Ashby's Law and the Dispute over Economic Planning

The dispute between proponents of central planning and the market economy is older than the implementation of the planned economy itself. Typically, it takes the form of an ideological battle between supporters of socialism and capitalism. Meanwhile, the subject of analysis—the economy—is primarily a complex system: dynamic, nonlinear, multi-element, susceptible to disturbances, oscillations, and delays. For this reason, it can and should be considered in the light of cybernetics—the science of regulating complex systems.

This article, inaugurating a series dedicated to the cybernetic analysis of the market-plan dispute, focuses on the question of which of these mechanisms better fulfills the role of a regulator in light of one of the fundamental laws of cybernetics—Ashby's Law of Requisite Variety. This law states that an effective regulator must have a variety at least equal to the variety of disturbances acting on the system.

A key distinction here is between regulation and adaptation. In the context of the market economy, these concepts are often equated: the market's ability to respond to stimuli and structural changes is often seen as evidence of its high regulatory variety. However, these reactions are post factum, lack a global goal, and are not based on a model of the system as a whole. This means that we are dealing with adaptation, not regulation in the cybernetic sense. Ashby's Law applies exclusively to the latter.

The variety of a regulator can be analyzed at several levels: the dimension of the state vector, the method of information aggregation, the existence of a system model, the degree of hierarchization, susceptibility to decisional entropy, the deliberate generation of positive feedback loops, and the ability to isolate disturbances.

Dimension of the State Vector

The first of these is the way the state of the system is described. The market economy describes the state of the system using a one-dimensional signal—price. Price serves as a universal hyper-aggregate, which is supposed to contain information about the availability of raw materials, the division of labor, the level of technology, the volume of production, the quality of goods, the relationships between sectors of the economy, and many other factors. At the same time, the economy is a multi-dimensional system where each of these aspects represents a separate degree of freedom, requiring a specific regulatory response.

Describing such a system using a single parameter leads to information collapse: many qualitatively different states become indistinguishable to the regulator. From the perspective of Ashby's Law, this means a drastic reduction in the effective variety of the regulator. Price does not so much simplify the description of the economy as it destroys the ability to distinguish disturbances of different structures.

In the case of central planning, the situation is different. The plan is not based on a one-dimensional signal but on a multi-dimensional description of the state of the economy, including both quantitative and qualitative parameters. Instead of reducing all processes to a single measure of value, the plan can operate simultaneously on quantities such as physical quantities, technical parameters, quality indicators, or directly expressed consumer preferences. For example, tons of steel, kilowatts of electrical energy, the number of floating-point operations per second, the number of transistors, the percentage distribution of assortments leaving stores, directly expressed consumer preferences in multiple-choice surveys, wear index, rejection coefficient, thermal insulation coefficient, percentage content of desired components, material density and purity, etc. This allows for much more precise control of economic processes.

The analysis of price as a one-dimensional regulatory signal is not just an abstract theoretical consideration. The same mechanism of variety reduction is clearly revealed in historical practice and in contemporary large organizations, which—facing the problem of coordinating highly complex systems—gradually limit the role of price as the primary carrier of regulatory information.

In the early years of industrialization in the USSR, planning as a new regulator encountered the same problem as any other new regulator—it could initially show only a small part of its effectiveness. Many industries largely operated on one-dimensional parameters: there were few diverse quantitative indicators, and many qualitatively significant degrees of freedom in the economy remained outside the direct observation of the regulator. This resulted in problems typical of a low-resolution regulatory system.

However, it is important to note that these problems did not stem from the nature of central planning itself but from its immature and incomplete form. The system did not then have sufficient regulatory variety to effectively compensate for economic disturbances. As the system developed and centralization increased, the number of directive parameters describing the state of the economy gradually increased—the plan ceased to be purely quantitative and began to include qualitative indicators, material intensity, energy intensity, and technical parameters.

In other words: the regulator increased its variety, trying to better absorb the variety of disturbances. This is a classic mechanism of regulator learning, also known in the market economy at its beginnings, when it generated very low efficiency. Compared to the market, however, it can be said that the process of adapting the Soviet central plan to variety was much faster, as it began as early as 1933-1934 (directive planning was introduced in 1928) and achieved exceptionally tangible effects in 1948-1955 in the form of very high growth rates for all sectors of the economy based on increased labor productivity, a significant reduction in material and energy intensity, and improved quality of goods.

This process was reversed during the period of decentralization initiated in the Soviet Union. Reducing the number of plan directives, loosening control over its implementation, the sovnarkhoz reforms, and the later Kosygin reforms limited the center's ability to impose multi-dimensional norms and parameters. At the same time, the process of further improving planning methods was stopped, despite the increasing complexity of the economy. As a result, the plan was simplified again, and the description of the state of the economy began to lose resolution.

