Epistemic Dimensions of Political Conservatism, I: The Social Knowledge Problem
or, How to Back Into Subsidiarity Without Really Meaning To
I am a conservative. So is Norman Podhoretz. So was Joseph de Maistre. But I did not support the Iraq war, either at the time or now. And I am no kind of monarchist. What, then, can be made of the content of the word "conservative," if it allows for such heterogeneity under its banner?
Much. There are many commitments that I think conservatives usually share. The ones I want to focus on are the epistemic commitments that a conservative like me has, ones which make me a conservative. There are a few, but this essay will focus on one, what, following Hayek, I’ll call the social knowledge problem.
Before getting started, I want to emphasize that I am not going to be arguing for conservatism. Nor will I really argue for any of the epistemic commitments that I lay out. Rather, my goal is explanation. Conservatives are often cast as hidebound, fearful of deep thought, overly-deferential to tradition to the exclusion of reason. Surely this is true of some. But I do not think it is true of all of us. If this essay (and the ones that will follow it) convince you of anything, hopefully it's that. There will probably be many lacunae, and many places in which my exposition will be prodded and poked. But since the goal is exposition, this doesn’t worry me.
Onward.
1 -- The Economic Knowledge Problem
Friedrich Hayek's paper "The Use of Knowledge in Society" is technically about a problem with the allocation of economic resources faced by central planners. In very brief form, the problem is this. The issue we have to address, when figuring out how to allocate resources, is "a problem of the utilization of knowledge not given to anyone in its totality."1 The issue faced by central planners is that the knowledge they have available to them in the moment is that of the general rules of economics, what Hayek calls "scientific knowledge." But this is not the only kind of knowledge that is necessary for economic planning. One also need what he calls "local knowledge," the "knowledge of the particular circumstances of time and place."2 To allocate economic resources efficiently, then, according to Hayek, one needs both scientific and local knowledge.
But the planner cannot ever have access to the local knowledge in a way that will be useful to him. This is so for a few reasons, according to Hayek.
1.1 -- Abstraction
Any transmission of information to planners in the form of statistical data inevitably glosses over numerous details relevant to the allocation of, say, widgets -- it abstracts away from circumstances, and must do so. A widget is a widget, no matter where it is produced, for the purposes of widget statistics. But if the planner says "make more widgets," he inevitably ignores numerous details about widget plants which need to be ignored to group these together as similar products.
Now suppose the planner wonders whether to increase the output of widgets. He looks at his stats, and concludes that not enough widgets are being made. Consequently he gets all the managers of the widget factories and tells them to make more widgets. Assuming his orders are obeyed, these factories will then produce more widgets.
But there's a problem here. There are innumerable details at any particular widget factory that affect the output. The decisions the factory manager has to make are not based on the general structure of the factory plus the laws of economics alone. They’re also based on facts like whether his floor manager has a new child and can't work at full capacity for full hours, whether the price of widget materials has risen a bit, whether there’s a cold snap coming in which will make the machinery stick and require extra attention, and so on. So there are constraints on whether these widgets can be produced at a standard cost, in a particular amount of time, or according to whatever other constraints the planner has in mind. And this information is not available
1.2 -- Time scale
You might think at this point that the answer is just more data or more computing power. If the planners had the relevant local knowledge, they could issue the correct -- that is, the most efficient -- instructions for the factory.
But here, according to Hayek, there's another problem. Planners think that the changes that affect economic factors like cost of production happen, and are therefore relevant, at a relatively long time-scale:
[T]here are few points on which the assumptions made (usually only implicitly) by the "planners" differ from those of their opponents as much as with regard to the significance and frequency of changes which will make substantial alterations of production plans necessary.3
He continues:
In a competitive industry at any rate-and such an industry alone can serve as a test-the task of keeping cost from rising requires constant struggle, absorbing a great part of the energy of the manager. How easy it is for an inefficient manager to dissipate the differentials on which profitability rests, and that it is possible, with the same technical facilities, to produce with a great variety of costs, are among the commonplaces of business experience which do not seem to be equally familiar in the study of the economist.4
The problem here is this. The aforementioned changes which affect the cost of production of widgets (cold snaps etc.) happen on very short time-scales. Once the planners are apprised of them, they may issue instructions. But of course there is no guarantee that their instructions will continue to be relevant by the time they reach the factory manager.
Now this part of the problem might in principle be solved by supposing the planners to have access to an oracle which instantly apprises them of whatever the factory manager knows. Once we suppose such a device, it might seem that we've left the realm of the feasible and entered the realm of science fiction. But it's worth noting that something much like this was actually tried in Chile in the early 1970s. This effort was Project Cybersyn, designed by British cyberneticist Stafford Beer. The idea was to connect the factory floor with the national government so as to allow near-real-time management of the relevant sectors.
