Title:	Dissociation between expected value and Bayesian posterior estimates during active avoidance in the brain.
DNr:	NAISS 2025/22-658
Project Type:	NAISS Small Compute
Principal Investigator:	Tobias Granwald <tobias.granwald@ki.se>
Affiliation:	Karolinska Institutet
Duration:	2025-04-25 – 2026-05-01
Classification:	30105
Keywords:

Abstract

In the proposed study (preregistered at https://osf.io/gwkj3), we will investigate the neural underpinnings of a posterior for active avoidance. In a previous study (Granwald et al., 2025; preprint; https://osf.io/preprints/psyarxiv/2zcjh), we established a method using a Bayesian model of learning to extract, from behavioural data, a prior expressing the probability with which negative outcomes are expected to be actively avoidable. In the current study, we will build on this approach to investigate the neural underpinnings of the posterior estimate of this prior, which is dynamically updated with experience. Previous studies have shown that the BOLD signals in the ventromedial prefrontal cortex (vmPFC) and hippocampus correlate with the value of the chosen option in decision-making tasks. As priors need to be maintained in memory, a process in which the hippocampus is centrally involved, we hypothesise that BOLD signals in the hippocampus will correlate better with the posterior at the time of deliberation than with the expected value of the chosen option. We will also explore the neural correlates of uncertainty updating at the time of the outcome.