The widely cited economic failures of the USSR in the 1960s-80s (such as the famous planning of chandelier production in tons) did not result from the absurdity of the method itself but were symptoms of the loss of the ability to control qualitatively significant dimensions of production. It was not central planning that was at fault, but its progressive erosion.

The limitations of price as a carrier of regulatory information are even more visible in the contemporary practice of large corporations. Corporations such as Amazon operate in a market environment but internally rely less and less on price as the main coordinating signal. Instead, they use multi-dimensional descriptions of the system's state: data on user behavior, search queries, reviews, delivery times, reliability, demand probability distributions, or logistical parameters. Price, if it appears at all, plays a secondary and auxiliary role.

From a cybernetic perspective, this is not a coincidence. As the complexity of the system increases, the one-dimensional price signal ceases to be sufficient to distinguish states important for the stability and optimization of processes. Large organizations, to effectively regulate their internal processes, replace price aggregation with multi-dimensional aggregation, thereby increasing the effective variety of the regulator. Thus, they confirm in practice the conclusion derived from Ashby's Law: the regulation of complex systems requires a richer description of the state than that offered by price.

Aggregation

An essential element of maintaining the controllability of a complex system is the reduction of variety reaching the regulator. This reduction should not be understood as the removal of information but as its processing in such a way as not to overload the regulatory channel. The regulator does not need to observe all the details of the system but only those distinctions that are significant from the point of view of regulation. One of the basic mechanisms implementing this type of variety reduction is aggregation.

Aggregation involves combining certain degrees of freedom of the system into aggregates with fewer dimensions. However, it can take qualitatively different forms. The key is to distinguish between aggregation, which condenses information while preserving regulatory distinctions, and aggregation, which destroys these distinctions, leading to a loss of control capability.

Consider a system whose state is described by three variables: temperature, pressure, and flow. Each of these quantities has its own permissible ranges and dynamics, but from the point of view of system safety, their common relation to one criterion—such as the safety margin—is important. Instead of transmitting three independent signals to the regulator, they can be aggregated into one value describing how far the system is from the critical state boundary.

In this case, aggregation reduces the number of observed variables, preserves information important for regulation, and enables a faster and more stable response from the regulator.

Although the details of individual parameters are hidden, the regulator can still clearly distinguish safe states from dangerous ones and respond appropriately. The reduction of variety does not lead to a loss of controllability but enables it.

Now consider another case. Suppose the system is described by four qualitatively different variables: durability, energy efficiency, output quality, and temporal stability. Each of these corresponds to a different type of disturbance and requires a different type of regulatory response. Instead of observing them separately, the regulator receives only one aggregated value—such as 'overall efficiency.'

In this situation, there are numerous system configurations that give the same aggregate value, even though they are qualitatively different: the system can be very efficient but unstable; stable but energy-intensive; energy-efficient but of low quality. For the regulator, all these states are indistinguishable, even though they require completely different actions.

Such aggregation does not preserve the structure of disturbances, collapses qualitatively different states into one signal, and leads to a loss of compensation capability. This is not a reduction of variety in the functional sense but its destruction. The regulator loses information not because it does not process it but because it was destroyed at the aggregation stage.

Price appears in this light as a special case of destructive aggregation. It is a one-dimensional super-aggregate that combines a vast pool of qualitatively different degrees of freedom of the economy into a single number. Parameters such as the division of labor, quality, durability, energy intensity, safety, environmental impact, technological structure, availability of raw materials, or supply chain stability are reduced to one signal that does not preserve distinctions important for regulation.

From a cybernetic point of view, price is not just a simplification of the system's description but a mechanism leading to a loss of observability. A regulator based on price does not see the difference between disturbances of different structures and cannot adequately compensate for them. Thus, price aggregation reduces the effective variety of the regulator below the level required by Ashby's Law.

The contrast between the market economy and central planning is particularly evident at the level of information aggregation. While the market relies on a single, universal aggregate in the form of price, central planning has the ability to perform multi-level, selective aggregation tailored to the structure of the regulated system.

In central planning, aggregation does not involve reducing all degrees of freedom to a single measure but combining variables of the same type into aggregates that preserve regulatory significance. Quantitative, qualitative, technical, and logistical parameters are aggregated separately, according to their function in the system. As a result, the reduction of variety does not destroy information important for control but condenses it into an operational form.

