In post #2, “Control,” I raised the question of how cybernetic thinking applied to the governance of society. Cybernetics was defined by Norbert Wiener as the study of communication and control in the animal and machine. I asked: what about social systems? Are they animals or machines?
This is a serious question. As digitization spreads and the role of AI-driven automation in social organizations advances, we will be asking variants of that question for the next 20-30 years.
In post #4, “The Progress of Digitization,” we saw that digital technology was founded on the ability to automate calculations using a specific technology (digital electronics running on semiconductors). As digital information systems spread into more areas, the number and type of activities susceptible to being controlled or governed by automated calculations increases.
Anyone making a rational, historically grounded assessment of this problem would not panic about it. It sounds scary until you realize that humans have been reducing social operations to procedures and rules for centuries. It is an essential part of what human societies do. There is a protocol for getting married. There are required procedures to follow when you are born (you are even assigned a number). We have been automating industrial production processes for three centuries. What’s different now is the comprehensiveness of the digital transformation. The automation capabilities are more powerful, more globally integrated technologically, than any ICT infrastructure human society has ever known.
I asserted in #2 that society is not a machine; and I still believe that. No one has argued with me about it, either. That may be because the normative implications of the opposite position are repulsive to most of us. If society is really a machine, then the machine has a designer (which is not us) and controller (also not us) who built the machine, defined its purposes and operates it to optimize them. The rest of us are just inputs or resources to be employed in the pursuit of those objectives – if there is any place for us at all. That “society as machine” sounds undesirable does not, of course, mean that it isn’t true; it could be a correct description, however undesirable. But we have too much firsthand knowledge of how society operates to believe that. Social cooperation and institutions are not designed; they evolve. Even the most oppressive regimes must contend with the entropic autonomy of human beings, individually and in groups, who either don’t follow the program or actively resist it.1
We can also agree that society is not an animal; that is, it is not a biological organism. Biologists, I am pretty sure, would agree with me about that. Humans are biological organisms, for sure, but their social systems are not. They are complex communicative, production, and authority relationships that replicate themselves, in some sense, but we don’t give birth to them or harvest them from farms. (Though one might, with Laurie Anderson, believe that language is a virus from outer space. We will set aside that issue for another day).
So, if human society is neither animal nor machine, does that mean that cybernetics has nothing to contribute to social science? I think not. It just means that the application of cybernetic principles to social systems is incomplete. They have not been applied in ways that fully understand and respect the distinctive features of humans in social systems. The solution to this shortfall, in my view, is political economy. By that I mean the infusion of cybernetics into institutional theory (though there is a surprising amount of it already there), and the use of political economy to analyze the digitization of society.
The term political economy is clumsy and has a somewhat antiquated ring to it. That’s because it is, in fact, a very old term; 401 years, to be exact. The origin is credited to a book published in 1615 by the French writer Antoine de Montchrétien. The term he is credited with creating – political economy – became the accepted name for a new social science for the next 300 years.
_written_by_Antoyne_de_Montchr%C3%A9tien.jpg){kind=link}
All the scholars that we now associate with the origins of Economics did not call themselves economists. They called themselves political economists. Adam Smith and Thomas Malthus in the 18th century, Ricardo and Marx in the 19th century, all referred to their field as political economy, not economics. Carl Menger, Alfred Marshall, William Jevons and other early neoclassicals also positioned themselves as political economists, but marginalism paved the way for dropping the ‘political’ part.
Montchrétien framed his new term (political economy) as a challenge to Aristotle, whose discussions of the topic used the Greek word Οικονομικά, a combination of the words for “household” and “management.” The Greek philosophers had, he said, “ignored the principle of public householding with which the responsibility of the state should be mainly concerned.” In other words, the scope of economic management goes beyond the household to a larger collective unit, the state.
Montchrétien was not just currying favor by assigning his King an elevated role in the governance of society. The invention of the term coincides with a turning point in Western institutional evolution. This is early modern Europe, and profound changes in the organization of society were underway. Jean Bodin and Thomas Hobbes, two political philosophers, were developing the theory of sovereignty and the state. It was also the earliest days of an industrial trading economy in Europe, with an organized division of labor capable of producing at scale for regional and international markets. As the trading economy grew, we see centralized territorial states replacing local, feudalistic forms of governance.
It is no accident that this mash-up of politics and economics, of states and markets, surfaces just as these momentous structural changes are taking place in the organization of Western society. As soon as you get something like a fixed-location, centralized territorial state linked by geographically extended markets, society is dealing with problems in political economy. There is now a duality in social governance, one set of controls responding to market exchanges, the other responding to and maintaining political authority.
