Numerous propagation models in social networks describing social influence can be found in the literature, making it difficult to select an appropriate model in a given situation. Selecting the most relevant model requires the ability to objectively compare them. This comparison can only be made at the cost of describing models based on a common formalism and independant from them. We propose to use graph rewriting to formally describe the propagation mechanisms as local transformation rules applied according to a strategy. This approach makes complete sense when supported by a visual analytics framework dedicated to graph rewriting. The paper first presents several models and illustrates them through selected simulations. We then show how our visual analytics framework allows to interactively manipulate models, and underline their differences based on measures computed on simulation traces.