Project Details
Projekt
Identifying mechanisms underlying decision making variability using data-driven optimal control frameworks
Applicant
Professorin Dr. Georgia Koppe
Subject Area
Experimental and Theoretical Network Neuroscience
Term
since 2026
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 437610067
Multimodal data-driven models that infer dynamical systems from neural and behavioral time series provide surrogate models of the brain that can be analyzed and perturbed to yield mechanistic insights into brain-behavior relationships. When combined with optimal control theory, these approaches can pinpoint the key brain regions or cell assemblies that drive goal-directed behavior and determine the specific perturbations required to achieve a desired outcome under given constraints. By integrating a formal framework that couples a dynamical model with a cost function - defining behavioral objectives along with time and energy limitations - optimal control theory offers a rigorous method to steer the system from an initial state to a target state, thereby revealing potential mechanisms underlying behavior. From this perspective, variability in decision-making between animals or humans may arise from multiple factors: Differences in the wiring or architecture of the dynamical system models, variations in classical control parameters (such as planning horizon and energy expenditure), differences in the optimization function (such as optimizing different behavioral strategies), or even from the system harboring multiple similarly effective solutions, leading to intra-individual variability. Here, we want to study these sources of variability. We aim to advance a data-driven framework that integrates dynamical system inference with optimal control strategies. Preliminary work demonstrates the efficacy of our models in reliably reconstructing neural dynamics from short and noisy datasets as well as identifying meaningful variability in control policy solutions. We will advance and deploy this approach to study decision-making variability in animals and humans, as well as refine neuromodulation strategies by identifying personalized control policies that optimally adjust neural dynamics.
DFG Programme
Research Units
