Project Details
Projekt
Signatures of network reset during dynamic belief updating in volatile and stochastic environments
Applicant
Dr. Jan Gläscher
Subject Area
Biological Psychology and Cognitive Neuroscience
Personality Psychology, Clinical and Medical Psychology, Methodology
Personality Psychology, Clinical and Medical Psychology, Methodology
Term
since 2023
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 461947532
Background: A fundamental problems of any agent is whether the current sensory evidence signals a change point in the environment that requires an adaptation of the current world model through dynamic belief updating (DynBU) or whether it represents a stochastic sample of a probabilistic, but stable process. These alternatives are referred to as unexpected and expected uncertainty and dissociating them is a hard problem, because for the learner both types of uncertainty manifest as noisier samples of sensory evidence. Unexpected uncertainty is triggered by unusually large prediction errors (PEs) indicating a change point. This activates the noradrenergic pupil-linked arousal system and leads to a cortical network reset. However, expected uncertainty can also elicit large PEs by sampling from the tails of a distribution, while the underlying mean remains unchanged. Furthermore, it remains unclear whether a network reset is triggered in the same way by distinct state and reward prediction errors (SPEs/RPEs) or whether the reset is limited to specific brain regions associated with their respective expectations and errors. Aims and objectives: This project aims to characterize DynBU and the network reset cascade in the context of unexpected and expected uncertainty in different PE domains (SPE/RPE). We will use the Confetti-Cannon-Task with changes the underlying dynamics of reward and state distribution and extend the existing computational model to include our two manipulations: joined inference about unexpected and expected uncertainty and separable neural correlates of change points in state and reward distributions. Main hypotheses: We hypothesize that only change points will trigger the full network reset cascade compared to changes in stochasticity. We also expect differential involvement of brainstem and midbrain nuclei when the network reset cascade is triggered by SPEs or RPEs respectively. Planned Methods: We will conduct 3 studies to investigate the activation of the pupil dilation responses (PDRs) and the cortical network reset in these four conditions. In Study 1, we will use eye movements as the response measure and investigate PDRs to unexpected and expected uncertainty for both types of PEs. We will then use combined EEG-fMRI measurements with eye-tracking to characterize the network reset cascade by contrasting unexpected and expected uncertainty (Study 2) and an RPE vs. SPE-induced network reset (Study 3). Expected impact: Because of its focus on the central mechanism of the entire Research Unit, the findings of this project will inform all other projects. Furthermore, our design that differentiates unexpected and expected uncertainty will provide the first empirical data and a computational model for how humans solve this joined inference problem. This will represent a seminal step forward in the field of human statistical learning and provide answers to the question how humans make decisions under both types of uncertainty.
DFG Programme
Research Units
