Belief networks, also referred to as Bayesian networks, are a fonn of artificial intelligence that incorporates uncertainty through probability theory and conditional dependence. Variables are graphically rep resented by nodes, whereas conditional dependence relationships between the variables are represented by ar rows. A belief network is developed by first defining the variables in the domain and the relationships between those variables. The conditional probabilities of the states of the variables are then detennined for each com bination of parent states. During evaluation of the network, evidence may be entered at any node without concern about whether the variable is an input or output variable. The probability of each state for the remaining variables, where the state is unknown, is evaluated. An automated approach for the improvement of construction operations involving the integration of belief networks and computer simulation is described. In this application, the belief networks provide diagnostic functionality to the perfonnance analysis of the construction operations. Computer simulation is used to model the construction operations and to validate the changes to the operation recommended by the belief network. This paper introduces belief networks, a form of artificial intelligence that may be described as a probabilistic-based expert system. Characteristics of belief networks are discussed followed by the evaluation of a singly connected belief net work. The next section describes an application that was de veloped to automatically improve the performance of construc tion operations by integrating belief networks and computer simulation. The computer simulation is used to model the con struction operations and provide performance measures. The belief network evaluates the performance measures, deter mines the most likely causes of poor performance, and rec ommends changes to the simulation model. These changes are automatically incorporated into the simulation model and the simulation is rerun. This iterative process was developed to provide automated support during the experimental phase of modeling to the simulationist. Belief networks were first developed at Stanford University in the 1970s. They fell out of popular research during the 1980s, and have experienced resurgence in the 1990s. In brief, belief networks are a method of representing the dependence and independence among a collection of random variables, and to calculate the probabilities of those variables as evidence about their values accumulates. Applications for belief net works, such as diagnostics, forecasting, and decision support, have been demonstrated in fields such as medicine and soft ware development (Heckerman et al. 1995).