We approach the heterogeneous treatment effect problem in a novel way. Instead of trying to obtain the optimal treatment regime, we seek an interpretable model for effect modification using the recently developed selective inference framework.