Hiveminds: Modelling and Pondering

im doing something actually interesting this time

Hiveminds are a common cultural phenomenon seen primarily in sci-fi, and are conceptually pretty awesome. However, I've always wondered about which of the forms of collective intelligence are most effective at ruling. I've always said that "twitch chat is the worst type of hivemind," but I've never actually theoretically verified that statement. So what I'll do here is try to model hiveminds, and maybe give some commentary on the moral/political value of hiveminds at the end.

Formalizing

buzz buzz (earthbound reference???)

OK, when talking to my friends about certain ideas I have about hiveminds, a lot of the discussion came around to arguing about what a hivemind actually is.
Mathematically, we can think about a hivemind as a network (or set)

M

of state machines

m

₁, m₂, ... m_i that each take input from some environment

E

and some arbitrary amount of other state machines. These state machines each have an output goal

g

that they will try to reach (whether it be making it to a certain position or collecting a certain volume of water or whatever combination of those that you wish).

g

should contain some "metadata" about how to achieve

g

, such as the energy required for achieving

g

.

Philosophically, we can think about a hivemind as any network of minds that have a connection between them and form a collective intelligence (that is, an agent that works to achieve goal that results from the efforts of several agents).

Scientifically, we can think about a hivemind as an entity capable of creating some productivity measure

P

at the cost of some total amount of energy

J

. What's interesting about this is, for some hiveminds, resources collected that contribute towards

P

will contribute towards

J

, thus creating a positive feedback loop.

The 3 types of hiveminds

legend of zorlda

What we perceive as a hivemind is actually 3 different categories of """collective""" (see Overmind) intelligence. These are:

Swarms
Democracies
Overminds

Swarms

I think that the two geniune hiveminds on this list can be differentiated by their "level of connectedness." While democracies are more focused on each agent's connection to

E

, a swarm is defined by the connections between each agent.
We can call a given hivemind a swarm if each agent's output is more defined by its state relative to other states around it than

E

. Note that this is a very vague definition, which is kind of the problem with categorizing these hiveminds. For example, think of a hivemind that exists to collect sap from trees, each agent in this hivemind will diverge from other agents if they are too close, but will head towards areas of forestry that other agents are heading towards, as it appears more fruitful of an area. This is a combined with the fact that the hivemind will head towards trees in their vicinity, and also deduce whether the tree contains a fruitful amount of sap. This is an example of the vague difference between a swarm and a democracy, as the status of "primarily determined by

E

" isn't as clear as it being "primarily determined by other agents' state."

A more quintessential example of a swarm hivemind is that of a... swarm. If we look at a boid swarm , for example, we see how collective action and goal can be formed out of 3 simple rules.
To see this for myself, I decided to use the previously linked tutorial as a guideline for making this : a simple boid simulation with a predator that tracks the center of mass of the boid swarm. The code is yours for the forking if you want to modify it/tune some constants. The result I wanted was that the overall swarm would collectively avoid the predator if any boid near the predator would move away from it. This is because each boid reacts to the movement of a boid around it, thus creating a collective reaction to the predator's encroachment. Because the agents in a boid swarm are so simple, they do not produce incredibly complex results. However, if we create an agent with more states and transitions between those states, we can create a collective that is capable of performing more complex tasks.

It is my opinion that this hivemind is probably most efficient at large sizes. This will be discussed more in a later section , but long story short, communication between 2 agents is fastest in a swarm hivemind, as information propagates at the speed of reaction and will reach a given location in the shortest amount of time that information moving at that speed can.

Democracies

Democracies seem to be the most esoteric and rarest type of hivemind. While agents in a swarm are more focused on their connection to other agents, agents in a democracy are most focused on

E