Do you think it increases complexity (or reduces randomness) when salt in solution precipitates into salt crystals? How about when super-saturated cloudless skies precipitate thunderclouds?
dohn, I think these are instances of self-ordering systems (e.g., like Bernard cells), not self-organizing systems (e.g., living systems). In the instances you describe, we are seeing increased ordering, which reduces randomness; but ordering processes seem to run in the opposite direction from processes that increase complexity. It has been observed that "ordering" is not particularly useful in living systems for the reason that "ordered" systems have comparatively low information content: there's a lot of informational redundancy in a repeating pattern, for instance.