We discuss a series of problems arising in the analysis of multi-agent interactions. In particular we present some mathematical tools that are useful in dealing with senile, myopic, silent and indistinguishable robots whose collaboration emerges by reacting to what they perceive in their neighborhood.