Using Protected Attributes to Consider Fairness in Multi-Agent Systems