For example, instead of a single price signal, the plan can operate simultaneously on aggregates such as material balances, energy consumption norms, quality indicators, assortment distributions, or time schedules. Each of these aggregates reduces the number of observed variables but does so in a qualitatively homogeneous manner, not mixing different types of disturbances into one signal.

The key difference is that planned aggregation is structural and hierarchical. Information is gradually condensed at successive levels of organization: from the plant, through the industry, to the system-wide level. At each level, distinctions important for that level of regulation are preserved, and only unnecessary details are eliminated. This allows the plan to simultaneously avoid overloading regulatory channels and maintain observability of the system as a whole.

From the perspective of Ashby's Law, this means that central planning has a mechanism for controlled reduction of variety, which allows the effective variety of the regulator to be matched to the variety of disturbances. Unlike price aggregation, this reduction is not one-step or global but distributed in time and space of the regulatory structure.

This does not mean that every planned system automatically meets the conditions of effective regulation. An insufficient number of parameters, incorrect hierarchization, or inappropriate aggregation criteria lead to the same problems as in the market: loss of regulatory resolution and increased uncontrollability. The difference, however, is that in central planning, these limitations are of a design nature, not structurally irremovable.

In this sense, central planning is not the opposite of aggregation but its development. Where the market is forced to use a one-dimensional super-aggregate, the plan has an architecture that allows for multi-dimensional, selective, and recursive aggregation. It is this architecture that gives it the potential ability to meet the condition of requisite variety in highly complex systems.

Such multi-dimensional aggregation enables the recognition and compensation of disturbances that remain invisible to the market mechanism. This allows regulation to occur before the appearance of classical signs of disequilibrium, which in highly complex systems is a condition for maintaining stability and avoiding information collapse.

Global System Model

Ashby's Law of Requisite Variety is related to another key law of cybernetics—the Good Regulator Theorem (Conant-Ashby). It states that every effective regulator must have a model of the system it controls. This is not surprising but rather a natural complement to the Law of Requisite Variety. The variety of a regulator is not a 'raw number of data' but the ability to distinguish states significant for regulation and to select adequate responses. Without a model, the regulator cannot assign observed signals with control significance, cannot distinguish disturbances of different structures, nor predict the effects of actions. Consequently, its effective variety falls below the variety of disturbances, making regulation impossible.

Central planning, understood as a system of information gathering and coordination at the system-wide level, has a structural advantage here: it enables the construction of a global model of economic processes. This does not mean the necessity of gathering complete information about every detail of the system but the selective acquisition of data significant from the point of view of the reference trajectory: balances, bottlenecks, technological dependencies, quality parameters, implementation times, critical resources. In other words: the plan can (and must) strive for a model that is not a complete reflection of reality but is sufficiently rich to distinguish states significant for regulation. This, by the way, is the essence of a model in cybernetics. The question of how to include local knowledge and not overload regulatory channels will be discussed later in the article (at the stage of hierarchization and regulatory recursion).

The situation is different in the case of the market. There is no single entity that gathers information about the system as a whole; instead, there is a heterarchy of many entities with local, fragmentary observations. A single entrepreneur can build at most a model of their own segment of the system: their industry, supply chain, demand segment. By necessity, it does not include system-wide networks of dependencies, especially feedback loops that only manifest at the level of the entire system. As a result, the market creates not a model of the system but a set of local micromodels.

The key point is that micromodels do not add up to a global model of the system. The properties of complex systems do not result from the sum of the properties of the parts because it is the relationships and feedback loops between the parts that generate the behavior of the whole—this is essentially the defining characteristic of a complex system. The Good Regulator Theorem requires a model of the system as a whole, not a loose collection of models of its fragments. The market thus remains a set of micro-control actions based on partial models that do not have a point of synchronization where they could be combined into a coherent regulatory model.

This can be illustrated with an analogy: imagine a house renovation undertaken by many independent craftsmen. Each has access only to the room they are working in and cannot communicate with the others. Although everyone does their job diligently and rationally from their point of view, a successful house renovation is not simply the sum of the renovations of individual rooms. In such a setup, a plumber's pipe might run through the middle of an electrical panel, and a mason might wall up access to a water valve. Similarly, consider the situation of ten drivers stuck in a traffic jam, each using GPS navigation to determine the optimal route. The system suggests the same solution to each of them—detour through a nearby village—which quickly leads to its blockage. If instead of summing local perspectives, a global model were used, this situation would not occur.

Finally, the argument about the supposed 'summation' of dispersed local knowledge is indefensible on the grounds of formal mathematics. It is based on the false assumption that the economic system is linear and additive, meaning that the influence of one element does not depend on the state of others.

