Friday, January 8, 2010

Of beating hearts part 2

CONTENT WARNING: This is an embarassingly long post, with a lot of long-winded analysis, caveats, pronouncements and occasional gems of insight. A nifty bulleted list is provided at the end, excerpting the main theses and arguments for the time- or attention-challenged.

-------------------------------------------------------------------

In "Of beating hearts" I expressed the opinion that in general, redundant, non-hierarchical or polycentric networks exhibit superior reliability and stability, compared to monopolistic, hierarchical ones, in most domains of technology, biology and social life, despite the polycentric systems having frequently an appearance of disorder and chaos. Yet, there is one particular form of network services that very persistently, in almost all societies, takes the hierarchical, monopolistic form - and that is the provision of law, order and defense (LOD).

This fact may be evidence that for some reason LOD services in principle cannot be provided by multiple, independent, competing providers. Indeed, without doubt there are particular societies and groups of people who cannot sustain competitive LOD networks, just as there are societies unable to develop e.g. the enterprise of natural science. The qualities of available network components are of course crucial for determining the range of network architectures that can be built. You can't have a Philosophical Society in a group of illiterate, IQ 85 people with high time preference, and you can't make a silk purse out of a sow's ear. These, and many other issues are good reasons why historically the vast majority of societies went the LOD monopoly way - but this shouldn't stop us from looking at the polycentric LOD issue, just as our lack of wings didn't stop us from eventually learning how to fly.

In fact, there are at least two historical examples of societies with relatively competitive LOD networking - medieval Iceland and the Wild West, described in David Friedman's "The Machinery of Freedom". Their development may have been a historical accident but their functioning is proof that competitive LOD is possible even in a society consisting of unmodified humans of average intelligence.

To state more clearly why I am writing this post, I am considering competitive LOD for two reasons: firstly, out of concern for long-term reliability in the provision of LOD services, and secondly and much more importantly, because the LOD regime tends to influence all aspects of social life. A corrupt LOD provider inevitably corrupts its charges, teaching fear, hatred, and violence, degrading society's ability to deal with physical challenges, the ability to discover truths about the physical world, to innovate and to survive. If it were possible to build a better LOD system, it would have far-reaching, cascading effects throughout the whole social fabric, allowing unprecedented improvements in the quality and probably duration of our lives.

Let me explain first what I mean by a polycentric LOD system through contrasting it with the modern state. Modern states combine judiciary and military components with a wide array of other aspects, from the regulation of zoos to the redistribution of income from workers to whoever happens to be at the levers of power. A constant feature is the unitary structure of the state: all persons in its area of influence are subject to direct control by organs of the state, with no recourse except by appealing to other organs of the state. You are truly surrounded by the state, and the cost of switching to another is large. In a polycentric system there would be no single organization capable of exerting direct control over more than a rather small fraction of the populace. A police organization could not wantonly attack any persons who happened to be in its reach, a single school system would not teach all students, a lawgiver could not proscribe the use of drugs except by persons voluntarily subscribing to his law. The polycentric LOD providers would be truly independent of each other, sovereign with respect to other providers in a manner reminiscent of the relation between sovereign states, yet spatially intertwined. These two features, sovereign independence and spatial intertwining would provide the conditions necessary (but not sufficient) to allow one more crucial element to be realized: individual, low-cost exit option. Should you be dissatisfied with your current laws and their enforcers, you could switch to an alternative with a single credit card authorization. This individual choice option is the crux of the matter as far as fullfilment of individual wishes is concerned - being able to choose the law to obey would maximize social utility (efficiency) of the system, since individual citizens are collectively in possession of a much better knowledge of their own preferences than possible for a group of even well-meaning and brilliant rulers. And, of course, individuals have on average a much better set of incentives to act in their own interest than any third parties. Increased efficiency in fulfilling wishes of users is to be expected in the polycentric system just as it always happens when a monopoly (say, AT&T, or the state grocery in East Germany) is replaced by a competitive ecology of providers.

Before I continue, let me dispense with one of the common objections to the possibility of building a polycentric LOD system - namely, free-loading (i.e. defection). It is important to realize that protective services are not non-excludable, therefore the system can easily be defended against defectors. In the polycentric system, anybody could choose not to have a protective contract with a LOD provider - but this would not mean an easy life under other people's umbrella. There would be no law protecting such persons from predators. Should you choose to cancel your protection policy, you could find yourselves robbed by random thugs, chopped up for spare organs, and nobody would come to your aid, except for a much higher price. While this might sound brutal, it isn't - just as getting yourself killed by your own choice, e.g. in an extreme skiing accident, is stupid, not brutal.

