The study of tax behavior is a research field which attracts increasing interest in social and behavioral sciences. Rational choice models have been traditionally used to account for that behavior, but they face the puzzle of explaining levels of observed tax compliance which are much higher than expected. Several social influence mechanisms have been proposed in order to tackle this problem. In this article, we discuss the interdisciplinary literature on this topic, and we claim that agent-based models are a promising tool in order to test theories and hypothesise in this field. To illustrate that claim, we present SIMULFIS, an agent-based model for the simulation of tax compliance that allows to combine rational choice with social influence mechanisms in order to generate aggregated patterns of tax behavior. We present and discuss the results of a simple virtual experiment in order to show the potentialities of the model.