthy et al, 2017;Rodenkirch et al, 2019). Changes in pupil size are also related to changes in the processing of sensory information and performance during perceptual decision making tasks (de Gee et al, 2014(de Gee et al, , 2017, with different effects of phasic and tonic pupil size (van Kempen et al, 2019) as in the present results.…”
Learning in a changing and 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 this confidence to regulate learning: for a given surprise, the update is smaller when confidence is higher. We explored the human brain dynamics sub-tending such a confidenceweighting using magneto-encephalography. During our volatile probability learning task, subjects' confidence reports conformed with Bayesian inference. Several stimulus-evoked brain responses reflected surprise, and some of them were indeed further modulated by confidence. Confidence about predictions also modulated pupil-linked arousal and beta-range (15-30 Hz) oscillations, which in turn modulated specific stimulus-evoked surprise responses. Our results suggest thus that confidence about predictions modulates intrinsic properties of the brain state to amplify or dampen surprise responses evoked by discrepant observations. Meyniel et al., 2015b), the weight of evidence (Rohe et al., 2019), the precision of predictions (Iglesias et al., 2013;Mathys et al., 2014;Vossel et al., 2014) discussed at the end of this article.Here, we propose to use optimal Bayesian models as a benchmark to formalize, at a computational level, the learning process. In particular, we formalize the notion of discrepancy between
“…The result that pupil diameter does not modulate sequential e ects seems to be at odds with previous work that has shown pupil-linked systems modulating how information from previous trials a ect choice (Krishnamurthy et al, 2017;Nassar et al, 2012). However, an important consideration is that in these tasks, trials were dependent over time such that participants should use information from past trials to maximize their reward.…”
Section: Cc-by-nc-ndcontrasting
confidence: 54%
“…Third, with regard to sequential e ects, pupil changes have been related to how humans integrate relevant information from previous trials to infer uncertainty and expectation, suggesting a role for pupil-linked arousal systems in modulating sequential e ects (Krishnamurthy, Nassar, Sarode, & Gold, 2017;Nassar et al, 2012).…”
Integrating evidence over time is crucial for e ective decision making. For simple perceptual decisions, a large body of work suggests that humans and animals are capable of integrating evidence over time fairly well, but that their performance is far from optimal. This suboptimality is thought to arise from a number of di erent sources including (1) noise in sensory and motor systems, (2) unequal weighting of evidence over time, (3) order e ects from previous trials and (4) irrational side biases for one choice over another. In this work we investigated whether and how these di erent sources of suboptimality are related to pupil dilation, a putative correlate of norepinephrine tone. In particular, we measured pupil response in humans making a series of decisions based on rapidly-presented auditory information in an evidence accumulation task. We found that people exhibited all four types of suboptimality, but that only noise and the uneven weighting of evidence over time, the 'integration kernel', were related to the change in pupil response during the stimulus. Moreover, these two di erent suboptimalities were related to di erent aspects of the pupil signal, with the individual di erences in pupil response associated with individual di erences in integration kernel, while trial-by-trial fluctuations in pupil response were associated with trial-by-trial fluctuations in noise. These results suggest that di erent sources of suboptimality in human perceptual decision making are related to distinct pupil-linked processes possibly related to tonic and phasic norepinephrine activity.
“…However, the main assays in that study (cortisol and self-report) probed stress fluctuations operating on longer time horizons than the faster acting learning needed to update beliefs about environmental richness. This time constant misalignment similarly affects a commonly used alternative assay of putative stress states -galvanic skin conductance -while further confounds such as arousal and spontaneous fluctuations complicate inferences regarding stress system contribution to pupillometry data (Bradley et al, 2008;Joshi et al, 2016;Krishnamurthy et al, 2017).…”
Section: Introductionmentioning
confidence: 99%
“…Connections with other brainstem and cortical regions provide a pathway for surprising stimuli to drive arousal rather than stress specific stimuli (Krishnamurthy, Nassar, Sarode & Gold, 2017).…”
Appraising sequential offers relative to an unknown future opportunity and a time cost requires an optimization policy that draws on a learned estimate of an environment's richness. Converging evidence points to a learning asymmetry, whereby estimates of this richness update with a bias toward integrating positive information. We replicate this bias in a sequential foraging (prey selection) task and probe associated activation within two branches of the autonomic system, sympathetic and parasympathetic branches, using trial-by-trial measures of simultaneously recorded cardiac autonomic physiology. In general, lower value offers were accepted during periods of autonomic drive, both in the sympathetic (shorter pre-ejection period PEP) and parasympathetic (higher HF HRV) branches. In addition, we reveal a unique adaptive role for the sympathetic branch in learning. It was specifically associated with adaptation to a deteriorating environment: it correlated with both the rate of negative information integration in belief estimates and downward changes in moment-to-moment environmental richness, and was predictive of optimal performance on the task. The findings are consistent with a parallel processing framework whereby autonomic function serves both learning and executive demands of prey selection.
Significance statementThe value of choices (accepting a job) depends on context (richness of the current job market).Learning contexts, therefore, is crucial for optimal decision-making. Humans demonstrate a bias when learning contexts; we learn faster about improvements vs deteriorations. New techniques allow us to cleanly measure fast acting stress responses that might fluctuate with trial-by-trial learning. Using these new methods, we observe here that increased stress -specifically sympathetic (heart contractility) -might help overcome the learning bias (making us faster at learning contextual deterioration) and thereafter guide us toward better context appropriate decisions. For the first time we show that specific building blocks of good decision-making might benefit from short bursts of specific inputs of the stress system.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.