Now back to justification for considering the polycentric option: I emphasize the efficiency of the polycentric system because there are many very smart thinkers who endorse the monopolistic state for efficiency, i.e. social coordination reasons. I believe they are factually wrong on this issue: The state is incapable of outperforming a polycentric LOD system, ceteris paribus, just as a central planner is structurally incapable of outperforming a market in apportioning of resources. The state relies on long feedback loops (elections, rebellions), does not utilize dispersed knowledge and does not align individual incentives effectively. The individual exit option in the polycentric system combines optimal utilization of dispersed knowledge about preferences (available to individuals through introspection) with optimal responsiveness to this knowledge through individual choice and the very short profit/loss feedback loop between them and the LOD providers. Furthermore, laws, just as any other complex ideas relevant the world, must be first discovered, and then validated in reality - the multitude of polycentric LOD systems therefore acts as a method of experimentally finding the best possible solutions in the social realm, a type of truth-finding device not unlike the collective enterprise of scientific discovery in the physical realm. Even the brightest state bureaucrats do not know as much as their subjects do, even the most altruistic bureaucrats don't care that much about the man in the street, even the best calibrated monopolistic lawgiver cannot reliably find the best solution - for greatest efficiency we need individuals who know about what they want, care about their desires and are free to experiment with their lives. To put is succinctly, the state exists not because it is better but because it is easier to build.

The notion of efficient individual choice of law leads directly to the question of legitimacy and some normative considerations here. Almost any extant social system relies on the notion of legitimacy, the idea that some aspects of organization are right and proper, independently of mere brute force acting on their behalf. But where does legitimacy come from? Olden rulers claimed a divine source. Our current batch claim to be empowered by the will of the voting people, however nebulously translated into the Federal Register. Some libertarian dreamers say they are inspired by "natural law" but hot debates immediately ensue as to its exact precepts. The legitimacy of the polycentric LOD would in my approach rest primarily on the claim of superior efficiency in attaining the goals espoused by its participants. Just as you can estimate the efficiency of a particular phone network at reliably shuttling voice between users, so it should be possible to estimate the efficiency of the polycentric LOD system at fullfilling the wishes of its customers, and I am confident that a quantitative comparison with monopolistic systems would at least for certain classes of customers strongly favor the polycentric system, for reasons partially enumerated in the preceding paragraph.

Furthermore, any LOD system is not only shaped by its customers but also influences them in a reciprocal fashion. A tyranny breeds the weak, the cruel and the corrupt. A capitalist system rewards boldness, industriousness and humility (but not greed, contrary to what many shallow minds assert). A polycentric system would strongly favor openness, honesty, conscientiousness, tolerance but not indolent indifference to dangers. The Homo Anarchicus (subsp. Critarchicus) is likely to blossom under the influence of benevolent laws generated by his provider, just as Homo Sovieticus was morally and socially stunted by the laws generated by the local monopolist. Instead of having their minds polluted by collectivist propaganda dispensed by state-run schools, a greater number of children could be exposed to a variety of systems of thought, thus helping with the development of critical thinking faculties, not to mention their skill in argumentative blogging.

Therefore, a polycentric LOD system would be legitimized both by its immediate short-term improvements in efficiency and by its benevolent long-term effect on the moral fiber of the citizenry. I understand that persons emotionally beholden to the monopolistic state might find this long-term feature abhorrent, a perversion of their basic moral ideas. I would counter their objection as follows: The polycentric LOD system is legitimized by its efficiency but then as a second-order effect it gives rise to a novel and for many persons quite counter-intuitive reformulation of basic moral concepts, such as violence, right, and crime. Learning to accept this new definition of violence is a form of moral progress on par with and going beyond the benevolence espoused by the New Testament, or universalist morality advanced by secular humanism.

The definition of what constitutes violence is very important in the makeup of any society. Understanding violence is at the very bedrock of our morality, and describing an action as violent opens the way for the legitimate (i.e. accepted as right and proper, that which should not be prevented) use of force, or any other means necessary to stop or punish the action. Infringing a right allows the use of force against the perpetrator. A crime is an infringement on rights that are defended by laws. The precise interpretation of actions as violent or criminal emerges in a long process of recursive interactions between many actors - lawmakers, enforcers, infringers, victims. The outcome is shaped by moral beliefs of the participants, and by objective regularities in their interactions (analyzed by e.g. game theory), as well as various local features of the physical world. Beliefs of actors shape each other and develop in sometimes non-intuitive ways, leading to ever-changing notions of right and wrong. Now, is there a way of determining which of the myriad socially constructed definitions of violence is in some way "best"?

