The effects of base rate neglect on sequential belief updating and real-world beliefs

Abstract

Base-rate neglect is a pervasive bias in judgment that is conceptualized as underweighting of prior information and can have serious consequences in real-world scenarios. This bias is thought to reflect variability in inferential processes but empirical support for a cohesive theory of base-rate neglect with sufficient explanatory power to account for longer-term and real-world beliefs is lacking. A Bayesian formalization of base-rate neglect in the context of sequential belief updating predicts that belief trajectories should exhibit dynamic patterns of dependence on the order in which evidence is presented and its consistency with prior beliefs. To test this, we developed a novel ‘urn-and-beads’ task that systematically manipulated the order of colored bead sequences and elicited beliefs via an incentive-compatible procedure. Our results in two independent online studies confirmed the predictions of the sequential base-rate neglect model: people exhibited beliefs that are more influenced by recent evidence and by evidence inconsistent with prior beliefs. We further found support for a noisy-sampling inference model whereby base-rate neglect results from rational discounting of noisy internal representations of prior beliefs. Finally, we found that model-derived indices of base-rate neglect—including noisier prior representation—correlated with propensity for unusual beliefs outside the laboratory. Our work supports the relevance of Bayesian accounts of sequential base-rate neglect to real-world beliefs and hints at strategies to minimize deleterious consequences of this pervasive bias.

Imprecise probabilistic inference from sequential data.

“…Instead, we should observe “order effects”: Two sequences that have the same length, t ≥ 2, and the same number of red rings, n R (with 0 < n R < t ), will not result in the same estimates if the red and green rings appear at different positions in the two sequences. We conduct an array of tests to look for such order effects in our subjects’ data and find that a majority of subjects do not exhibit significant order effects (for details, see the Method section), contrary to the results obtained by authors such as Ashinoff et al (2022) using the more familiar “balls and urns” paradigm. And indeed, when we estimate the parameters of the recursive quasi-Bayesian model (as discussed below), the value of the parameter ρ that best fits subjects’ data is close to 1.…”

Action prediction in psychosis

“…We speculate that patients’ reduced calibration of confidence to accuracy during the up-to-20% integration period may share a common mechanism with, and potentially contribute to, this bias. Combined with underweighting of prior beliefs 36 , reduced calibration could lead to reduced data-gathering, premature decisions, and endorsement of odd beliefs in their daily lives 37 . To explore this hypothesis, future studies could probe the influence of priors on the integration of information in movement kinematics (for an example of prior manipulation, see ref.…”

A Distinct Neural Code Supports Prospection of Future Probabilities During Instrumental Information-Seeking

“…A second important task feature is that we limited participants to a single additional sample rather than allowing them to request multiple samples. While this distinguishes our approach from studies examining how people terminate sampling (i.e., decide how much information to gather before making a choice (Edwards 1965; Huq, Garety, and Hemsley 1988; Roitman and Shadlen 2002; Furl and Averbeck 2011; Hanks, Kiani, and Shadlen 2014; Baker et al 2019; Kaanders et al 2021; Ashinoff et al 2022)), it allowed us to understand with greater experimental control how participants prospect about information gains over a single time step (e.g., avoiding systematic distortions and noise that may gradually accumulate over samples (Ashinoff et al 2022)).…”