Emergent behaviors can be thought of as a form of distributed computing. It seems that for many large patterns of coordination, the computation required to achieve it - which may be very intensive to manage globally - can be safely and more efficiently offloaded to the parts.

The classic example of this is the economy. To centrally manage a non-trivially sized economy efficiently is a monumental task, both in terms of the information flow and the processing required. Yet through implementing a few simple rules, such as the private ownership of property and allowing prices to float freely, this “management work” can be successfully distributed to achieve a reasonably efficient emergent outcome. The result isn’t perfect by any stretch of the imagination, but in terms of the difficulty of implementation and the efficiency attained, the emergent way handily beats the centrally coordinated way.

More abstractly, let’s say we have some number of individual agents, each with a computation capacity of 100, working towards some goal at a desired velocity. To manage their coordination at the group level explicitly, it would require a planning agent with 1000 computation capacity. But no single agent has that much computation capacity (in this case computation capacity is like intelligence, and cannot simply be added together), so this seems like an impossible task. But if the goal they are working towards is an emergent property, then the goal becomes obtainable. By distributing the computation required, the agents can operate independently (within their individual computation capacity of 100) and still manage to produce a system that would take 1000 computation capacity to plan. This is the efficiency that emergence affords us.

To characterize the gain from emergence compared to explicitly coordinating things globally, we could imagine an emergent efficiency constant of the universe, similar to the cosmological constant in physics. A value greater than 1 means that it is possible for parts with low computation capacities to produce a system that has a higher effective computation capacity, without explicit coordination. And a value less than 1 would mean the opposite.

Our universe obviously has an emergent efficiency constant value greater than 1. We take it for granted because it seems so natural, but I find it pretty miraculous that we live in a world like this. In such a world, an emergent phenomenon is often outside the comprehension of each individual agent within the system, since the whole has a much higher computation capacity than the individuals. But through the greater than 1 emergent efficiency constant, we can nonetheless reap the benefits of such a complex system, something beyond our own understanding. It’s like cheating in a way, and it helps us bootstrap to higher levels of coordination faster.

It is difficult to picture what a universe with an emergent efficiency constant less than 1 would look like. But we can try with the free market economy example. In such a world, it would not be possible for an efficient free market to emerge out of some simple set of rules. Each of us acting independently creates more chaos and noise than useful information at the system level. Equilibriums may still exist, but they probably tend to random values, and certainly don’t result in the efficient allocation of resources. Centrally coordinating our trade is the best we can do, which is not very good, even at our current level of computation capacity. More over, the central planning would have to fight the additional noise created from the low emergent efficiency constant, so it’s even harder than it appears to us.