If you are a highly rational person like me, you won't be able to escape the notion of efficiency as the basis for translating moral intuitions and concepts into practical injunctions. To want or wish something (a state of the world) means to act within your capabilities to bring that state of the world into existence. Efficiency is choosing actions so as to realize the maximum number of (weighted) wishes that can be attained given the actions available to you. A rational person is by definition always acting efficiently, no matter what particular wishes drive his behavior. If desires are the matter of morality, then efficiency is the matter of meta-morality. Applied to the social realm, the criterion of efficiency is meta-judicial: The desires and beliefs of citizens must be translated efficiently into action, thus various independent legal approaches to this translation can be compared and evaluated by their levels of efficiency. Rational actors must then choose the most efficient possible approach (or else they are not rational). For example, if capital punishment is efficient in achieving whatever goals are to be achieved (deterrence, revenge), then it should be a part of the system. Otherwise it should not be used. The decision to use this action or not is thus driven by an analysis of desires among participants in the system (network) and the degree of efficiency of the action. I would claim that the most efficient system provides the best definition of violence, contingent on current beliefs, and superior to definitions produces by less efficient systems. Continued operation of the efficient system is likely to recursively modify beliefs, leading to more advanced definitions of violence, not attainable in less efficient systems.

I noted that a polycentric LOD system, like any market system, should be very efficient at aggregating dispersed information about desires. At the same time, through the pricing mechanism this system effectively communicates information about the efficiency of actions, about the trade-offs and costs inherent in the action. Insisting on punishing minor infractions (e.g. stealing a loaf of bread) by death could be very costly for subscribers to a private LOD organization, since their provider would frequently find itself in to-the-death battles with other providers. Still, there might be persons willling to pay high prices to save their loaves, and the polycentric LOD system would not a priori exclude any specific solutions to moral problems from consideration (aside from solutions that directly produce a monopolistic LOD system). Dissemination of information about the full prices of actions would most likely have an impact on the moral beliefs underlying such actions, in a recursively self-modifying fashion. If Americans had to pay the full price for killing hundreds of thousands more-or-less random brown people in far away lands, there might be fewer voices favoring sending troops out for no clear reason. Closer to home, the polycentric LOD system would most likely put a high price on envy. It would be difficult for the inept and the unscrupulous to extort resources from the hard working and the successful, since instead of simply voting for higher taxes and "benefits", the parasitic classes would actually have to go to war with LOD providers protecting workers from exploitation. Given enough time the polycentric LOD system would most likely reshape many cherished beliefs and desires, acting as a technique of moral progress, damping the thirst for violence and easy disregard for the wishes of others that characterizes monopolistic LOD systems.

You may protest that a modern state is not really that violent. After all, hardly anybody gets killed by the police in the US, right? Compared to the state of nature, when 20 to 50 % of all males died by homicide, our current levels of violent death are minuscule, even including fatalities among foreigners. I would would however contend that this is not a good measure of the levels of violence in our society. There is a subjective component to the notion of violence - the feeling of being surrounded, threatened with suffering, of having no recourse, of being violated. The polycentric LOD system, like any market system, is very efficient at measuring such feelings and translating them into price signals. Those who feel offended by an action are willing to pay to prevent it in proportion to the gravity of offense, and as long as it is easy to choose various sets of laws (outlawing different actions as being offensive to the users of the law), their willingness to pay will be translated into the existence of LOD providers protecting their patrons from the action in question. Currently however, living in a monopolistic system, I am extremely offended and feel violated by the notion that a municipal tax collector can demand access to my house to "assess" its "value", and then demand various arbitrary payments from based on his assessment. I would be willing to pay a lot to a provider to protect me from such indignities but, unfortunately, all providers in my area are members of a monopoly organization which includes the tax collector himself, making it impossible for me to express my preferences. Thus, looking merely at the statistics of slaughter does not fully describe the level of violence in a society. Since the monopolist makes a lot of people feel violated, the social level of violence is actually quite high, even if most of the time the body count is low.

This situation persists because state-inflicted violence is extremely cheap – once a coercive monopoly exists, it is as cheap as paying a uniformed thug 15$ an hour to enforce whatever law the state machine came up with. On the other side there is an enormous harm inflicted on the victims of the law, the harm of being exposed to and broken by violence. Therefore, whoever manages to get a hold of the levers of state power will impose his notions of what constitutes violence or harm, while disregarding the desires of others, and cheaply inflicting almost arbitrary levels of violent harm on his victims. In fact, the state is *built* to prevent any form of measuring the relative moral weights of violent harms inflicted on a law’s victims vs. the “harms” allegedly prevented by the law. It is the antithesis of moral efficiency – because of the massive imbalance in power between the state and its victims there is no bargaining process or any other effective way of measuring the relative weights of harms involved (short of a bloody uprising), and therefore using the state to achieve one’s goals is extremely likely to increase the overall harms inflicted on members of the ingroup.