So, from Montchrétien onward, Western society embarks on a longstanding analysis of, and debate over, the relationship between the economy and the state. Different schools of political economy developed different positions on what that relationship should be, and those debates still rage today. But they all share the idea that economy and state are systemically related. And they all agree that the purpose of linking politics and economy together is to give us some macro-level control over the society we live in, through the exercise of public policy in a context where human beings have both collective and individualized modes of decision making.
How is this deep dive into the origins of political economy relevant to cybernetics? My contention is that a social cybernetics never evolved because cyberneticians had no concept of political economy. And this explains why computing, cybernetics and information theory led to incredible progress in the production and operation of information systems and cyber-physical systems – machines, in other words - but fostered no equivalent revolution in social science or the theory and practice of public governance. On the contrary, it fueled visions of technocracy, central economic planning, and other promises of “scientific” control of society, as if society were a machine, yet never really provided practical guidance on how to build or run this machine.
If we look at prior attempts to apply cybernetics to society, we can see four critical disconnects or failings:
They did not recognize markets, money and the price system as a social governor;
They did not have a theory of the state;
They did not recognize the state-market duality as the presence of two distinct control logics in social organization;
Except for the developers of so-called second-order cybernetics, they did not understand that human policy makers and governors are in social systems and not an external designer or controller of them. (Those who did recognize this, however, still missed the first three points.)
Taken together, these four problems add to one conclusion: cybernetics needs a political economy.
Norbert Wiener himself embodies all these problems. A humanist as well as a scientist and mathematician, Wiener sensed that communication, control and feedback were central to the functioning of human society. But he knew too little of economics and politics to integrate them into cybernetics. His 1947 book, ostensibly presenting Cybernetics as a mathematical science of communication and control in animals and machines, contains many pages of ruminations and speculations about its relevance to human society. His next book, The Human Use of Human Beings (1950), focuses almost entirely on that, but in it, we also see Wiener recoiling from the notion of society as a cybernetic machine. Wiener also seemed to believe that studies of society could not be scientific at all, as we will see in section 4 below.
So, let’s go through each of these failings one by one to see in more detail how they reflect the absence of political economy
In all his mathematical formulations of feedback and control mechanisms, Wiener never once recognized the price system as a steering system for the economy. He dismissed neoclassical price theory because “[t]he mathematics that the social sciences employ and the mathematical physics that they use as their model are the mathematics and mathematical physics of 1850”. (God and Golem, p. 90) Nor did he see money as a critical medium of social communication. He anticipated digital robots but never imagined digital money.
Near the end of Cybernetics (1947, p. 158), Wiener rejects the idea that market competition is a homeostatic process governing the economy. With a vague appeal to von Neumann’s game theory, he asserts that markets inevitably result in monopoly (which is untrue), but at the same time makes the contradictory claim that the result of complex games was “one of extreme indeterminacy and instability.” Wiener’s discussions of economics exude disdain for commerce, businesspeople, and markets. He was equally cynical about political games – in fact, he did not much differentiate between the two.
In the end, Wiener’s complaints about the economy’s “lack of efficient homeostatic processes” is more of a moral judgement than a scientific one. Society did not run like a well-engineered machine, and that disappointed him. Social actors were self-interested and could cheat on each other to achieve power or wealth. No shit. These problems occur precisely because people are not machines. They have autonomy, divergent views, conflicting interests, and sometimes act irrationally. Wiener the rarified scientist-mathematician seems to have been astounded by these simple facts about human society.
Once one dispenses with the expectation that markets should work like a conflict-free machine, it becomes easy to find communication, feedback and control concepts all over economic governance and in economic theory. Menger’s 1892 account of the origins of money as a generalized medium of exchange would make a nice supplement to Wiener’s chapter on language in The Human Use of Human Beings. Starting in the 1920s, two other Austrian economists (Mises, Hayek) redefined prices not as numbers that could be used by mathematicians to govern the economy, but as signals endogenously generated by a massively distributed set of human exchanges. No central planner could gather and comprehend all the information perceived and generated by this distributed system, yet without these signals, the efficient allocation of resources was impossible. Unaware of this debate, Wiener could not appreciate its relevance to notions of a self-governing social system.
Positive feedback loops, in the form of asset bubbles, have been recognized as a phenomenon in business and economics since the 17th century; more recently they have been formally modeled (e.g., network externalities increasing returns, path dependence). Human perceptions and expectations are recognized as critical economic forces by all schools of economic thought, and where there are perceptions and expectations, there are distortions, errors, and major fluctuations governed by positive feedback loops.
But it is in the practice of monetary policy where the presence of cybernetic concepts – and the aspiration for mathematically grounded technologies of societal regulation– is most obviously evident. The Taylor rule used by the U.S. Federal Reserve to regulate interest rates is a simple example. It is a (somewhat weak and arbitrary) attempt to define a homeostatic monetary policy mechanism by manipulating the federal funds rate – a point of leverage created by a central bank. The Taylor rule models how to close the gap between the desired and actual rates of inflation and/or GDP growth, a form of deviation-reducing feedback (which relies on human decisions but is supposed to work best when these decisions are “automatic”).
