Neural Mechanisms for Integrating Prior Knowledge and Likelihood in Value-Based Probabilistic Inference

Abstract

In Bayesian decision theory, knowledge about the probabilities of possible outcomes is captured by a prior distribution and a likelihood function. The prior reflects past knowledge and the likelihood summarizes current sensory information. The two combined (integrated) form a posterior distribution that allows estimation of the probability of different possible outcomes. In this study, we investigated the neural mechanisms underlying Bayesian integration using a novel lottery decision task in which both prior knowledge and likelihood information about reward probability were systematically manipulated on a trial-by-trial basis. Consistent with Bayesian integration, as sample size increased, subjects tended to weigh likelihood information more compared with prior information. Using fMRI in humans, we found that the medial prefrontal cortex (mPFC) correlated with the mean of the posterior distribution, a statistic that reflects the integration of prior knowledge and likelihood of reward probability. Subsequent analysis revealed that both prior and likelihood information were represented in mPFC and that the neural representations of prior and likelihood in mPFC reflected changes in the behaviorally estimated weights assigned to these different sources of information in response to changes in the environment. Together, these results establish the role of mPFC in prior-likelihood integration and highlight its involvement in representing and integrating these distinct sources of information.

Adaptable history biases in human perceptual decisions

“…In games where reinforcement learning has been used to model behavior, animal (48,49), and human (8) subjects often display success-stay/fail-switch strategy similar to those observed in our experiments. Our task differs from typical reinforcement learning tasks in that it requires weighing sensory evidence and choice options where the latter have no explicit value or probability associated with them (50). Nonetheless, humans may perform the task by assigning value to choice options (51) or by treating successful trials as rewards.…”

Sub-optimality in motor planning is retained throughout 9 days practice of 2250 trials

“…Little is known about the neural representation involved in the integration of these 2 sets of information. Ting et al 49. investigated the neural mechanisms underlying Bayesian integration using a decision-making task in which participants were required to estimate the reward probability by combining past experience (prior knowledge) with current sensory information (likelihood), and showed that the mPFC plays a role in this Bayesian integration.…”

Base-rate neglect and neural computations for subjective weight in probabilistic inference

“…They found that multivoxel pattern of activity in OFC represents the log posterior probability distribution on latent causes. In our previous study (Ting et al, 2015), we found that mPFC correlated with the mean of the posterior distribution on probability of reward. The present study adds to this literature by showing that mPFC represents the subjective weight necessary for the integration of prior and likelihood information when these two sources of information were presented.…”

“…In cognitive neuroscience, there is a growing interest in studying the neurocomputational substrates involved in a variety of inference tasks, from how people use cue reliability as prior information to guide perceptual decision making (Forstmann, Brown, Dutilh, Neumann, & Wagenmakers, 2010;Mulder, Wagenmakers, Ratcliff, Boekel, & Forstmann, 2012), make financial decisions based on prior information about partner's reputation (Fouragnan et al, 2013), combine prior and likelihood information about reward probability (d 'Acremont, Schultz, & Bossaerts, 2013;Ting, Yu, Maloney, & Wu, 2015), to infer latent causes (Chan, Niv, & Norman, 2016) and other people's intentions (Chambon et al, 2017). These studies pointed to the role of medial prefrontal cortex (mPFC) and orbitofrontal cortex (OFC) in inference, from representing prior information (Forstmann et al, 2010;Fouragnan et al, 2013;Ting et al, 2015;Vilares, Howard, Fernandes, Gottfried, & Kording, 2012), current observation or sensory evidence (d' Acremont et al, 2013;Ting et al, 2015) to the combination of prior and current information (Chambon et al, 2017;Chan et al, 2016;Ting et al, 2015).…”

“…First, in task paradigms suitable for neurobiological investigations, the field has yet to see paradigm(s) that robustly reveals or replicates base-rate neglect. Across the studies highlighted above, subjects either achieved near-optimal performance in combining prior and current information (Chan et al, 2016;Ting et al, 2015) or it is unclear whether subjects achieved near-optimal performance (Chambon et al, 2017;d'Acremont et al, 2013;Forstmann et al, 2010;Fouragnan et al, 2013;Mulder et al, 2012;Vilares et al, 2012). To address this challenge, it is critical to identify the statistical properties of prior and likelihood information such that, when manipulated, produce robust base-rate neglect.…”