Abstract Dynamic Bayesian networks (DBNs) are a general model for stochastic processes with partially observed states. Belief filtering in DBNs is the task of inferring the belief state (i.e. the probability dis tribution over process states) based on incomplete and uncertain observations. In this article, we explore the idea of accelerating the filtering task by automatically exploiting causality in the process. We Figure 1: Example o consider a specific type of causal relation, called pas observation variabl sivity, which pertains to how state variables cause t + 1, respectively changes in other variables. We present the PassivityThe arrows describe based Selective Belief Filtering (PSBF) method, which maintains a factored belief representation and exploits passivity to perform selective updates as the prior in the the belief factors. PSBF is evaluated in bo unrolled variant be thetic processes and a simulatedinmulti-robot the successive house, where it outperformed alternative states and approxim methods by exploiting passivity.methods were developed propagated over tim Often, the ke