Some issues and varieties of computational and other approaches to understanding socially embedded phenomena are discussed.  It is argued that of all the approaches currently available, only agent-based simulation holds out the prospect for adequately representing and understanding phenomena such as social norms.