Forming global estimates of self-performance from local confidence

Rouault, Marion; Dayan, Peter; Fleming, Stephen M.

doi:10.1038/s41467-019-09075-3

Cited by 114 publications

(208 citation statements)

References 61 publications

Supporting

Mentioning

166

Contrasting

Order By: Relevance

“…In other words, we contest that while first-order reaction time information is, under some experimental settings used by participants to generate a confidence judgement, when motor information is not available at all, it may be replaced by other, equally precise sources of information, closer to the strength of evidence (such as the probability of being correct (Sanders et al, 2016), the internal signal noise (Navajas et al, 2017) and the evidence in favour of the chosen response alternative (Peters et al, 2017). This admittedly speculative account is compatible with our capacity to form confidence estimates about decisions that are not directly linked to a transient motor action, for instance when controlling a brain machine interface (Schurger, Gale, Gozel, & Blanke, 2017) or when making global confidence judgments in ecological contexts (Rouault, Dayan, & Fleming, 2019).…”

Section: Differences With the Existing Literaturementioning

confidence: 86%

Response-related signals increase confidence but not metacognitive performance

Filevich

Koß

Faivre³

2019

Preprint

View full text Add to dashboard Cite

AbstractConfidence judgements are a central tool for research in metacognition. In a typical task, participants first perform perceptual (first-order) decisions and then rate their confidence in these decisions. The relationship between confidence and first-order accuracy is taken as measure of metacognitive performance. Confidence is often assumed to stem from decision-monitoring processes alone, but processes that co-occur with the first-order decision may also play a role in confidence formation. In fact, across a broad range of tasks, trials with quick reaction times to the first-order task are often judged with relatively higher confidence than those with slow responses. This robust finding suggests that confidence could be informed by a readout of reaction times in addition to decision-monitoring processes. To test this possibility, we assessed the contribution of response-related signals to confidence and, in particular, to metacognitive performance (i.e., a measure of the adequacy of these confidence judgements). In a factorial design, we measured the effect of making an overt (vs. covert) decision, as well as the effect of pairing a motor action to the stimulus about which the first-order decision is made. Against our expectations, we found no differences in overall confidence or metacognitive performance when first-order responses were covert as opposed to overt. Further, actions paired to visual stimuli presented led to higher confidence ratings, but did not affect metacognitive performance. These results suggest that some of the relationships between first-order decisional signals and confidence might indeed be correlational, and attributable to an upstream cognitive process, common to the two of them.

show abstract

Section: Differences With the Existing Literaturementioning

confidence: 86%

Response-related signals increase confidence but not metacognitive performance

Filevich

Koß

Faivre³

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…We now discuss those results in the context of previous studies, which requires to delineate precisely what aspects of confidence we investigated here. It is the confidence that accompanies probability estimates on a trial-by-trial basis, thus it is not "global" or related self-confidence (Rouault et al, 2019). It is also not related to whether the stimulus is perceived clearly or ambiguously, which is important in general for inference (Kepecs et al, 2008;Mathys et al, 2014; 10/36 Meyniel et al, 2015b;Pouget et al, 2016;Bang and Fleming, 2018) but does not play a role here since stimuli were perceived without ambiguity.…”

Section: Discussion (1500)mentioning

confidence: 99%

Brain dynamics for confidence-weighted learning

Meyniel

2019

Preprint

View full text Add to dashboard Cite

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

show abstract

“…It has been previously proposed that next to perceptual evidence, inferences about "the state of the decider" (i.e., one's own actions 19 , and prior or global estimates of performance 92,93 ) are important for metacognitive decision-making. In addition, to adequately compute an estimate about the quality of a decision it is necessary to know the broader task context or infer "the state of the world" (i.e., value for an action at a certain state of the (task) environment) at the moment of the decision [94][95][96][97] .…”

Section: Models Of Metacognitive Decision-makingmentioning

confidence: 99%

Action information contributes to metacognitive decision-making

Wokke

Achoui

Cleeremans

2020

Sci Rep

View full text Add to dashboard Cite

Metacognitive abilities allow us to adjust ongoing behavior and modify future decisions in the absence of external feedback. Although metacognition is critical in many daily life settings, it remains unclear what information is actually being monitored and what kind of information is being used for metacognitive decisions. In the present study, we investigated whether response information connected to perceptual events contribute to metacognitive decision-making. Therefore, we recorded EEG signals during a perceptual color discrimination task while participants were asked to provide an estimate about the quality of their decision on each trial. Critically, the moment participants provided their confidence judgments varied across conditions, thereby changing the amount of action information (e.g., response competition or response fluency) available for metacognitive decisions. Results from three experiments demonstrate that metacognitive performance improved when firstorder action information was available at the moment metacognitive decisions about the perceptual task had to be provided. This behavioral effect was accompanied by enhanced functional connectivity (beta phase synchrony) between motor areas and prefrontal regions, exclusively observed during metacognitive decision-making. Our findings demonstrate that action information contributes to metacognitive decision-making, thereby painting a picture of metacognition as a process that integrates sensory evidence and information about our interactions with the world.

show abstract

Forming global estimates of self-performance from local confidence

Cited by 114 publications

References 61 publications

Response-related signals increase confidence but not metacognitive performance

Response-related signals increase confidence but not metacognitive performance

Brain dynamics for confidence-weighted learning

Action information contributes to metacognitive decision-making

Contact Info

Product

Resources

About