Something Different This Way Comes

In which I attempt to renegotiate rationalism as a political philosophy, and offer my alternative—A king is not a micromanager—Decentralisation can save the world

Introduction

In this part, I lay out what I see to be the political implications of rationalism, and try to offer an alternative worldview for solving high level cooperation and coordination problems.

The Ugly Side of Things

So far, I have not laid my cards on the table as to why I am writing this essay. It seems clear to me that the four core beliefs I laid out in “Defining Rationalism” place rationalists and their ideological neighbours at a high risk of ideological radicalisation to conservative and far-right causes. This is generally done by extrapolating the third principle to the idea that the smart and agentic should direct the affairs of the hoi polloi due to their inability to self-govern, along with adopting a dim view of human cooperation that leads to the adoption of “might makes right” strategies. It also leads to a libertarian impulse which resists social constraints and sees social conventions as entirely retrograde and not worthy of respecting, eroding the social ties of rationalist participants in the process. There is also a large element of potential overlap with scientific racism (so-called “Human Bio-Diversity”): Once it can be “determined factually” that one sub-group of humans is more intelligent than another, the intellectual supremacism that has already been established justifies depriving the less intellectual of equal standing in society. It is a lie to say that rationalism is compatible with every value set, because not every value set values “winning”, and the supremacy that entails ¹.

Of course, the rationalists would say that, however repugnant such conclusions might be to conventional morality, they are the best way of analysing the world we have before us. However, what if there was an alternative?

There is another way

It seems clear to me that, if we aim to dethrone an ideology constructed on the supposed rigorous study of reality, then we must make sure that any alternative world-model we propose have at least as much explanatory power as the original. It should also enable instrumental and ethical outcomes superior to rationalism, otherwise rationalism would still remain the best way of achieving your goals even if its world-model is incorrect. We must win more and win better, so to speak. In that light, I tried to generalise some of the ideas behind conscious reinvestment to broader systemic principles. Here’s my best shot:

And hence the fourth principle - By thinking about this problem, you take on the responsibility to explain this problem to others and get as many brains online to tackle this problem as possible. Loosely speaking, one million minds are better than one hundred, no matter how smart the one hundred are.

The Perfect Case of Governance is No Governance

In the previous section I have brought up the idea of decentralisation. It is necessary for me now to define it more rigorously.

The ideal case of governance is often presented as a supremely wise philosopher-king, one who makes all decisions justly and whose word is law, turning those perfect decisions into perfect actions ². But, if we accept the constraint that such a perfect being could exist and be identified at all, surely the best case scenario is that all citizens of a kingdom are such perfect beings? In such a world, the kingdom governs itself without error, every citizen instantly knowing the best solution to cooperate to solve any issue that arises. So we establish that the perfect case of governance is that no governance is necessary. How does it apply to the real world?

Fucking around and finding out

As an example of a useful heuristic I have learned through painful experience, I will now present my account of why erroneous theories and systems persist for so long, and cause so much harm when they fail (e.g. the financial order behind the 2008 financial crash). The heuristic is this: the longer you fuck around, the worse it is for you when you find out. To be slightly more specific, the longer it takes for a theory or proposition to be tested, the worse the consequences are if it is invalid.

Imagine wearing a backpack with only one strap. This is less secure than wearing your backpack with two straps, and if you trip the backpack may dramatically swing out and throw you off balance, but in general this is not a problem. If you trip early on while wearing a backpack like this and catch yourself, you’ll probably put the backpack on properly and continue on your way. However, if you don’t trip early on, you might fool yourself into thinking that this is a stable situation. Then you pick up a drink with one hand… a box of chicken nuggets in the other… maybe a shoelace eventually gets undone… When the tripping hazard eventually comes it’ll be a lot more awkward than a momentary wobble.