To demonstrate this, consider a simple interaction model. Let a key characteristic of the system (e.g., the efficiency of a complex process) be described by a product function: $f(x_1, x_2) = x_1 \cdot x_2$

Assume that entity A observes and optimizes only the variable $x_1$, and entity B only the variable $x_2$. This is a classic market situation: division of knowledge and lack of a common model.

The key mathematical fact, however, is that it is impossible to reconstruct the global function of the system by simply adding the knowledge of these entities. There are no local functions $f_1$ and $f_2$ that satisfy the equation: $f(x_1, x_2) = f_1(x_1) + f_2(x_2)$

This is a phenomenon of non-additive separability. It means that in a system containing interactions (multiplication), the value contributed by the variable $x_1$ is inseparably dependent on the level of the variable $x_2$. The sum of local optimizations does not yield a global optimization here because none of the entities see the interaction component—only the global model sees it.

In mathematical analysis, the following property holds: if a function of many variables is additively separable (i.e., it can be written as the sum of functions dependent on single variables), then its mixed derivatives are zero. In other words, the lack of couplings between variables is formally expressed by the vanishing of mixed derivatives.

For the function $f(x_1, x_2) = x_1 \cdot x_2$ the mixed derivative is non-zero, which means there is a coupling between the variables and formally excludes the possibility of an additive decomposition of the function.

The application of this example is not metaphorical. In the cybernetic and economic description of the economy, it is assumed that the global properties of the system are a function of its state variables. If this function were additively separable, it would mean there are no relations and couplings between the elements of the system. In the economy, this situation does not occur: productivity, demand, stability, and growth rate depend on the mutual relations between sectors, technologies, and resources.

Formally, this means that in the economy, the mixed derivatives are not zero. The economy is therefore a nonlinear system in the strict sense, and its global properties cannot be reconstructed by the sum of local models.

Hierarchization and Regulatory Recursion

Another way to reduce the variety of the regulator is through hierarchization. This involves delegating different types of tasks and decisions to different levels of control. Each level processes a specific type of information and performs actions best suited to its level of generality. This way, information does not overload the regulatory channels, and the regulator retains the ability to respond to disturbances of different scales.

Central planning naturally reaches for hierarchization and regulatory recursion at different levels of planning institutions. The central office collects, aggregates, and filters information from regional offices, which in turn aggregate and filter data flowing from local offices. Local units collect information about the state of plants in the area, the availability of raw materials, consumer sentiments, the state of the workforce, technical equipment of production, or local disturbances. This information is then sent up the structure, where it is aggregated and filtered according to the level of control.

Based on data aggregated at the global level, the central planner makes decisions regarding the directions and profiles of production of individual plants in such a way as to maximize the realization of the reference trajectory of the economy. These decisions are then transmitted down the structure, where they undergo further concretization. The mechanism of reporting from the bottom up and decision-making from the top down creates a negative feedback loop for the system, stabilizing its dynamics.

Hierarchization simultaneously solves two seemingly contradictory problems. On the one hand, the center does not suffer from a lack of local knowledge because it is systematically reported. On the other hand, it is not overloaded with an excess of details because, thanks to regulatory recursion, it has only information significant for regulation, free from noise. At the same time, lower levels make decisions within the framework of a global model of the system, not based on models of only local segments.

The situation is completely different in the market economy. Its natural heterarchy prevents the formation of a vertical hierarchical structure in the cybernetic sense. Although there are organizations, corporations, and contractual networks within the market, they do not form a coherent regulatory system covering the economy as a whole. There is no unified mechanism for reporting, aggregating, and filtering information from the bottom up and making decisions from the top down based on a global model of the system.

Information in the market economy remains dispersed and local. Individual entities process data significant only from the point of view of their own interests and their own segment of the system. There is no level at which local information is systematically combined into a coherent description of the state of the economy as a whole. As a result, the heterarchy of the market does not lead to a hierarchical reduction of variety but to its dispersion: each entity has its own partial model, and none has a global model.

The lack of a regulatory hierarchy also means the lack of regulatory recursion. Decisions made locally are not embedded in the structure of superordinate goals and system constraints, and their effects are not corrected by higher levels of control. Feedback loops are indirect, delayed, and often positive, instead of designed negative feedback loops stabilizing the system's trajectory.

At this point, the argument is often raised about the supposed greater flexibility of the market and its ability to make rapid structural changes. This argument, however, concerns the problem of adaptation, not regulation in the cybernetic sense. The analysis of adaptive mechanisms is beyond the scope of this article and will be addressed in a separate text. In the context of regulation, the key point remains that the heterarchy of the market does not enable the integration of local knowledge into a global model of the system.

