Centralisation

Benefits of Centralisation

I will be arguing the seemingly obvious point that man centralises action when there is material benefit from doing so, with one major caveat being that things of scale have a will of their own. Just like the saying that says; "things that you own end up owning you" - actions can end up owning you. Think of a war, a business, a movement. To take action at a certain scale is to bring forth momentum and god a daemon or a God into being.

This is worthy to muse upon, as some that seek decentralised solutions to societal problems will at the same time depend on centralised systems. By musing on what centralisation is, its strengths and its negative aspects, one can better understand what is meant when it is said that a decentralised solution is sought, because invariably there is some context and nuance to the issue being spoken to that is not articulated.

The Evidence

It is not controversial to think of the benefits of centralisation being a motivator for the behaviour of peoples in the age of tribes and warlords. People wanted protection, and a good leader with whom men could win a war hoard. Nor is it controversial when we think of today's age of grift and fraud, when large companies employ psychology graduates to trick and pressure people into buying not-needed subscriptions and premium services so the company has better metrics in earning reports. People join such companies because they can earn a living to support their families.

When you seek to explain centralising behaviour in the age of the church, there does however exist controversy. An honest look at Italic oligarch practices of offering collection rights to mercenaries implies it is no exception. There is again controversy when you look at the highly centralised decision-making in Western societies, but again, an honest look reveals that the challenge of competently running a school, for example, within budget and balancing risk, is too great for a regular managerial schooled person, and so they instead quickly give up sovereign decision-making in favour of receiving grants and promissory notes of insurance in exchange for ticking boxes.

There are certain modes of behaviour, or playable actions in an action space, and these behaviours present themselves differently in different action spaces. The workforce of the capitalist company grifting is fundamentally no different to the thanes in the time of warlords, and is no different to the members of a religious congregation, and is no different to conditional grant money accepting school managers. Appropriate actions that provided needed security were taken by people in their individual circumstances.

As things have become more complicated, primarily owing to the action space becoming crowded and competition increasing, the form of this centralising herding behaviour has been forced to move into new realms. On a note of overcrowding and competition, it is striking to hear in the English text Winner and Waster how it was an issue then, in that time. The author bemoaned that standards of hospitality toward strangers had lessened - the best foods and choice cuts were no longer given to guests to show off.

How is Centralisation Presenting Itself in the Information Age?

In search for revenue, there is a push to commodify everything, and this has married with a user need to outsource risk and management of complicated systems. Companies do not want the overhead of having a system administrator on staff or running a server room, and so cloud services and remote administration become highly appealing.

So-called progress itself is driving centralisation by fuelling ever more complicated systems that require management with overheads beyond the capabilities of non-specialists. There are several concepts that articulate aspects of this process:

• Jevons Paradox: In 1865, William Stanley Jevons first described a paradox ... He maintained that more efficient steam engines would not decrease the use of coal in British factories but would actually increase it.
• Path dependence: The phenomenon of past events or decisions constraining or defining later events or decisions.
• Induced innovation: The relative prices of the factors of production are themselves a spur to invention ... to economising the use of a factor which has become relatively expensive.

Many thought technology would have met a bottleneck and stagnated long ago, but it still seems to push forward ... Although there are signs that some bottlenecks have been reached, and are just not spoken of. An example is ISPs lying about internet speeds and bandwidth - multiplexing and throttling traffic to deal with periods of high demand. Put plainly, as long as the infrastructure can scale, and the market is able to grow, centralisation will continue. Even as the risks of centralisation present themselves - such as with Cloudflare disruptions taking down large sections of the internet - the risk and domain knowledge outsourcing offered to individuals and companies still prove worthwhile. Admittedly, some of that could be artificially imposed via government diktats.

Within the realm of software development, centralisation is pivotal. Developers want to acutely control what the end user is able to do, often hiding or masking functionalities, as correct behaviour can yield poor user experience, which then translates to programming time. Further still, developers will often deploy background telemetry so bug reporting is wholly controlled by the developer. By centralising this control, the developer is reducing the risk of their time being wasted. The use of Large Language Models as an automation tool has further fuelled centralisation in programming, as the development team can simply leverage automation and autocomplete tools to increase their output and complete tasks that otherwise would have been delegated to others.

Contention

Where this accelerating centralisation meets resistance is when a local decision-maker has more local domain knowledge than the centralised body, and an identified correct course conflicts with the will of the centralised body. This is a signal that decentralisation should take place. Unfortunately, a signal can be ignored, just as it can be acted upon. The status quo will only change when there is a forced hand demanding a change.

When you study computer science, you are taught that decentralisation is a scaling solution. In the realm of decentralised systems, take blockchain, for example, there is acute awareness of the network effect. Could that great filter that the network effect represents be the same thing as the need for centralised systems to decentralise, with the difference being that decentralised projects audaciously seek to sidestep the centralisation stage altogether?

