Swarms permeate our technological, biological, and sociological worlds. In swarms, individual members are  limited to sensing or communicating with nearby peers, and no direct group-wide coordination is possible. Yet swarms can exhibit coherent collective behavior, and are often highly effective at achieving a collective goal, even when individual members are unaware of it.  Swarms have inspired generations of researchers who seek to understand how a single mind becomes part of a collective, and practictioners who seek to improve and impact the world outside academia.

A common assumption in existing research on swarms is that the swarm is *homogeneous*, meaning all members have the same capabilities and decision-making procedures. In this talk, I will highlight some of my group's work, which in contrast explores the role of diversity in  swarms. Specifically, I will show how (1) human pedestrian traffic is better modeled by swarm algorithms taking individual differences into account, (2) reinforcement learning by individual swarm robots leads to swarms that are *rational*, and achieve better results, because the robots no longer make identical decisions, and (3) molecular robots (nanobots) can be built to obey Asimov's laws, despite individually lacking the computational  complexity to do so.