Emergence as Self-Domestication

Status: Idea

We classically think of emergence - simply defined as when something is larger than the sum of its parts - as a bottom-up concept. We have some parts that interact in some way, and then somehow magically they start to behave in a very complex, unpredictable way. But where does the magic come from?

The alternative viewpoint is top-down, where we assume some large-scale concept, like the self, is controlling the lower-level parts. I make a decision, and I move my arm, and then my arm moves. This approach has more explanatory power - assuming such an abstraction often gives better predictions - but it also seems weird. How can I have free will if physics only determine the behavior of the lowest-level parts?

Which is it then?

I am proposing that all emergent behavior arises from a very simple interaction, which I call self-domestication.

First, assume that any system follows evolution. Or energy minimization, gradient descent, physics, or however you wanna call it. Everything obeys some form of gravity through some space. For example, take the simple system of a rolling marble. It’s falling through space, in this case, actual physical space. If it can roll down, it will roll down., and if there is something in the way, it will just not move.

(This space can be abstract too. For example, the germline of an animal is moving through evolutionary space. If the environment changes - let’s say when humans introduce fishing nets of a certain size - then the germline “falls down” and fish get smaller.)

But this space is relative and we can do this:

We can have systems that control the environment, the space, of another system. We can domesticate the marble. (domestication, from domus: home, to bring something into one’s household)

We can understand the space, the environment, the constraints, the allowable rules, the energy landscape, the tech tree. And we can understand the will, or physics, economics, energy flow, etc. In short, we understand what a marble can do, and what it wants to do. And that gives us full control over the system.

We can take a pig, or rather its germline, and move it through morphological space, using breeding (direct evolution). We can make its children fatter, by only pairing it with fat pigs. Each individual pig might still do exactly what it wants, but we disallow pairing with skinny pigs, and so the marble rolls. We can even reverse it by pairing skinny pigs and make it go in a loop! Control through environment modification.

(We can even model a chain of systems. Economics, which roughly maps to the wants of the individual, and culture, which maps to the human action space will tell you that pigs have gotten larger, the richer the world got.)

But what happens if you have a system that is self-domesticating?

Let's take the example of an ant colony. This ant colony is made of a bunch of ants (duh). How do you control them? Well, you figure out what they want (follow the smell) and you get:

(this is actually real)

The problem is, ants do this all the time. Ants change other ants’ environments. They are reinforcing each other's paths. They even come together and make group decisions, aka, they copy their environments from a collective. This leads them to do these highly no-linear things, like pack up all their stuff and just … move their nest. Or this:

In short, we can’t predict what ants will do next. The state they “fall through” contains many unsolvable differential equations, it is inherently divergent, chaotic, emergent.

I argue that humans are basically this. On multiple levels. Or even further: this bilateral, self-referential environment building, seems to be what it means to be alive. Crystals don’t have it (the crystal shapes the environment for individual ions, but the ions don’t shape what it means to be a crystal, it’s only top-down). Deserts don’t have it (the desert is the environment for the life inside, which rarely changes the desert itself, it’s only bottom-up). But organisms have it. Coral reefs. Ant colonies. Us.

P.s.: Does this also define consciousness? I’m not sure. It seems that consciousness only arises when the “emergency hierarchy” of a system has a certain shape, mainly that it has a single node at the highest layer, which does all of the information aggregation (introducing most of the non-linearity). I don’t think coral reefs are conscious, and the more I think about it maybe octopi are also not, or if they are they contain 8 consciousnesses. Ants clearly aren’t, but ant colonies might.

P.p.s: please someone help me formalize this?