Choices Change the Temporal Weighting of Decision Evidence

Abstract

Decisions do not occur in isolation, but are embedded in sequences of other decisions, often pertaining to the same source of evidence. Here, we characterized the impact of intermittent choices on the accumulation of a protracted stream of decision-relevant evidence towards a final decision. Human participants performed two versions, based on perceptual or numerical evidence, of a decision task that required two successive judgments at different times during the evidence stream: an intermittent response consisting of a binary choice, and a continuous estimation at the end of the evidence stream.In a control condition, subjects executed a choice-independent motor response instead of binary choice as the intermittent response. In both, perceptual and numerical tasks, the intermittent choice reduced the sensitivity of subsequent evidence, and flipped the relative temporal weighting of early and late evidence in the final estimation judgment. The individual extent of the choice-induced overall (nonselective) sensitivity reduction predicted the extent of the selective down-weighting of subsequent evidence inconsistent with the initial choice, a form of confirmation bias. In sum, active decisions about a protracted evidence stream profoundly alter the dynamics of evidence accumulation, consistent with an active, modulatory mechanism triggered by the choice.

Controllability boosts neural and cognitive signatures of changes-of-mind in uncertain environments

“…Similar benefits of active sampling have been reported in terms of memory performance (Voss et al, 2010), and in terms of transfer of prior knowledge to novel object learning (Xu and Tenenbaum, 2007). Active sampling has even been reported to bias perceptual sensitivity in favor of choice-consistent evidence (Talluri et al, 2020). Importantly, in contrast to this earlier work, we designed our paradigm to create tightly matched conditions of active and passive sampling in terms of the information provided, and we offered participants a limited space of hypotheses to choose from (Weiss et al, 2019).…”

Bump attractor dynamics underlying stimulus integration in perceptual estimation tasks

“…This transition could also be gradual, for example caused by a ramping I 0 (Finkelstein et al, 2019). Moreover, it has been shown in combined discrimination and estimation tasks that stimulus estimation is influenced by a categorical decision causing post-decision biases (Jazayeri and Movshon, 2007;Luu and Stocker, 2018;Talluri et al, 2018Talluri et al, , 2020. One hypothesis is that these bias effects are mediated through attention signals and through changes in global gain (Talluri et al, 2018(Talluri et al, , 2020).…”

“…Moreover, it has been shown in combined discrimination and estimation tasks that stimulus estimation is influenced by a categorical decision causing post-decision biases (Jazayeri and Movshon, 2007;Luu and Stocker, 2018;Talluri et al, 2018Talluri et al, , 2020. One hypothesis is that these bias effects are mediated through attention signals and through changes in global gain (Talluri et al, 2018(Talluri et al, , 2020). Our network model provides a comprehensive computational framework for investigating the neural mechanisms underlying stimulus estimation and perceptual categorization and their interaction in future studies.…”

“…the average orientation; but see (Ratcliff, 2018)). They are thus not well suited to study stimulus estimation tasks that require continuous perceptual judgements or tasks that require both the report of a continuous estimate and a categorical choice (Jazayeri and Movshon, 2007;Luu and Stocker, 2018;Talluri et al, 2020).…”

Filling the gaps: Cognitive control as a critical lens for understanding mechanisms of value-based decision-making

“…However, we don't make choices in a vacuum, and our current choices depend on previous choices we have made (Erev & Roth, 2014;Keung, Hagen, & Wilson, 2019;Talluri et al, 2020;Urai, Braun, & Donner, 2017;Urai, de Gee, Tsetsos, & Donner, 2019). In the standard view of value-based decision-making, the only way choices influence each other is through option value learning, where the evaluations of the outcome of one choice refines the value assigned to that option in future choices (Fontanesi, Gluth, et al, 2019;Fontanesi, Palminteri, et al, 2019;Miletic et al, 2021).…”