How do robots in a collective know what the group as a whole is doing? How can connected devices make sense of the world around them with so many interconnections? How can a robotic arm composed of many independent parts understand how its body is behaving as it reaches for an object? When intelligence is distributed across many parts, it can be tricky for the bigger picture to emerge. Yet answering these questions is key to making collective systems that are easy to design, monitor and control.