Bayesian inference with incomplete knowledge explains perceptual confidence and its deviations from accuracy

Khalvati, Koosha; Kiani, Roozbeh; Rao, Rajesh P. N.

doi:10.1038/s41467-021-25419-4

Cited by 45 publications

(37 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Instead, we were able to explain decision confidence within the decision circuitry by simply changing the prior beliefs within this framework (see also Khalvati et al, 2021). Thus, our findings add to the ongoing debate about the need for a separate metacognitive module to explain dissociations between accuracy and decision confidence.…”

Section: Dissociations Between Accuracy and Confidencementioning

confidence: 65%

“…previously that priors in these models might be an important factor to understand deviations in the computation of confidence (Moreno-Bote, 2010; Drugowitsch et al, 2014;Khalvati et al, 2021), empirical evidence for this claim has been lacking so far. Here, we provide the first empirical demonstration that inducing under-and overconfidence by means of changes in prior beliefs can be readily accounted for within dynamic probabilistic models.…”

Section: Discussionmentioning

confidence: 99%

“…Although we are not the first to acknowledge the role of priors in the computation of confidence (Moreno-Bote, 2010;Drugowitsch et al, 2014;Khalvati et al, 2021), so far empirical support for this notion has been lacking. Therefore, in the current work, we aimed to provide direct evidence that prior beliefs underlie under-and overconfidence by explicitly manipulating prior beliefs about task performance.…”

Section: Introductionmentioning

confidence: 95%

See 2 more Smart Citations

Manipulating prior beliefs causally induces under- and overconfidence

Marcke

Denmat

Verguts

et al. 2022

Preprint

View full text Add to dashboard Cite

Making a decision is invariably accompanied by a sense of confidence in that decision. There is widespread variability in the exact level of confidence, even for tasks that do not differ in objective difficulty. Such expressions of under- and overconfidence are of vital importance, as they are related to fundamental life outcomes. Yet, a clear account regarding the computational mechanisms underlying under- and overconfidence is currently missing. In the current work, we propose that prior beliefs in the ability to perform a task can explain why confidence can differ dramatically despite similar task performance. In two experiments, we provide evidence for this hypothesis by showing that manipulating prior beliefs about task performance in an induction phase causally influences reported levels of confidence in a testing phase, while leaving actual performance unaffected. This is true both when prior beliefs are influenced via manipulated feedback and by manipulating the difficulty of the task during the training phase. These results are accounted for within an accumulation-to-bound model by assuming an additional parameter controlling prior beliefs about task performance. Our results provide a fundamental mechanistic insight into the computations underlying over- and underconfidence.

show abstract

Section: Dissociations Between Accuracy and Confidencementioning

confidence: 65%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 95%

See 1 more Smart Citation

Manipulating prior beliefs causally induces under- and overconfidence

Marcke

Denmat

Verguts

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Again, the bias-free observer shows an apparent overconfidence bias. In addition, this bias increases as type 1 performance decreases, reminiscent of the classic hard-easy effect for confidence ( Lichtenstein and Fischhoff, 1977a ; for related analyses, see Soll, 1996 ; Merkle, 2009 ; Drugowitsch, 2016 ; Khalvati et al, 2021 ). At chance level performance, the overconfidence bias is exactly 0.25.…”

Section: Resultsmentioning

confidence: 99%

Reverse engineering of metacognition

Guggenmos

2022

eLife

View full text Add to dashboard Cite

The human ability to introspect on thoughts, perceptions or actions − metacognitive ability − has become a focal topic of both cognitive basic and clinical research. At the same time it has become increasingly clear that currently available quantitative tools are limited in their ability to make unconfounded inferences about metacognition. As a step forward, the present work introduces a comprehensive modeling framework of metacognition that allows for inferences about metacognitive noise and metacognitive biases during the readout of decision values or at the confidence reporting stage. The model assumes that confidence results from a continuous but noisy and potentially biased transformation of decision values, described by a confidence link function. A canonical set of metacognitive noise distributions is introduced which differ, amongst others, in their predictions about metacognitive sign flips of decision values. Successful recovery of model parameters is demonstrated, and the model is validated on an empirical data set. In particular, it is shown that metacognitive noise and bias parameters correlate with conventional behavioral measures. Crucially, in contrast to these conventional measures, metacognitive noise parameters inferred from the model are shown to be independent of performance. This work is accompanied by a toolbox (ReMeta) that allows researchers to estimate key parameters of metacognition in confidence datasets.

show abstract

“…Finally, our model suggests that the uncertainty estimation can be resolved at a perceptual level. A recent study introduced a value-based Bayesian framework based on partially observable Markov decision processes to explain the results in the earlier confidence study (Khalvati et al, 2021). In the framework, both the perceptual decisions and the opt-out decisions were based on the same hidden belief updated with the sensory inputs.…”

Section: Discussionmentioning

confidence: 99%

An Optimal Policy Model for Concurrent Uncertainty Estimation During Decision Making

Chen

et al. 2021

Preprint

View full text Add to dashboard Cite

We often postpone or even avoid making decisions when we feel uncertain. Uncertainty estimation is not an afterthought of decision making but a dynamic process that accompanies decision making in parallel and affects decision making. To study concurrent uncertainty estimation during decision making, we adapted the classic random-dots motion direction discrimination task to allow a reaction-time measure of uncertainty responses. Subjects were asked to judge whether a patch of random dots was moving left or right. In addition, they could seek assistance by choosing to look at a second stimulus that had the same direction but high coherence any time during the task. The task allows us to measure the reaction time of both the perceptual decisions and the uncertainty responses. The subjects were more likely to choose the uncertainty response when the motion coherence was low, while their reaction times of the uncertainty responses showed individual variations. To account for the subjects' behavior, we created an optimal policy decision model in which decisions are based on the value functions computed from the accumulated evidence using a drift-diffusion process. Model simulations captured key features of the subjects' choices, reaction times, and proportions of uncertainty responses. Varying model parameters explained individual variations in the subjects and the correlations between decision accuracy, proportions of uncertainty responses, and reaction times at the individual level. Our model links perceptual decisions and value-based decisions and indicates that concurrent uncertainty estimation may be based on comparisons between values of uncertainty responses and perceptual decisions, both of which may be derived from the same evidence accumulation process during decision making. It provides a theoretical framework for future investigations, including the ones that aim at the underlying neural mechanism.

show abstract

Bayesian inference with incomplete knowledge explains perceptual confidence and its deviations from accuracy

Cited by 45 publications

References 69 publications

Manipulating prior beliefs causally induces under- and overconfidence

Manipulating prior beliefs causally induces under- and overconfidence

Reverse engineering of metacognition

An Optimal Policy Model for Concurrent Uncertainty Estimation During Decision Making

Contact Info

Product

Resources

About