More recently, central bank research teams are using artificial intelligence to develop their models. IMF economists, for example, use Deep Reinforcement Learning (DRL) to simulate optimal monetary policy under nonlinear conditions. A literature review in PNAS discusses the application of DRL to create more desirable social mechanisms for people, such as “optimal tax policies,” or “redistribution policies that win the popular vote among human users.” The authors claim that “this work opens new opportunities for AI systems to act as ‘intelligent institutions’ that propose new ways of optimizing human interaction for collective social benefit.” Here we see an open embrace of modeling society as a complex machine.
The role of cybernetic principles in regulating the economy is thus deeply entangled with debates over the role of mathematics in the social sciences. This is an issue I will try to take up in more detail later. Note that Menger and the Austrian School, which we might call the human-grounded school of economics, rejected the complete mathematization of economics. On the other hand, if human market activity really can be reduced to numbers, then in a digitized society mathematical tools provide the basis for a technology of social control, a science of social engineering.
Wiener had no theory of the state, either. But why should we expect a mathematician to be a social scientist? Karl Deutsch, a renowned political scientist and younger contemporary of Wiener, did try to apply cybernetics to political institutions. His work illustrates both the promise and the failings of early efforts to arrive at a cybernetic political economy.
Deutsch’s first book, Nationalism and Social Communication (1953), was written “under the influence” of Wiener, whom he credits for the theoretical lens that society can be understood by means of its shared information channels and communication flows. Nationalism started with a puzzle that is central to political economy: what explains state formation around a national identity? One cannot find the answer to this question in economics. No economies of scale or scope or variations in supply and demand can explain why there is a Belgium and a Russia, or why Belgium and the Netherlands used to be part of the Spanish empire and later became separate states.
Deutsch’s answer was that communities are formed by communications, and more specifically by shared history and complementary capabilities for storing, recalling, disseminating and processing information.2 Deutsch described a process by which a people become a nationality, and a nationality becomes a nation-state when it acquires “a measure of effective control over the behavior of its members.” A positive feedback loop generates a collective political identity, which leads to the formation of political institutions, which in turn gain the monopoly on force. Deutsch recognized the two-sided nature of nationalism as something that brought people together but also divided people into those who belong and those who don’t.
Deutsch’s book The Nerves of Government, published in 1963, tried to present a less evolutionary but more explicitly cybernetic model of the political process. The book suggested that one “could look upon government somewhat less as a problem of power and somewhat more as a problem of “steering.” Government was not just about rulemaking; it was about guiding the ship of state in a way that was responsive to the feelings, needs and beliefs of citizens. Thus, he recognized that the power to steer can be democratically distributed across a broader population.
When Deutsch tries to describe the specific way a population “steers” itself, he of course concentrates only on collective or political steering, not on steering through decentralized economic choices. Thus, he ignores the price system, the economic dimension of social governance, as did Wiener. But his account of political steering is quite fascinating and surprisingly contemporary.
Steering implies values and goals toward which the machine is directed. In his attempt to define political governance in cybernetic terms, Deutsch conceptualized values in terms of decision theory; in his cybernetic political theory, “a ‘value’ is a tendency for a particular class of messages or data to be received, transmitted, or acted upon in preference to others. They were routed through different channels, assigned different priorities, and left different “associative trails” within the network. Deutsch referred to this selection process as the “switchboard.” The values are not permanently hardwired in, however; they are subject to a “learning” process and may be revised as the system receives feedback about the effects of its policies and actions. In response to these changes, it alters the operating rules of its switchboard.
What makes this 1963 observation fascinating, 63 years after it was written, is that Deutsch could have been describing the operation of a neural network. Decisions about how to handle inputs sent signals through different routes, and that choice of routes trained the system to behave in a certain way. That is exactly what a neural network does; only it is implemented physically and symbolically in an electronic information system. So, Deutsch was equating the control of a social institution, a government, with the way a neural network is trained.3
One of Deutsch’s contemporary critics, a young sociologist named Langdon Winner, complained in 1969 that “this theory of politics is based on the design of the digital computer.” Well, yes, Langdon, it was, but it was just an idea and not an implementation. The question of where the decisions that program the digital computer came from was still open. Based on that, it’s not clear whether Deutsch was proposing a machinic society or not. By comparing government to a computing machine informed by feedback, was he just describing the way he thought society works? Or was he also telling us we could program the social computer to manage the behavior of people and organizations in the same way computers and telecommunications manage bits, bytes and automatic devices?
