This document discusses user models and predictive modeling techniques for interface design. It covers Fitts' law for predicting pointing and selection times based on target size and distance. The keystroke-level model is presented for estimating times of complex tasks as the sum of basic operations. Guiard's model of bimanual control and neural networks are described as ways to predict more complex interactions like menu placement. Predictive models allow mathematical performance predictions while descriptive models provide frameworks for understanding problems. Both have limitations and should be used alongside user testing.