To summarize the foregoing analysis, a polycentric LOD system would act as a moral research engine driven by considerations of efficiency, reshaping our notions of violence, rights and crime through a process of recursive modifications of laws and beliefs. The polycentric LOD system could make us better humans.

The interesting question then is what kind of conditions would have to be fulfilled to purposefully create a stable competitive LOD regime, and whether there is a feasible pathway leading from some of today's systems of mutual oppression into the shining future of untrammeled freedom. A few very bold thinkers tried to answer it: David Friedman, Mencius Moldbug, Patri Friedman, and Robert Nozick come to mind. Fans of science-fiction might be familiar with Vernor Vinge, Karl Schroeder, and John C. Wright in this context.

I will not summarize the ideas set forth by my illustrious predecessors in this field of armchair social philosophy. Invincible, cryptographically controlled robot armies would be a fine addition to our political armamentarium but, certainly, a stable polycentric LOD system would require other ingredients as well.

The most important indispensable ingredient that is currently missing is probably faith: a belief, shared by a sufficient number of actors, that the polycentric LOD system is a good of great importance, one that must be defended against all kinds of influences. Our Holy Mother Church depends on parishioners' belief in her divine origin for paying her electric bills. Democracy depends on the belief that participating in the spectacle of elections is right and proper. A polycentric LOD system is no different in this respect: There must be a critical mass of polycentrists, both in absolute numbers and as a fraction of the society at large, to allow for the system to come into existence and to maintain itself. The first problem on the path to polycentrism is then convincing a sufficient number of people of the validity of the concept.

The ability to appreciate the benefits of polycentrism depends on sufficient intelligence, exposure to a large body of historical information, and having the right set of emotional attitudes. In my experience, to follow the argumentation in favor of polycentrism one needs to have an IQ in excess of 120 - I have been completely unsuccessful in communicating the issues of polycentrism to persons of average intelligence. The advent of the internet gives the inquisitive 120- and up person the ability to sift through a lot of sources of information and to develop the critical faculties needed to collect data relevant to polycentrism (including game theory, history, evolutionary psychology, economics). It is now also much easier to find other persons of similar persuasion and to accelerate one's development by interacting with them. Thus the number of potential polycentrists is probably growing, although I doubt that there will be enough of us to effect any positive action before the singularity wipes the social and political slate clean. It may be possible for some small areas with polycentric LOD to emerge (a la charter cities) but of course their sovereignty would be dubious, as they would exist by the sufferance of other states, rather by virtue of their own defensive capabilities (about which I might write another post). On the other hand, the humans (if any) who survive the AI singularity will be most likely more intelligent than us, and will have also the ability to modify their own psychology to remove atavistic tendencies that made sense in the environment of evolutionary adaptiveness but are completely useless now, such as instinctive xenophobia, or the strong yearning to belong to a hierarchical, authoritarian structure that is exhibited by at least 50 - 70 % of modern humans. If polycentrism is possible in the post-Singularity computational substrate (no AI singleton is in charge of everything), there may be a good chance that the requisite numbers of post-humans will form themselves into polycentric networks.

Once a polycentric system is created it must be stabilized and defended against various internal and external influences. It is useful to look at extant sociopolitical belief systems and the mechanisms contributing to their stability. There is a number such mechanisms, operant in parallel and in various contexts. Altruistic punishment, banishment of non-believers, early indoctrination, control of reproductive access (directly or indirectly, e.g. by controlling financial resources), modern security organizations (i.e. the Gestapo and its congeners), and propaganda are capable of perpetuating the most odious, destructive and useless social organizations, seemingly limited only by their liability to be physically supplanted by less inefficient ones. Certainly, some techniques are adequate to the task of maintaining a polycentric system as well.  Of course, using e.g. concentration camps to enforce polycentrism would be a strange contradiction so not all of the commonly used methods would be appropriate or possible here. Probably the best stabilizing technique would be a social norm allowing any citizen, or group of citizens, to attack any LOD provider that exceeds some preset market share, and to divide its property (as well as the property of the provider's patrons) among themselves. This may sound strange - a call for brutal aggression against a successful corporation and thousands of its clients - but it is not any different from the aggression that is allowed in the direct defense of any political system. All persistent sovereign entities and social systems are willing and capable of inflicting extreme harm in protecting themselves from being supplanted, or else they are supplanted by other, more resilient systems. Once a tyrant blinks, his enemies pounce. Once the king, the politburo or the senate lose their mojo and no longer feel good about slaughtering their enemies, their demise is near. In this respect the polycentric system is not different from a state but the stabilizing action would be achieved by independent organizations of citizens rather than by a hierarchical institution. Another technique would be ostracism of individuals who do not subscribe to polycentric beliefs, or fail to ostracize such individuals. Both of these techniques are a common feature of many stable and successful societies. They are forms of individually initiated but collectively effective punishment of defectors, allowing the indefinite maintenance of both primary public goods, i.e. the beliefs about polycentrism, and the secondary public goods, i.e. the willingness to punish defectors.

