Can proximity make friendships more diverse? To address this question, we propose a learning-driven friendship formation model to study how proximity and similarity influence the likelihood of forming social connections. The model predicts that proximity affects more friendships between dissimilar than similar individuals, in opposition to a preference-driven version of the model. We use an experiment at selective boarding schools in Peru that generates random variation in the physical proximity between students to test these predictions. The empirical evidence is consistent with the learning model: while social networks exhibit homophily by academic achievement and poverty, proximity generates more diverse social connections.