This thesis introduces Dynamic Bayesian Mean Estimation (DBME), a decentralized probabilistic framework for distributed mean estimation in networked multi-agent systems. The framework is designed for scenarios where a set of agents, each collecting local observations, cooperatively infer a shared latent quantity while communicating only with their neighbors. DBME offers an alternative to consensus and PAC-based methods by expressing distributed inference through a probabilistic formulation that explicitly represents uncertainty. Each agent maintains a parametric belief whose dynamics follow Bayesian principles of evidence accumulation and variance contraction. Nodes start from a bootstrap prior encoding an empirical or subjective belief about their local mean.