Of course, relying on such ad hoc movements would make the system somewhat prone to instabilities caused by variations in commonly held beliefs and attitudes. In a tyranny you can have a stable, if brittle, structure consisting primarily of individuals opposed to its existence. In a polycentric system, which keeps aggregating information about preferences and translates it into action, as I extensively expounded upon previously, even small shifts in crucial beliefs could have a destabilizing effect. Since I expect that large polycentric systems will come into existence only after the AI singularity, I can wave the magic futurist wand here, and declare that the requisite stability of beliefs woud be achieved by a priori designing participants' minds to hold polycentric beliefs in a stable manner, subject to modifications only is some special circumstances - e.g. if a Bayesian analysis of available data strongly favors a different social system. Such stabilization would probably require programming techniques similar to what the SIAI would like to develop for building the Friendly AI but applied to (post)-humans rather than sentient artifacts. Post-Singularity knowledge of mind functioning would also result in a whole new level of psychological transparency and belief verifiability, making mutual, reciprocal enforcement of necessary beliefs less daunting than it is now. Importantly, in contrast to the current situation, only a very small number of meta-legal beliefs, directly related to the prevention of instability in the polycentric network, would need to be maintained in this way. The myriad of legal norms, desires and proclivities that form the bulk of what we want would still be free to evolve, enabling the moral progress machine to reach new levels of functioning.

OK, this is a monster of a post, almost an unqualified reservation. Let me finish by providing a nifty little bulleted list encapsulating my main theses:

- The polycentric LOD system consists of sovereign and independent but spatially intertwined providers of law and law enforcement

- It may be infeasible in many societies but there are clear historical examples of its emergence and relatively long-term stability

- Free-loading is not the prime obstacle to the functioning of a polycentric LOD system

- The main feature leading to increased efficiency of law generation in the polycentric system is the low-cost individual exit option

- The mechanism of increased efficiency further relies on aggregation of dispersed knowledge, a transparent pricing mechanism allowing rational trade-offs, shortening of feedback loops, and improved alignment of incentives

- The polycentric LOD system promotes high levels of legal innovation and discovery of novel, useful laws, leading to accelerated social progress

- An efficient legal system translates individual preferences into rules in such a way as to maximize achievement of participants' goals

- Normative claim: Laws of the polycentric LOD system are morally superior, because they are more efficiently translating human preferences into concrete rules than is possible using other sources of law

- The criterion of efficiency is meta-judicial

- High efficiency legitimizes the polycentric LOD system

- The polycentric LOD system acts a legal and moral discovery engine

- The polycentric system recursively modifies moral thinking of its participants, leading to moral progress and a blossoming of humanity

- The polycentric LOD system's ability to transform common notions of rights, violence, and crime could lead to a moral quantum leap on par with the greatest ethical revolutions of the past

- From the moral vantage point of persons accepting polycentric LOD, our current monopolistic system (i.e. the liberal democratic state) is marred by unacceptably high levels of violence

- High levels of violence delegitimize the state, independently of its claims to justification through general franchise

- The main ingredient necessary for the stable formation of a polycentric LOD system is the belief in its efficiency shared by a sufficient number of rational actors

- Additional features, such as norms for maintenance of secondary public goods through individual punishment of non-cooperators, are most likely needed to further stabilize the polycentric system

- Due to limitations of intelligence and rationality, it is unlikely that a large, sovereign polycentric LOD area would form in the near future, although smaller enclaves with limited sovereignty could form by conventional political mechanisms

- A polycentric organization may be a feature of post-Singularity societies, relying on augmented intelligence, software-stabilized beliefs about polycentrism and complete psychological transparency among AI or post-human sentients