Flexible Relations Between Confidence and Confidence RTs in Post-Decisional Models of Confidence: A Reply to Chen and Rahnev

This is the authors' abstract. We don't add key points for this paper.

When making decisions, humans are able to provide well-calibrated estimates of their accuracy, based on their decision confidence (Fleming, Weil, Nagy, Dolan, & Rees, 2010).This ability allows humans to guide subsequent decision-making behavior (Desender, Boldt, & Yeung, 2018;Folke, Jacobsen, Fleming, & De Martino, 2016).Studies investigating the underlying computations of decision confidence frequently point to a post-decisional locus: Post-decisional evidence confirming (contradicting) the decision tends to increase (decrease) decision confidence (Fleming, van der Putten, & Daw, 2018).In line with this, theoretical work suggests that confidence is computed exclusively post-decision (Pleskac & Busemeyer, 2010).It should be noted, however, that there is currently no strong empirical evidence for the claim that confidence is "exclusively" computed post-decision (see Xue, Zheng, Rafiei, & Rahnev, 2023 for evidence to the contrary).Many theories of decision making assume that humans accumulate noisy perceptual evidence sequentially, until the accumulated evidence reaches a predefined decision boundary (also known as accumulation-to-bound models; see Gold & Shadlen, 2007, for a review).Because the duration of the accumulation of evidence to the decision boundary depends on the strength of the evidence and the height of decision boundaries, such models naturally account for the relationship between decision response times and stimulus difficulty, and for speed-accuracy tradeoffs (e.g., Ratcliff & McKoon, 2008).One common approach to model decision confidence within accumulation-to-bound models is to allow the accumulation process to continue after boundary crossing (Pleskac & Busemeyer, 2010).This extension allows accumulation-to-bound models to explain varying confidence ratings for the same decisions and post-decision error monitoring (Yeung & Summerfield, 2012).When further assuming that the process of post-decisional accumulation continues until reaching a second set of confidence boundaries, it follows that confidence response times (cRTs) should be informative with respect to the process of post-decisional evidence accumulation, and the resulting confidence judgment (Herregods, Denmat, & Desender, 2023;Moran, Teodorescu, & Usher, 2015).In line with this hypothesis, Baranski and Petrusic (1998) already observed consistent, non-monotonic relationships between the level of decision confidence and cRTs across participants.Recently, Chen and Rahnev (2023) investigated the notion that confidence and cRTs are intrinsically related and showed a wide variety of correlations between confidence and cRTs, with some participants showing a negative relationship (i.e., low confidence associated with long cRTs), but others showing a positive relationship (i.e., high confidence associated with long cRTs).The authors revealed that these individual differences were related to the frequency with which each of the confidence options was used by each participant: Participants tended to have lower cRTs for confidence ratings they chose more often (and vice versa).Furthermore, they stated that "the crucial hypothesis underlying post-decisional evidence accumulation models is that high-confidence responses are inherently made faster" (Chen & Rahnev, 2023, p. 1).Given that our recent post-decision model Herregods et al. ( 2023) was explicitly mentioned as an example of such a model, we here put this claim to the test by fitting their data with our model.Below, we show that, contrary to the claim made by Chen and Rahnev, the Herregods et al. ( 2023) model can capture both positive and negative confidence-cRT correlations.