The document proposes using neural networks to help an intelligent software agent called IDA learn to make high-quality decisions for assigning US Navy sailors to new jobs. Specifically, it evaluates using feedforward neural networks, multilayer perceptrons, and support vector machines to model the decision-making of human Navy personnel called detailers. The neural networks are trained on data from detailers involving sailors, jobs, and which jobs were offered to learn the constraints and preferences that guide human decisions. The goal is for the neural networks to help IDA's constraint satisfaction module make more human-like decisions over time as situations change.