To be clear, this is a limited heuristic: your theory that you can outrun a car will be swiftly tested and the consequences will still be extremely dire from your perspective despite less than 3 seconds transpiring between theory and test. However, when constructing systems, I propose that the longer a system has to wait before it comes under stress the worse the results will be. In general, systems grow labyrinthine with no pressure to stay resilient and errors compound silently in non-stressed states. A casual example of this is the global network of Just-In-Time logistics, founded on the belief that ships will always arrive on time and that stockpiles need only carry exactly as much stock as is needed to manufacture a product because losses during production are predictable and minimal ³. Constructed in a time before the consequences of climate change were evident, the system has had some 70 years to accrue errors of judgement and critical dependencies ⁴. The worsening weather will finally show us how bad those errors of judgement are.

This heuristic is a quasi-adaption of Taleb’s concept of antifragility, where small shocks to a system actually strengthen it if it can learn from those shocks. Here, I position that systems that can rapidly iterate over feedback (i.e. learn from mistakes without too much time lag for their errors to become established doctrine) are more efficient than those systems that will not or cannot. In practice, while some degree of organisation is needed to coordinate effectively, having local units free to react to situations as they arise allows them to iterate more rapidly and get a better grip on the situation. We contrast this against the failures of top-down systems of control such as those in the Soviet Union, or the failure of the Federal Reserve to prevent economic crises in America. ⁵ If we accept the general proposition that those closer to an event (physically or socially) have more information about it, the idea of decentralisation becomes a lossy approximation of the kingdom of perfect philosopher-kings. Each unit is still making decisions quasi-independently, but under the aegis of a larger whole which manages coordination and larger issues ⁶.

Levels of Organisation

At this point, much has been offered in terms of high level abstraction. How do these supposed principles play out in practice? To even partially answer that question I will have to do something that I have in the last part accused rationalists of doing: turning complex real world phenomenae into abstract systems. Specifically, I would like to compare decision making systems to computer networksr or signal transmission systems.

Lest I be accused of hypocrisy, I will now point out the constraints such a high level model places on the environments I am seeking to model. I am generally modelling a system with no overtly malicious actors or saboteurs, or at least one where they are sufficiently rare as to be anomalies. The components in a signal processing system have fixed efficiencies and fixed utility functions, things real people rarely possess as constants for long amounts of time. In general, I assume that people are generally decent, working to something like their overall capabilities, and largely working cooperatively. An example case where this analogy fails apart: common strategy in signal transmission networks is to throw in a repeater every now and then to strengthen the signal and suppress noise. When people do it, it’s called Chinese Whispers, and the signal gets way, way worse.

What does this framework offer? Once we see individual bureaucrats or social officers as components in a greater information network, some features of organisation immediately become clear. Many video games and media that feature kings portray them as effectively nothing more than national micromanagers, resolving their petitioner’s claims one by one and setting standards that apply to the whole of their domain. In practice, such a system would be almost impossible without computational capacities far beyond that of a human or group of humans: total top down control would mean that a kingdom becomes a star network, every terminal requesting information and depositing information to a central server, quickly overloading it at scales of more than even fifty users. Similarly, the slowdowns brought about bureaucracy become easy to model: the speed of information transmission in a chain of signal repeaters is bottlenecked by the efficiency of the slowest repeater. Systems where multiple parties must sign off on the same task or approve the same form also become upper-bounded by the speed of the slowest worker.

A birds-eye view of societies as information systems also allows us to appreciate the statistical function of the state, what Foucault refers to as biopower ⁷. This is best illustrated, as Foucault does, with the example of “law and order”: Consider the mundane securitarian, the beat cop. For him, crime and order are individual questions, discrete assignments he can succeed or fail at. A thief let loose is a single, real, entity who shall commit tangible harms, a person with whom it is possible to have a relationship of admiration, anger, or hatred. Insofar as he concerns himself with the community’s continued security it is in real and conceivable terms of knowable persons who live in that locale. So-and-so has been robbed, so-and-so murdered.

