We have built some simple, but useful, cooperative Artificial Life age
nts. Based on this experience and by contrasting our work with compute
r viruses, we argue that Artificial Life (the simulation of life inclu
ding evolution) can only remain reliably and indefinitely cooperative
if it adheres to esplicitly-specified social conventions. Breaking or
neglecting these conventions results in systems that are worse than us
eless; in fact, malicious with respect to human social values.