While theoretical models increasingly use Bayesian frameworks to explain neural processing, we lack understanding of how neuromodulators like serotonin implement these computational principles biologically. This thesis proposes that serotonin serves as a biological mechanism for precision modulation in predictive processing, tested through experiments on exploratory behavior, and psychedelic-altered states, and expanded with perceptual illusion models. Investigation of serotonergic modulation in rodent whisking behavior reveals how precision weighting influences sensory processing and active sensing. Using active inference, the research demonstrates that serotonin modulates the precision of sensory inputs and prior habits, regulating exploratory behavior. Robotics work translates these insights into artificial systems, implementing precision-weighted processing in a humanoid robot to enable adaptive sensorimotor control. Mathematical modeling of the rubber hand illusion shows how the brain arbitrates between competing sensory models via precision-weighted inference, explaining body ownership experiences. Analysis of psilocybin's neural effects demonstrates that this serotonergic agent elevates chaotic brain responses and increases transition rates between brain states, suggesting psychedelics decrease precision in hierarchical neural communication. The thesis concludes by exploring how the serotonergic system influences proprioceptive integration and body sense, evaluating alternative computational frameworks for understanding serotonergic function across neural and behavioral domains.
Filip Novický, born in the Czech Republic, completed his Bachelor's in Psychology at Masaryk University in 2020. A 2019 exchange at Bilkent University sparked his neuroscience interest, leading him to pursue a Master's in Cognitive Neuroscience at Maastricht University (2021). During his master's, he interned at University College London in Karl Friston's lab, developing a bistable perception model using active inference. Currently a PhD candidate in Computational Neuroscience at the Donders Institute, Radboud University, Filip's research is supported by the Marie Curie Serotonin & Beyond programme. Supervised by Fleur Zeldenrust, Judith Homberg, Pablo Lanillos, and Thomas Parr, he focuses on Bayesian modeling of brain dynamics and neuroimaging analyses. His research interests span computational modeling, neurobiology, and robotics, with expertise in precision-weighted Bayesian inference. Filip completed research internships at the University of Florence (2023) studying spatiotemporal brain activity patterns in mice, and at Monash University (2024) investigating psychedelics' effects on brain dynamics. He has published first-author papers in Frontiers in Neurorobotics and Cerebral Cortex.