Evidence integration and decision-confidence are modulated by stimulus consistency

Abstract

Evidence-integration is a normative algorithm for choosing between alternatives with noisy evidence, which has been successful in accounting for a vast amount of behavioral and neural data. However, this mechanism has been challenged as tracking integration boundaries sub-serving choice has proven elusive. Here we first show that the decision boundary can be monitored using a novel, model-free behavioral method, termed Decision-Classification Boundary. This method allowed us to both provide direct support for evidence-integration contributions and to identify a novel integration-bias, whereby incoming evidence is modulated based on its consistency with evidence from preceding time-frames. This consistency bias was supported in three cross-domain experiments, involving decisions with perceptual and numerical evidence, which showed that choice-accuracy and decision confidence are modulated by stimulus consistency. Strikingly, despite its seeming sub-optimality, this bias fosters performance by enhancing robustness to integration noise. We argue this bias constitutes a new form of micro-level, within-trial, confirmation bias and discuss implications to broad aspects of decision making.

Transient Value Refinements during Deliberation Facilitate Choice

“…This interpretation is consistent with a computational mechanism by which coherence shifts might occur. In their comparison of evidence integration models of choice, Glickman, Moran, and Usher (2020) showed that the best model was one that over-weighted new evidence when it was consistent with previous evidence. Surprisingly, such apparent distortion of evidence led to choices with both higher accuracy and higher confidence.…”

“…While the distortion discussed in their study was related to the difference in value between options, it is possible that a similar distortion might take place within the individual value signals (for each option) before they are compared. Such distortions might provide a benefit to the decision system by dampening the impact of neural noise (Glickman et al, 2020), causing information about the positive and negative attributes of each option to be accentuated. Future work will be needed to directly test this intriguing hypothesis.…”

Within and Cross-Domain Effects of Choice-Induced Bias

“…One prominent theory suggests that CB might serve to facilitate action implementation [46]. Strikingly, a recent paper suggests that CB-driven evidence accumulation can sometimes enhance performance from purely computational perspective in both perceptual and higher-order inference tasks [12]. On a broader level, CIB is often considered as an instance of confirmation bias [11,61], which has been proposed to to facilitate group cooperation and stability [48].…”

“…(subjective preference: [3,5,6,7], perception: [8,9,10], higher cognitive inference [10,11]), and across timeframes, from immediate, trial-level effects [11,12], to long-lasting changes in preference [13]. CIB can affect both cognitive representations [14,15,16] and neural activity [17,18,19,20].…”

Value Certainty in Drift-Diffusion Models of Preferential Choice

“…Future work will be needed to examine the neural mechanism that extracts the drift rate from fluctuating values (sampled from memory or prospective imagination; Bakkour et al, 2019;Poldrack et al, 2001;Schacter, Addis, & Buckner, 2007) and that reduces the drift rate of more strongly fluctuating items. One interesting possibility is that the effective drift rate might be modulated by the temporal congruency of the evidence in successive samples (Glickman, Moran, & Usher, 2020); the higher the value certainty, the lower the variance of the value signals, thus leading to a higher probability that successive samples will provide consistent choice evidence.…”