A common way to look at society is as a network graph. The nodes are individual people, and the edges are the connections between people. This view can be further expanded by adding some more details, borrowed from machine learning (ML) style neural networks. Each node can have a set of properties specifying how the individual behaves in the network, very much like the weights and biases on the neurons in a neural network. The properties of a node will determine which other nodes to connect to, and how information is ingested, transformed, then emitted to other nodes, via the edges.

Networks of sufficient size and complexity can begin to have emergent properties. In neural networks from ML (in supervised learning at least), the desired emergent properties are specified beforehand (e.g. the categories we want from the output), and the specifics of the network configurations are determined through the process of training. This is, in a sense, a reverse engineering operation. In the abstracted society networks however, the process goes the other way. The state of each node and the configuration of the network comes first, from which the emergent properties, unknown at first, arise organically and evolve over time, often in unpredictable ways. One can see these properties as a major component of culture.

One could make the case that depending on the state of the network, there will be a set of emergent properties that is stable or natural, in the equilibrium sense, at any given time. Perturbations to the parameters of the network will cause the corresponding shifts in the equilibrium state of the emergent properties. Making things more complex is the fact that these perturbations sometimes come from the emergent properties of the network itself, in a self-feedback loop. Some changes to the network could cause drastic shifts to the emergent equilibrium state of the network, which then causes new changes to the network, and so on. This process can result in both extremely positive or devastating outcomes for the network.

Using this conceptual visualization, we can begin to frame some recent developments in the world. Here I’ll take rise of the Internet in the past two decades as an example. The Internet’s fundamental effect on the network version of our society can be seen as a massive network-wide expansion in the number of edges as well as the bandwidth of the edges between nodes. At the same time, the natural processing capacity of the individual nodes (i.e. the people) remained mostly constant, given our limited cognitive and social abilities.

In order to deal with the explosion in the number and size of the edges, the nodes have several strategies. One is to simply ignore the new and bigger edges of the expanded network, and carry on with the previous set of node parameters (this is more of less the Amish way). Another is to filter information randomly, which although is hypothetically possible, I don’t think anyone actually carries this out. Yet another way we can deal with this issue is by expanding our capacity to process information, that is, making the nodes bigger, more complex, and capable of operating at full capacity in the new network. This involves drastic cognitive improvements for humans, perhaps in the form of cybernetic enhancements, which unfortunately isn’t feasible with today’s technology. The strategy that was possible and became the most popular is to outsource the information filtering process to third parties. These would be the various ranking and recommendation algorithms, run by other people or companies, that determine what we see every day.

One thing to realize is that regardless of how the nodes choose to behave, this drastically altered version of the network (with more and larger edges) has a very different set of emergent properties and equilibrium state from the one before. In addition, depending on the we (the nodes) react, the network’s new emergent equilibrium state can shift to different positions. We have been moving towards some form of a new state ever since the initial change occurred, and the position of this new state is still in flux based on what the nodes do.

The question to ask then: is the emergent equilibrium state of our new network, one where there are too many edges for each node to deal with, and so nodes rely on third parties to process information, the one we want in the long term?

I don’t know the answer to this question. But I can bring up a point about networks in general, which is that the less influence an individual node has on the overall network, the more stable the network. In most neural networks of sufficient size, changing the weight and bias of a single neuron will not dramatically alter the behavior of the network (unless in rare pathological cases). Having most nodes outsource information processing to a few large, centralized entities effectively gives certain nodes (i.e. the leaders of these large companies) super influential powers in determining how the network as a whole functions. This could be good, as any improvements can quickly propagate throughout the network; but it could also be disastrous, as any pathology can also spread fast and wide.