Social scientists now regularly employ computational models, most often to create proxy variables for theoretical concepts that are difficult to observe directly. This chapter discusses how the application of these models differs from traditional quantitative methods common to the social sciences. It then introduces a particular method—neural networks—by building from a model familiar to social scientists: logistic regression. To build intuition about how neural networks flexibly approximate complex functions, the chapter compares the performance of logistic regression to neural networks in a simple simulation. It then further demonstrates the flexibility of neural networks by introducing convolutional and recurrent neural networks and their application to image and text classification, respectively. Throughout, the chapter relates approaches to machine learning with neural networks to common statistical practices in social science. It concludes with a discussion of new methodological challenges posed by reliance on neural networks and open areas of research.