Contrary to more dirigiste economic systems, the Cybersyn model emphasized laborer participation in the system. Beer reports Allende's desire that the system be "decentralizing, worker-participative, and anti-bureaucratic."5 This was in keeping with the spirit of Allende's particular form of socialism, with its emphasis not on a dictatorial regime which would enforce socialism but instead on a political revolution in which labor forces would find emancipation from the control of capital owners and ownership over the same capital.6
The failure of Cybersyn appears (I am not an economic or cybernetics historian, so I am relying here on the work of others) to have been that, first, it underestimated the inputs needed to decide efficient economic allocation, and second, that in order to function properly it required a level of social transformation simply not attainable in Chile at the time. Eden Medina writes:
Cybersyn fell victim to the instability accompanying Allende’s programme for socialist reform.
Project engineers found themselves attempting the impossible: modelling an economic system that refused to remain constant using only a subset of the variables needed to understand the system. Production, as gauged by flows of raw materials and finished goods, constituted only one aspect of the Chilean economy – one that increasingly paled in comparison to the economic dislocations of inflation, consumer shortages, political infighting,
US foreign policy, black-market hoarding, labour strikes and increased social unrest. Labour, in particular, did not behave as just another factor of production, but rather as a corpus of self-conscious individuals able to criticise and resist state operations.7
Stafford Beer's assessment echoes this in part:
[T]he model we were using until then could not adequately represent changes that had come about during Allende’s term, and which had crystallized around the events of October, because these were changes in economic management that had nothing to do with ownership in the legal sense.8
Both of these failures are interesting, and I'll have more to say about the second later. But for now, let's press on.
1.3 -- Tacit knowledge
Suppose now that these two problems could be overcome. Suppose that there is a Marvelous Machine, with functionally limitless computing power, designed with the most advanced machine learning in mind, which fulfills the function of a Cybersyn. It instantly connects the planners with the factory workers so that they can adjust in real time to the changes in cost of materials, functionality of equipment, and other factors. Everything that the workers know, the Marvelous Machine lets the planners Would such a machine be able to solve this problem?
No, according to Hayek. For he thinks that this knowledge can in principle not be given to anyone in its totality.9 This is because, according to him, much of the necessary information is not consciously known at all. This isn't explicit in this article, but it appears as a theme in others of his works. For instance, in his essay "Rules, Perception, and Intelligibility," he writes that
there will always be some rules governing a mind which that mind in its prevailing state cannot communicate, and that, if it ever were to acquire the capacity of communicating these rules, this would presuppose that it had acquired further higher rules which make the communication of the former possible but which themselves will still be incommunicable.10
In the argot of analytic philosophy, what Hayek is saying here is that knowledge-that presupposes knowledge-how. Indeed, he explicitly uses the terminology in his book The Sensory Order and cites Gilbert Ryle as his source.11 He takes this anti-intellectualist position on broadly Rylean grounds:
[Conscious] thought must, if we are not to be led into an infinite regress, be assumed to be directed by rules which in turn cannot be conscious -- by a supra-conscious mechanism which operates on the contents of consciousness but which cannot itself be conscious.12
This anti-intellectualism is crucial for Hayek's analysis of the knowledge problem. For the following of these rules, and the results of following such rules, form a crucial part of the data needed to coordinate economic arrangements. And since the rules themselves cannot ever be consciously articulated, they are in principle unavailable for report to the Marvelous Machine. Hence, in principle, a distributed network of economic actors, even with incomplete knowledge, can do things that the Marvelous Machine simply cannot in principle do.13
2 -- The Social Knowledge Problem
So far I've said a lot about Hayek's epistemology, and how it applies to economics, but not a whole lot else. What does this have to do with conservatism? It might be thought that I mean to make an argument for free markets (these are distinct from capitalism to my mind) but that's not my intention. Instead, I want to make the case that there is a social knowledge problem, analogous in key respects to the economic knowledge problem, which suggests that society is best managed when the power to manage it is devolved as locally as feasible. Let me now lay out my case.
2.1 -- Social Abstraction
Recall that Hayek's argument against the transmission of local knowledge in statistical form is that it necessarily abstracted away from many relevant details, details crucial to the efficient allocation of economic resources. My claim in this subsection is that there is a similar problem with central political control. Let me lay out this claim in more detail.
A key problem for a peaceful and stable state is how to manage its citizenry such that they do not find their circumstances too intolerable. There are moral dimensions to this problem. For instance, a state may impose draconian police control of thought, expression, and so forth so as to keep the people in line. We may think that this will be unjust because it violates the rights of the citizenry. Or maybe a state might, as in Plato's Kallipolis, raise children so as to conform to their particular social class or occupation by certain means. We might think that this too is unjust, for it infringes upon the liberties that all people ought to possess by birthright.
But that's not the point that I'm making here. The point is not "what is the just way to manage a populace," but instead "what is the way to manage a populace so as to most efficiently take into account both the good of the nation as a whole and the desires of the individual in such a way that a state remains stable?" And it is here where the question of abstraction comes in.