In the context of there being contention with ever-encroaching centralising forces, it is possible that decentralised projects have a different purpose to what the language implies. That being, it is an attempt to function beyond the encroaching centralising forces. Rather than being decentralisation, it is a defensive attempt to organise, or, in other words; to centralise.

Single Points of Failure.

The obvious downside to centralisation is the creation of single points of failure. There are many an argument saying that this is a major risk that needs to be avoided. The real world, however, says otherwise, as there are multiple single points of failure in systems, and through sheer will and momentum, things are forced to keep working. And all the sweat and pain that keeps it working goes unsung. In the real world where results are paramount, the risks coupled with single points of failure simply have to be accepted. A small company may have only one person on staff who knows how to do a key operation - recruiting a replacement would be a major challenge, but training up someone else in the company is not viable due to the lack of overlap of employee expertise, and so this risk just has to be accepted - work goes on.

A small company with delegated responsibilities is highly self-organising. In a small workspace - everyone knows what domain is whose responsibility, and everyone knows what is and is not being done. When operations get bigger, self-regulation and self-delegation of tasks is less pronounced. Centralisation and single points of failure are not so apparent. Consider a bunch of small business garages in a town, but unbeknown to residents, they are all owned by private capital, and so if that one capital firm gets in trouble, all those garages go under all at once.

Mitigation

Management being globally centralised, and company operations being operated in a decentralised manner, is abhorrent when looked at in statistics, but it is the situation that came to be. Revenue streams are materially valued more than financial assets on the balance sheet, yet it is asserted that the valuations each are true. Money power manipulation games. That one still-standing small family-owned garage in the area looks run down compared to the centrally private capital-owned garages that have new fascias and paintwork on their company cars. The revenue is likely similar between the two, but while one has ready access to capital to do renewals, the other does not. Obviously, some fraud is taking place.

Such a fraud can only be dealt with by those that have a monopoly on violence, wielding it against the perpetrators of the fraud. That is, of course, assuming that the perpetrators of the fraud and those that wield a monopoly on violence do not ultimately have the same boss. A more tractable problem is where a single node in a small group of interdependent nodes, being tasked with a central responsibility, becomes a single point of failure and a liability to the network. To mitigate this, there are devised policies of monitoring, redundancy, or decentralisation. Calling back to the real-world example of a small company where the risk has to just be accepted, the only practical policy left there is one of monitoring. That is what is seen to take place in the hu-man domain, be it informally or formally. People chat by the coffee maker to check if colleagues are happy, or there might be a formal end-of-year review, or tests to meet industry standards. A compromised member of a group needs to be identified, and removed or rehabilitated. This, in practice, means constant monitoring, and or, conditioning of behaviour. An attack vector to undermine this self-regulation is to ensure that all the members of the group are too busy on their own tasks to do the informal decentralised monitoring of the group.

In modern society, we have pointless things like filling out voter registration, and regular scheduled shock events that fill news headlines, which act as behavioural conditioning. Outliers can be identified by the society mass and brought back in line, rejected outright, or closely monitored if needed. This is exactly how a system administrator looks after a network, or a security guard looks after private property. If that monitoring then itself becomes centralised, then the obvious route of attack is to go after the system node that bears the responsibility of doing the monitoring. It is thus important that the monitoring be kept decentralised, incentivised to be honest, or that policies are put in place to monitor the monitors.

A policy of monitoring the monitors is what seems to have come to be in the West; policies, standards, conditional grants, and rebates all act to constantly keep a watch on people and companies, concentrating centralisation and reducing the local autonomy of people. Again, this may well not be the state that is wanted, but it is what came to be. People are all too willing to forfeit responsibility to fuel greater centralisation, as responsibility is fundamentally a burden, as it represents work to be done.

Again, we have a point of contention - most will gladly give up responsibility, since it is a burden - but there will be a proportion who will insist that the burden that responsibility represents must be borne by as many people as possible. That proportion are, in effect, held captive by the passive will of the masses. It is from this group of people that we hear talk of a need for decentralisation. Framed in the context just outlined, it is not decentralisation that they are wanting, but a want to centralise their dispersed and overshadowed will.

Conclusion

Centralisation happens because it benefits people, and it is fuelled by technological development. In its truest sense, de-centralisation is a scaling solution, that, as the name suggests, comes after centralisation. People that talk of the need for de-centralised solutions could, in fact, be speaking to a want to centralise their will to escape the tyranny of the mass permitted centralised systems built around them.

To mitigate single points of failure in systems that employ centralisation, policies of monitoring, isolating, repairing, or replacing compromised points of failure must be in place. In the hu-man domain, not suited to rigid policy, consensus-building and self-regulation through social ritual can be employed.