When intelligent systems are built out of multi-agent systems, and explanations are required, self-explaining agents are the only sensible way out.