From the perspective of Ashby's Law, this means that hierarchization in central planning is not an organizational solution or a bureaucratic addition but a necessary mechanism for the distribution of the regulator's variety among the levels of control. It is this that allows the use of local knowledge and the retention of the ability to regulate the system as a whole.

A frequently raised objection to central planning is the claim that local knowledge is lost in the process of aggregating information to the central level. This objection, however, overlooks the basic cybernetic fact: in complex systems, information aggregation is a necessary condition for regulation because without it, the regulator would be overloaded with noise.

The task of aggregation and filtering is not to preserve all information but to extract information significant for regulation—which certainly does not disappear—and to reject noise. In a well-designed hierarchical system, this process is precise and purposeful: each level of control passes up only that information which is significant from the point of view of decisions made at that level—significant information for a higher level does not disappear but is carefully selected for transmission. Information that is crucial for operational micro-decisions remains at the local level because, from the center's perspective, it has no regulatory value and would only be a source of noise.

Even in the most centralized systems, operational micro-decisions are not and cannot be centralized. Decisions regarding the current organization of work, minor technological adjustments, or local conditions for the implementation of the plan remain the responsibility of lower levels. Centralization concerns only regulatory decisions of system-wide significance, not complete information about every smallest detail of the system's functioning.

Moreover, if we accept the argument that information aggregation inevitably 'kills' local knowledge, then we would have to consider the market price, which is an extreme case of hyper-aggregation, as the most destructive regulatory mechanism. Price reduces multi-dimensional information about the state of the system to a single scalar, losing almost all the structure of relations and conditions. Compared to it, hierarchical planning aggregation, which preserves many dimensions of information at appropriate levels of control, is a mechanism that is incomparably less destructive of information.

Decisional Entropy

One of the greatest threats to complex systems is high decisional entropy. The more complex the system, the greater the number of potential decisions that can be made at any given moment, and thus the greater the number of possible disturbances generated by the control decisions themselves. Decisional entropy directly increases the variety of disturbances acting on the system. This means that for a complex system not to be destabilized by its own decisions, it must have a regulator with exceptionally high effective variety.

The market economy is characterized by systemic decisional entropy. The multitude of autonomous actors making one-dimensional decisions based on local signals and without access to a global model of the system means that each decision is burdened with great uncertainty regarding its system-wide effects. The lack of decision synchronization further amplifies this phenomenon: decisions made independently and asynchronously generate a huge number of possible system configurations, a significant portion of which are disruptive.

As a result, the market not only responds to external disturbances but also produces disturbances on a massive scale, resulting from uncoordinated allocation decisions. From the perspective of Ashby's Law, this means a drastic increase in the variety of disturbances, with a simultaneous lack of a mechanism for their reduction at the system level.

The situation is different in the case of central planning. Access to a global, multi-dimensional model of the economy enables top-down synchronization of decisions and their alignment with the reference trajectory. The central regulator does not completely eliminate the risk of decision errors but significantly narrows the space of possible decisions, thereby reducing the system's decisional entropy. Decisions are not made independently but within the framework of a coherent model of dependencies and constraints.

For example, the owner of a private enterprise, making an allocation decision, is burdened with a series of potential disturbances. They may duplicate the efforts of other entities, wasting resources. They may make a decision leading to the collapse of their own plant, which entails the breaking of supply chains and disruptions among contractors. They may also produce goods for which there is no real demand, leading to the waste of labor and raw materials. The central planner, having knowledge of the production of all plants, the network of industry connections, and the structure of demand, is able to limit or prevent such errors through the coordination of decisions already at the planning stage.

It is often argued that in the case of central planning, a decision error can affect the entire system, whereas in the market economy, the effects of errors are limited to individual entities. This argument, however, overlooks the key cybernetic difference between the scope of the effects of an error and the frequency and structure of errors.

In the market economy, decision errors are dispersed but occur massively and permanently. Each independent decision burdened with high decisional entropy generates potential disturbances that can propagate through the network of economic connections. The cumulative effect of thousands of local errors often leads to systemic disturbances, even though none of them were 'global' in intent.

Central planning works the opposite way: it reduces the number of possible decisions and their randomness, drastically decreasing the frequency of errors. From a cybernetic point of view, this is not a flaw but a controlled centralization of risk that enables its identification, correction, and compensation through negative feedback loops. A centralized error is a visible and correctable error; an error dispersed in the market's heterarchy often remains invisible until the moment of crisis.

Consequently, it is not the scale of a single error but the structure of error generation and correction that decides the stability of the system. From the perspective of Ashby's Law, central planning limits decisional entropy at the source, while the market internalizes it as a permanent source of disturbances.

