A Bayesian network is a probabilistic graphical model that represents variables and their conditional dependencies using a directed acyclic graph. It supports reasoning under uncertainty by combining graph structure with Bayesian probability computations.