Enviar por SMS: Reward-guided learning with and without causal attribution