"lets try and describe some properties that would be desirable for the dynamics of the community to satisfy and then try and implement a system which provably satisfies them."

This might actually be a good idea. If LessWrong could beget the formulation of some theory of good online communities (not just a set of rules that make online communities look like real-world communities because they work), that would certainly say something for our collective instrumental rationality.