Many econometric works have attempted to identify a peer effect in a population. In this paper, we propose to compare the results obtained by two types of interpretable predictive approaches: (1) an averaging linear model which assumes that individuals are affected in a linear fashion by the actions and mean characteristics of their peers and (2) a generalist firstorder logic rule mining approach that has been adapted through the use of specific language biases. The experiments were carried out on a real data set concerning the choice of a medical specialty by medical students when they have to select their internship position.