Is kind of paradoxical when you think about it.
How can you design local interactions that add up to unpredictable global outcomes, when by definition the outcomes are unknowable.
This paper establishes a research question :
Suggests that there are 3 approaches :
- borrow from nature (bio-inspiration)
- GeneticAlgorithms / adaptation
- trial and error
(not sure I see much difference between 2 & 3 to be honest)
Futher thoughts, along a HerbertSimon hierarchy / modularity lines.
What if you decompose the gap between the micro-scale and the macro-scale into intermediate MiddleScales?
For example ... you want an effect when 1000s of agents combine. But can you experiment with a unit of 10 agents? And then another experiment with 100s of those groups of 10?
Maybe hopeless and lose all the virtues of emergence.
OTOH, DivideAndConquer is a classic DesignStrategy