Dual strategies in human confidence judgments

Abstract

Although confidence is commonly believed to be an essential element in decision making, it remains unclear what gives rise to one’s sense of confidence. Recent Bayesian theories propose that confidence is computed, in part, from the degree of uncertainty in sensory evidence. Alternatively, observers can use physical properties of the stimulus as a heuristic to confidence. In the current study, we developed ideal observer models for either hypothesis and compared their predictions against human data obtained from psychophysical experiments. Participants reported the orientation of a stimulus, and their confidence in this estimate, under varying levels of internal and external noise. As predicted by the Bayesian model, we found a consistent link between confidence and behavioral variability for a given stimulus orientation. Confidence was higher when orientation estimates were more precise, for both internal and external sources of noise. However, we observed the inverse pattern when comparing between stimulus orientations: although observers gave more precise orientation estimates for cardinal orientations (a phenomenon known as the oblique effect), they were more confident about oblique orientations. We show that these results are well explained by a strategy to confidence that is based on the perceived amount of noise in the stimulus. Altogether, our results suggest that confidence is not always computed from the degree of uncertainty in one’s perceptual evidence, but can instead be based on visual cues that function as simple heuristics to confidence.

Subjective confidence reflects representation of Bayesian probability in cortex

“…Earlier work on statistical confidence has manipulated evidence reliability by varying physical properties of the stimulus, such as its contrast. This left open the possibility that observers simply monitor these image features as a proxy for uncertainty 7,[10][11][12] , without considering an internal belief distribution over the latent variable. For this reason, we held stimulus properties constant, relied on fluctuations in internal noise, and extracted probability distributions directly from cortical activity.…”

“…Eye tracking data were acquired using an SR Research Eyelink 1000 system for 62 out of 64 sessions. For 11 of these sessions, data were collected for 4-12 runs (out of a total of [10][11][12][13] due to technical difficulties with the eye-tracking system. Gaze position was sampled at 1 kHz.…”

“…However, a major limitation of this work has been the use of physical sources of noise, such as a variation in image brightness or contrast, to manipulate uncertainty. This is problematic because it could be that observers simply monitor such stimulus properties as external cues to uncertainty and confidence 7,[10][11][12][13] .…”

Suprathreshold perceptual decisions constrain models of confidence

“…While in some cases, the heuristic label has been applied to variants of the Evidence-Strength metric (Aitchison et al, 2015;Li and Ma, 2020), the majority of "heuristic" confidence models have focused on the use of stimulus cues to infer decision uncertainty. Support for this method of computing confidence comes from a series of studies demonstrating that observers over-or under-weight external noise in the stimulus when reporting confidence (De Gardelle and Mamassian, 2015;Spence et al, 2015;Boldt et al, 2017;Bertana et al, 2021). That is, they display a dissociation between Type 1 and Type 2 performance such that, if perceptual performance is matched for two stimuli with di↵erent levels of external noise, confidence is not equated.…”

Variance misperception under skewed empirical noise statistics explains overconfidence in the visual periphery

“…through the triple-threshold-versus-contrast function (triple-TVC) approach (Dosher & Lu, 2000, 2017; Lu & Dosher, 1998, 2008) and double-pass procedures (Awwad Shiekh Hasan et al, 2012; Gold et al, 2004; Levi & Klein, 2003; Ratcliff et al, 2018; Vilidaite & Baker, 2017) to quantify internal noise. It would be necessary to conduct these procedures under a range of attentional manipulations, in both the center and visual periphery, and across a large range of tasks in the same observer (i.e., a within-subjects design) in order to accurately measure the shape of the internal noise distributions and whether and how they may change conditioned on visual field location or even stimulus type (Bertana et al, 2020). Second, this approach would need to be coupled with established procedures for ‘recovering’ the prior used by an observer in psychophysical tasks in order to compare the true empirical prior to the prior used by the observer.…”