The argument that an error in central planning affects the entire system, while an individual error in the market affects only a single entity, is based on a false understanding. It is analogous to stating that since everyone can use any methods to extinguish a fire chaotically, an error by one participant will only affect them, whereas in the case of an organized fire brigade, a wrong decision can endanger everyone, so it is better to eliminate the fire brigade and extinguish fires chaotically.

This reasoning overlooks the fact that in the first case, the number of errors is enormous and uncontrolled, and their cumulative effect leads to the spread of the fire, while in the second case, errors are rarer, visible, and correctable. The problem is not the scale of a single error but the structure of its generation and the system's ability to correct it.

It is often assumed that the errors of micro-actors in the market economy have only local effects, while an error by the central planner affects the entire system directly. This distinction, however, is oversimplified and does not reflect the actual dynamics of complex systems. In practice, the errors of micro-actors affect the entire system indirectly but through the network of connections and couplings can accumulate and propagate, leading to system-wide disturbances. The locality of a decision does not mean the locality of its effects.

Simultaneously, the argument about the 'catastrophic scale' of an error in central planning overestimates both the probability of such an error and its durability. In a hierarchical system, the risk of making a fundamental error is relatively low because decisions are made based on a global model of the system and are subject to continuous verification through feedback loops. The vast majority of errors are ordinary deviations that can be quickly detected and corrected thanks to central coordination. The key here is not the hypothetical possibility of a large-scale error but the system's ability to identify and correct it.

A good illustration of this difference is the comparison of energy systems. A nuclear power plant is an object with a potentially large scope of impact in case of failure, but at the same time, it is characterized by a very low probability of a serious error and an extensive system of safeguards, monitoring, and corrective procedures. Starting a fire chaotically, without control and coordination, carries smaller consequences of a single error but generates them massively and without a mechanism for their isolation. From the point of view of risk theory, it is not the maximum possible damage that is important but the product of the probability and scale of the effects, as well as the system's ability to limit and correct deviations.

Similarly, in the economy: central planning centralizes risk but also controls and minimizes it, while the market disperses decisions but generates a permanent overproduction of uncoordinated errors. From a cybernetic perspective, it is not whether an error can be 'large' that matters but whether the system has a regulator capable of its rapid detection and compensation.

Finally, note that in complex systems, local disturbances can undergo nonlinear propagation to the level of a systemic catastrophe. In the market economy, the lack of isolation mechanisms means that an error by a single actor—a plant's bankruptcy, a credit decision, a contract breach—can spread through the network of financial, logistical, and informational connections to a much wider scale. This phenomenon is well known in systems theory as the butterfly effect, where a small initial deviation leads to disproportionate global effects. This means that not only can a central plan generate an error that affects everyone. In fact, it was such a butterfly effect in the form of the collapse of Lehman Brothers in 2008 that caused the great financial crisis of the capitalist world.

Deliberate Production of Positive Feedback Loops

Moreover, the market economy not only fails to fully compensate for disturbances but endogenously produces positive feedback loops, increasing the variety of disturbances without simultaneously increasing the variety of regulation. This mechanism is not an anomaly or pathology but results directly from the structure of market competition.

An inherent feature of competition is the elimination of some entities from the economic process. This elimination means not only a change in the ownership of resources but often their temporary or permanent exclusion from circulation: production downtime, breaking of supply chains, loss of employee competencies, degradation of technical and organizational capital. Each such exclusion is a disturbance that propagates through the network of economic connections, generating positive feedback loops.

The most extreme form of this mechanism is economic crises, which serve the function of market 'cleansing.' However, crises do not reduce disturbances in a controlled manner but cause their violent accumulation and propagation, leading to mass destruction of material, human, and organizational capital. From a cybernetic point of view, a crisis is an example of a strong positive feedback loop in which local disturbances are amplified to a systemic scale.

An additional source of positive feedback loops is the deliberate actions of market entities aimed at local optimization at the expense of the stability of the system as a whole. These include, among others, the externalization of external costs, planned obsolescence of products, blocking or destruction of innovative capital by sectors of the old production structure threatened by it, as well as the creation of demand for goods harmful from the point of view of the system's long-term reference trajectory. These actions increase the variability and instability of the system, generating disturbances that the market does not have a mechanism to neutralize at the system-wide level.

In contrast, central planning does not have endogenous mechanisms for the deliberate production of positive feedback loops. Its regulatory structure does not reward the elimination of entities or the destruction of resources as a method of coordination but is based on designed negative feedback loops aimed at stabilizing the system relative to the given reference trajectory. This means that disturbances are treated as deviations requiring compensation, not as a tool for selection.