Now consider the Director of Police, or the Minister for Security. His concerns stretch out both in terms of space and in time, he concerns himself not with neighbourhoods but national districts, not with moment-to-moment assignments but broad trends of crime and punishment. The success metric here is probabilistic, that is to say non-discrete. It does not matter if a single cop catches a single thief, what matters is the percentage of cops who catch a percentage of thieves. Returning to the metaphor of social systems as information systems, mesh networks such as the internet demonstrate a similar property in that the general success of a message-passing directive depends not on the individual performance of any node, but rather on the overall performance of sufficient nodes in the mesh network.

But it would be a mistake to conclude that only Ministers of Security view these problems in such a way. Compared to the individual doctor, the Minister for Health in the neighbouring building takes a similar approach to his securitarian counterpart: while for a local doctor it matters personally if their patient lives or dies, the Minister for Health deals in broad statistical categories. They do not consider Joanna the overworked single mother with stage III lung cancer, they consider the group of all single mothers in long term care. Insofar as these directors and ministers and presidents concern themselves with individual cases, it is because they have become prominent enough to affect the psychological landscape and therefore the broader trends which form the system—or that they have become obsessed with some model problem which is often not in fact the broader phenomenon it is a cypher for. The paradigms are incomparable, more has shown itself again to be different.

And it is not even that these two perspectives are in conflict: if the local doctors no longer cared about individual patients, the overall trends of national healthcare would suffer, and if the overall trends were not considered local doctors would not receive adequate centralised resources with which to care for individual patients. If the beat cop started reasoning in overall trends and no longer caring about individual assignments, he would fail in his securitarian task and harm the broader securitarian project. This micro-macro syncopation explains the so-called Voter’s Paradox: while in the broadest sense voting trends are based on population level distributions, these distributions are constituted through the local-level belief that individual decisions are important and matter; and the broader scale implications of these distributed local decisions gives each local decision weight and importance. One cannot exist without the other ⁸.

Society as a Black Box

A further benefit of considering society as an information system with defined processes and utilities (each with their preset inputs and outputs) is that society itself can now be considered as a totality, a system with its own defined set of inputs and outputs. In aggregate, then, societies take in resources of various kinds (labour, capital, natural) and produces services for its members (food, water, shelter, electricity). We can describe the efficiency of a society as the disparity between, for example, the energy it takes in from livestock, agriculture and transport; and compare that to the amount of food it produces with that energy. However, we must be careful to separate overproduction and waste from actual useful production: in other words, we measure calories taken in and compare that to calories produced and consumed by a society’s members, rather than also accounting for, say, wasted food that goes straight to a landfill. Similarly, we can aim for reducing duplicate labour, wasted time etc.

If alarm bells aren’t ringing already, they should be. The crude version of this optimisation process led to the so-called optimised supply chains of Just-in-Time logistics, which we have discussed already. The optimal amount of fraud is not zero, and the optimal amount of slippage/redundancy is definitely not zero in a world with freak storms and heat waves in February. Furthermore, we can begin to quantify statements like “capitalism is not fit for purpose” by examining what outputs the existing information networks of society accelerate and incentivise. A full version of this analysis is beyond the scope of this essay, but suffice it to say that we perhaps owe more to the legacy pre-industrial contexts that we have inherited than we do to capitalism the bread on our table today ⁹. And, of course, once one obtains this high level view of society’s functions, one further understands the impossibility of directing a million, much less eight billion, people.

Other Pragmatic Arguments for Decentralisation

Beyond such high level, conceptual arguments, more pragmatic arguments can be made for the “chunking” or decentralisation of society. Consider the following chain of argument (and the contextual presumptions it takes as axiomatic!):

This suggests that, rather than a centralised “beamtersstadt” relying on anonymous bureaucracy to manage millions, a society consisting of self-organised social units that work not by abstract principles of ideology but rather by close-knit community bonds would feature less fracturing and conflict. Taleb addresses this approach in his description of the cantonal organisation of Switzerland, which he refers to as the most successful country in the world. He attributes this success to their relatively weak central government and strong local social participation. The removal of cold bureaucratic abstraction would also reduce the likelihood of harmful decisions being inflicted on people by anonymous others with no understanding of the impact of their actions.