Deutsch did not ask where this state, this social computer, comes from. How did it get there? Who built it? The Nationalism book provides some clues, but Nerves of Government avoids the issue of violence and war. If the state’s monopoly on violence is rooted in war and conquest, and the historical record indicates that it is, we need to ask how far down the road of institutionalizing, taming and governing the use of violence has it come? If the system learns, what does it learn from? Ironically, the political scientist who provided such a nuanced, historically grounded theory of nationalism did not really have a theory of the state.
Lacking both a theory of the state and an understanding of markets, cybernetic theorists were incapable of recognizing dualism, the co-existence of two distinct logics of social regulation. Market governance is based on individual decision-making and decentralized transactions; political governance is based on collective decision making and centralized choices. Human action routinely takes both forms. People exercise agency by acting individually, and by mobilizing or forming groups with shared goals and joint action. The label “political economy” embeds this distinction in its name, while at the same time indicating that social systems inextricably combine both of them.
The co-existence of individual and collective choice in social systems is unavoidable. Ideologically motivated attempts to drive markets, money and trading out of society, advanced so fervently by Marxists, utopian socialists and some religious fanatics, have been largely abandoned in practice but persist in ideologies. Whether it is Mao’s China in the 1960s, Pol Pot’s Cambodia in the 1970s, or today’s “progressive” democratic socialists, efforts to get rid of market economy have proven time and again to lead to economic stagnation and authoritarianism, if not genocide. They may suppress but never eliminate markets.
By the same token, the efficiency and individual freedom fostered by markets promotes a tendency to believe that markets can and should replace states altogether. Ideologically motivated attempts to drive the state out of existence, however, also fail. The form taken by states can vary widely, but a state of some form is inevitable. This is because states institutionalize the human capacity for violence. Physical force will never disappear as a factor in human society; the only question is what forms it takes and how well it is controlled. At best, a state makes the use of violence limited, lawful and fair. At worst, a state is not much different from organized crime. Either way, the absence of a monopoly on violence will foment instability and confusion about the governing rules, leading to violent competitions for power among contending groups. Markets need stable property rights and protection against expropriation, exploitation, and fraud. The absence of a state may even expose a society to invasion by an external state.
In all their discussions of self-organizing systems and homeostatic governance, the cybernetic theorists never recognized the state-market duality.
This brings us to another common failing in the application of cybernetic models to society. The people who saw society as a machine and thought cybernetics would provide the technology for scientific management of it assumed that control was exogenous to the system. The people exercising control, in other words, were outside of the social system they were controlling. That assumption, it should be obvious, is wrong. One cannot steer society, its people and its organizations using some external mechanism, as if the controller was a deity. Lacking anywhere to stand outside of the system, no one person can exert Archimedean leverage on it. In human society, all controllers are inside the system, which means that their actions cannot help but reflect their own distinct perspectives and interests within that system. In many ways, and in varying degrees, the social system has as much control over them as they have over it.
Social engineering assumes that other humans are means to the ends of the engineer(s). But in the real world, there are many other people who, like them, are trying to exert control for their own ends. When it comes to human society, institutions and public policy are not simple technologies, but sites of negotiation, bargaining and struggles for control; a place where purposes coordinate, mix, conflict, and converge. Once you detach the notion of cybernetic control from the assumption that you are the system’s designer-creator; once you realize that you are within it, then you can no longer ethically assume a hierarchical control relationship over social systems as a whole.
This perspective, which places the human social scientist and policy maker within the system being governed, is sometimes credited to Margaret Mead, who is said to have challenged cyberneticians to apply their insights about system organization to themselves in 1968. Heinz von Foerster gave it the label “second-order cybernetics” and in the 1970s shifted cybernetics’ focus from the study of “observed systems” (first-order) to the study of “observing systems”. But this shift locked second-order cyberneticists into an epistemological stance that was even farther removed from policy and political economy than before. Its primary message was that “observers only observe themselves.” (Krippendorf, 1987)
Interestingly, it was Wiener himself, not Mead, who first recognized the importance of human endogeneity. And in his view, it called into question the scientific status of any social cybernetics. In Cybernetics (1947, p. 162-3), Wiener was already criticizing the “excessive optimism” of those who thought that the scientific achievements that gave birth to digital communication and information technology could be extended to anthropology, sociology and economics.4 He made this point precisely because the observer or policy maker is part of the system being observed. He said, “all the great successes in precise science have been made” when there is a “high degree of isolation of the phenomenon from the observer.” In the social sciences, in contrast, “the coupling between the observed phenomenon and the observer is hardest to minimize.”
The pioneers of digital technology realized that communication and information were critical to how all things were organized and controlled, and if that was true of machines and biological systems, it should be true of social systems as well. Once linked to political economy, cybernetics provides a useful entry point into analyzing the way changes in information and communication technology change the governance of social systems, while recognizing the autonomy and agency of the humans affected, and the limits and indeterminacy of efforts to shape society from within it.