Isolation of Disturbances

An important element in minimizing disturbances in complex systems is the regulator's ability to isolate them so that a local deviation does not take on a global scale. To meet the optimality condition in light of Ashby's Law of Requisite Variety, the regulator must be able to absorb disturbances at the lowest possible organizational level. Otherwise, the rate of increase in the variety of disturbances far exceeds the rate of increase in the effective variety of the regulator, leading to the destabilization of the system.

The ability to isolate disturbances is possessed only by a regulator with a global model of the system and a hierarchical control structure that enables the synchronization of the actions of its elements. In conditions of autonomy of many actors, lacking access to a comprehensive model and acting independently, effective isolation of disturbances becomes impossible. The lack of coordination means that responses to a disturbance are not synchronized, and their effects can mutually reinforce each other.

Central planning structurally meets the conditions for the isolation of disturbances. If a failure, downtime, or other type of disturbance is detected in a given plant, region, or sector, this information is reported to the central level. Based on the global model of production and logistical connections, a retuning of activities takes place both at the site of the disturbance and in units located on the potential trajectory of its propagation. The aim of these actions is to minimize the effects of the disturbance and prevent its further spread.

The market economy, on the other hand, does not have an authority standing above the current activity of entities that could coordinate such actions. The market's self-regulation mechanism is based on the actions of heterarchical actors who have only fragmentary knowledge of the system and are guided by their own decision-making interests.

In such a setup, a disturbance arising in one place may remain outside the informational horizon of other entities, and units located in the line of its impact do not have any systemic safeguards against its effects.

Consequently, disturbances in the market economy tend to propagate in a cascading manner, and in extreme cases, they can take the form of disproportionate systemic effects known in systems theory as the butterfly effect. From the perspective of Ashby's Law, this means that the market not only does not isolate disturbances but facilitates their spread, increasing the effective variety of disturbances beyond the regulatory capabilities of the system.

Does the Multitude of Actors in the Market Mean Greater Variety?

A frequently raised argument in favor of the market economy is the claim that it is an exceptionally diverse regulator because it consists of millions of independent actors: enterprises and consumers. This argument, however, is based on the false equation of the number of elements in the system with the effective variety of the regulator in the sense of Ashby's Law.

First, it should be noted that the same physical base of actors also exists in the planned economy. Production plants, enterprises, workers, and consumers do not disappear with the introduction of central planning. On the contrary—they are directly included in the regulatory system as units of the plan, equipped with specific roles, tasks, and reporting channels. The number of actors, therefore, does not constitute any distinguishing feature of the market compared to the plan.

Second, in central planning, there is an extensive, multi-level hierarchy of regulatory institutions: from the plant and local level, through the regional, to the central level. Each level processes information and makes decisions within the scope of its competencies. This means that the number of actors participating in the regulation process is not smaller in the plan than in the market but differently organized. The difference does not lie in the number of elements in the system but in the structure of regulatory relations between them.

Third, the market as a regulator is not a set of actors but a set of coordinative relations between them. It is these relations—not the number of participants—that determine the effective variety of the regulator. In the market economy, these relations are extremely simplified and are mainly based on the one-dimensional signal of price and derivative supply-demand stimuli. In the planned economy, regulatory relations are much richer: they include direct quantitative and qualitative parameters, schedules, technological norms, material and energy connections, and hierarchical feedback loops.

Fourth, the multitude of market actors does not translate into an increase in the variety of the regulator because their decisions are not independent in the regulatory sense. They are strongly correlated through common price signals, expectations, and informational stimuli, which leads to the convergence of behaviors, herd effects, and mass reactions. From a cybernetic point of view, this means a reduction, not an increase, in variety, despite the apparent number of decision-makers.

Fifth, the lack of hierarchy and synchronization means that the variety of market decisions is not functionally distributed among the levels of control. Instead of absorbing disturbances, the market generates high decisional entropy and increases the number of potential disturbances, without having a mechanism for their coordination relative to the system's reference trajectory.

Central planning works the opposite way: it does not maximize the number of autonomous decisions but maximizes the effective variety of the regulator through a multi-dimensional description of the system, hierarchization of decisions, and synchronization of actions at all levels. In light of Ashby's Law, the regulatory capability is not determined by the number of actors but by the richness and structure of the control relations that the regulator can generate.

Feasibility of Central Planning

A frequently raised objection to central planning is the claim that building a global model of the system, limiting the decision space, global coordination, top-down gathering of local knowledge, and processing multi-dimensional information exceed the real capabilities of any regulator. This objection, however, is based on the false assumption that effective regulation of a complex system requires a complete, photographic representation of all its details. In reality, cybernetics has always assumed that the regulator operates on a simplified model containing only information significant for regulation, not on a complete copy of the system.