Dignity without Supremacy

If we take into account the concept of decentralisation, then we begin to dissolve the intellectual hierarchy/meritocracy framework that underlies rationalism’s affinity with the right. After all, the principle of meritocracy is that a society needs to elect or promote those most qualified (by some metric ordained from above) to operate it in a centralised fashion. By contrast, a decentralised society might have many such centres of operation at a mesoscale, with intervention between clusters or groups occurring only in times of shared need. The goal becomes not to subjugate, control, or otherwise manage those less intelligent than you, but to empower them to contribute to the collective thinking-effort of society at whatever level suits them best. Once the oppositional framing is removed, cooperation becomes much easier, even straightforwardly rational.

There are those who will say that these words are only possible in our modern, degenerate, toothless age, where all conflict is mediated and the inner brutality of man is hemmed in by civilised watchmen on all sides. There are those who would further posit that, should this veneer of civility be stripped back, the strong (by which they always mean themselves) will inevitably emerge and reshape the world in their image. However, I would venture to argue the opposite: that it is this anonymised, alienated society that allows cruelty to flourish. In a vast and impersonal system the worst that can happen (socially, at least) is that you are required to relocated to another anonymous corner to continue your cruel acts. Society creates systems of faceless workers to enable monsters to access food, water, and fresh victims by accruing power and money, establishing barriers of plausible deniability between sociopath CEOs and those they wound and injure. Whereas, in the counterfactual world where the community is small and survival difficult, a tyrant’s grip is (though harder perhaps to disturb initially) much more of a brittle thing, with no recourse for the Roman king if his subjects revolt. Writing in just such a time of tumult, Macchiavelli understood this principle well:

Therefore take care not to assume that what is ruthless and bloody is natural, nor that what is soft and gentle-seeming is a sign of weakness.

Recall again that rationalism is path-dependent, that games where you can win are poor metaphors for life, that perfect prediction and control are near-impossible. In this schema, I argue, any top-down plan for world optimisation is misguided—doomed to failure by the sheer scope required. I understand that the alternative I present can be attacked as its own particular form of baseless theory. But it did not seem right to criticise without offering my own alternative, so I hope it is of some interest, at least, to think about.

For more analysis on rationalism’s ties to scientific racism and eugenics, see Timnit Gebru’s TESCREAL analysis: https://firstmonday.org/ojs/index.php/fm/article/view/13636/11599 ↩
It is, of course, a particular kind of blindness that assumes that, if such a man were real, he’d be able to administer perfectly the affairs of a thousand people, much less several hundred million. The kings of old ruled by delegation and the modern lords of finance rule by decentralisation not out of benevolence but by necessity ↩
An article with more detail for reference: https://www.theguardian.com/commentisfree/2021/oct/11/just-in-time-supply-chains-logistical-capitalism ↩
Like, for example, being disruptable by both global pandemics and a single ship ending up in the wrong part of a canal ↩
Of further interest, naturally, is James C. Scott’s seminal Seeing like a State, which goes into much more detail about such topics. ↩
For those who will protest that this is some delusional form of leftist optimism, a similar argument is also put forth by Hayek as to why decentralised markets are more efficient than central state control. ↩
Cf. Section 1 in Security, Territory, Population by Michel Foucault: https://en.wikipedia.org/wiki/Security,_Territory,_Population ↩
Another excellent account of this stochastic or statistical approach to crime can be found in the field of financial Know Your Customer/Anti Money Laundering regulations, as described by Patrick McKenzie here: https://www.bitsaboutmoney.com/archive/kyc-and-aml-beyond-the-acronyms/ ↩
As a taster of arguments in this direction, in a perfect void rational agents aiming to maximise earnings would prefer to work in finance rather than agriculture. Think about that for a second. ↩

Something Different This Way Comes - Part 2