Second, this belief is based on a false intuition of the discrepancy between the seemingly 'simple' process of control and the gigantic complexity of social life. This, however, is an untrue simplification resulting from the observation of everyday life, where we constantly encounter the vast complexity of society and rarely encounter complex regulatory mechanisms that are performed in specialized centers, and thus outside the source of everyday intuition.

History provides numerous evidence that complex economic and logistical processes were effectively coordinated without the use of modern computational tools, solely through analog processing, organizational hierarchy, and reporting procedures. Examples include large-scale economic and logistical operations related to global conflicts: the Seven Years' War, the Napoleonic Wars, and World War I and II, in which the production, transport, supply, and mobilization of millions of people and enormous material resources were coordinated. The fact of their implementation proves that the processing of information necessary for the regulation of systems of enormous complexity was possible even with very limited technical means.

An additional empirical confirmation of the feasibility of planning is the practice of socialist countries, particularly the Soviet Union during the period of industrialization, post-war reconstruction, and the economic miracle of the 1950s. Despite technical limitations and institutional immaturity in the early phases, the planned system was able to coordinate production processes on a scale and complexity unattainable for market economies of that period. Achievements in industrialization, electrification, the development of heavy industry, nuclear energy, and the space program are evidence that central economic regulation was not only possible but also effective.

Such high rates of growth in national income, industrial production, labor productivity, TFP, and the scale of own, pioneering technological solutions as the USSR achieved in 1928-1961 (especially in the more mature, later phase of this period) were never achieved by any capitalist country to such a large geographical extent under such unfavorable conditions. This is evidenced not by Soviet propaganda but by analyses valued in the global academic community by authors such as Grigorij Khanin, Vladimir Popov, or Robert William Davies, as well as CIA reports.

The later decline in dynamics in the 1960s-80s resulted— as already mentioned—from the dismantling of the planned economy, which began during Khrushchev's time, long before the beginning of perestroika, and was silenced in the mainstream narrative.

Importantly, as the complexity of the system increases, so does the potential computational power of the regulator. Contemporary digital data analysis tools, Big Data systems, and artificial intelligence algorithms enable the processing of information on a scale and detail incomparable to the possibilities of past eras.

An empirical proof of this thesis is the practice of large corporations, which already today coordinate global production, logistics, and distribution chains based on a centralized plan powered by modern technologies, not on the market mechanism in the classical sense.

This means that the argument about the infeasibility of planning loses its validity with the development of regulatory tools. If large private systems on the scale of countries are able to effectively model and control complex economic processes using modern information technologies, then it is even more possible to apply the same principles at the system-wide level. From the point of view of cybernetics, the problem is not the lack of technical possibilities but the regulatory architecture and the way decisions and information are distributed.

Summary

The conducted analysis of the dispute between the market economy and central planning in the light of Ashby's Law of Requisite Variety leads to clear conclusions. The economy, as a complex system, requires a regulator capable of absorbing the variety of disturbances through a multi-dimensional description of the state, functional aggregation, hierarchization of decisions, reduction of decisional entropy, and isolation of disturbances. These criteria are not ideological but result directly from the formal properties of complex systems.

The market does not meet these conditions. As a coordination mechanism based on the one-dimensional signal of price, heterarchy of decisions, and lack of a global model, it reduces information in a destructive manner, generates high decisional entropy, and facilitates the propagation of disturbances through positive feedback loops. The multitude of market actors does not increase the effective variety of the regulator but leads to the convergence of reactions and the overproduction of uncoordinated decisions that the market cannot absorb.

Central planning works the opposite way. It does not eliminate the actors of the system or local knowledge but includes them in a hierarchical regulatory structure, enabling effective, selective aggregation of information, synchronization of decisions, and local suppression of disturbances. Thanks to the existence of a global model of the system, central planning can limit the decision space to variants consistent with the reference trajectory, reducing decisional entropy and preventing the cascading propagation of errors.

Arguments about the alleged infeasibility of planning lose their validity both in the light of historical experiences and contemporary practices of large corporations. Since complex economic processes were effectively coordinated using analog tools, and today private entities use Big Data and artificial intelligence for the central coordination of global production and distribution systems, the problem is not technical limitations but regulatory architecture.

In the light of Ashby's Law, the dispute between the market and the plan ceases to be an ideological dispute. It becomes a question of which regulatory mechanism has sufficient variety, structure, and synchronization capability to control a system with the complexity of the modern economy. Cybernetic analysis leads to the conclusion that these conditions are met by central planning, not the market mechanism.