I occupy many social positions. I am a son, a teacher, a driver, an American, a resident of Connecticut, and so forth. I may be classified, as a matter of statistical aggregation, according to any of these categories. But this will inevitably leave some things out -- perhaps quite important things! For instance, I may be classified as a driver, and hence the relevant details that go in the spreadsheet will concern my car's make and model, my driving record, my license number, and so forth. But this will inevitably leave out other details, ones which are central to my self-conception and hence my desires and goals. I think of myself more readily as an American than as a driver; that categorization is more relevant to my goals and desires than my being a driver.
Now, in any statistical aggregation, such details will be left out. And so if I live subject to what I'll call an allocatory state -- one that both has the power specifically to allocate to each citizen their role, their class, etc -- unless the allocators are provided with a sufficiently exhaustive list of my categorizations, together with information about which of these is most relevant for my self-conception and goals, they won't be able to slot me into the proper role. Now imagine performing that task for 300 million people.14
So if such allocation can't be left to the state as a whole, to whom may it be left? There are various levels of organization at which this might be done -- the province, the city, the neighborhood, the household -- but I think that the most plausible candidate is the individual level. All else being equal, each of us is a better judge of what our self-conception, goals, and desires are than anyone else is. The self-conception may be mistaken, the goals and desires misdirected, but no one else can think my thoughts. Hence, that knowledge is most readily available to me, and hence all else being equal I have more authority over what's the best way to fulfill them. And, all else being equal, a population left to fulfill their own goals and desires has a better shot at being stable in the long-term than that controlled by an allocatory state.
This goes, not only for individuals, but for localities as well. Certain sorts of communities will want to live in certain ways that might seem odd or even deviant to people not in that community, such that if they were forced en masse to adopt a different way of life they would revolt. Or, if not revolt, at least not flourish. I'm operating on the assumption that people flourishing, or at least being able to pursue activities that will allow them to do so, is all else equal a good thing. That, together that the people who all else equal know best what this flourishing is and how to accomplish it, at least in a rough way, are the people involved, generates a strong presumption in favor of community self-determination. This, in turn, generates a strong presumption in favor of a very strongly decentralized system of governance, as well as law.
2.2 -- Socal Time-Scale
Before getting started here I want to draw an important distinction. As a conservative, I believe that large societal changes may have relatively immediate effects, but they well might not. Usually, the only proper judge of the stability or wisdom of a given large-scale change is experience and time. We'll deal with this in more detail in a later essay.15
But when I talk about the time-scale of changes here, I don't mean the time-scale of large-scale social change. I mean the time-scale of variations in local matters of fact. Often these change quite quickly. Your wife gets pregnant, someone crashes into your car, you get laid off at work, and so on and so on. These often happen on the time-scale of minutes! And sometimes they are reversed within just as small a time interval.
Allocatory states are mostly a fiction these days. We are not quite in danger of our every move being dictated by the Marvelous Machine. So at the individual level, this is not quite a worry. It is a worry, however, on the level of the community or town. Suppose that you live in an agricultural town and a blight has hit your corn crop. How should you react? Suppose most townsfolk have a stockpile of corn that they could sell, while preserving some for a future crop, but that because of certain laws dating back to the Great Depression they couldn't legally sell these reserves.16 Should they follow the law? If they do, they run the risk of the federal government coming after them. But if the townspeople decline to sell these surpluses and let them rot or merely be taken, the town will suffer severe economic downturn.
This thought experiment shows that often large-scale laws can fail to be properly responsive to short time-scale changes, sometimes in ruinous ways. As a result, if a state is concerned with management of resources at short time-scales, all else equal there is a strong presumption towards strongly decentralized legislation. In other words, except for some broadly laws, the articulation of a body of enforceable law that is responsive to the multitude of short time-scale changes is best left to the smallest feasible level of social organization. Sometimes this will be the town, as in the case above. Sometimes it may even be at the neighborhood or family level. This might be balanced by bills of rights, generally applicable to all centers of law. But otherwise, the presumption is in favor of the local over the global.
2.3 -- Social Tacit Knowledge
But as above, even if these two problems were soluble by the Marvelous Machine, we would still be left with the fact that much of the information necessary to most efficiently attain a stable, peaceful state is out of our reach. If knowledge-that presupposes knowledge-how, and actions and reactions according to these tacit rules form a great amount of the body of data relevant to the calculations of the Machine, then even were the time-scale and the abstraction problem solved, the information base would still be incomplete. Just as Spinoza thought that we are "conscious of [our] actions and ignorant of the causes by which [we] are determined," (EIIp35s / C.I.473 / G.II.117) on the anti-intellectualist view we are ignorant of many of the rules according to which we act. We thus cannot articulate them in any useful way.
This poses an essential limit on the knowledge that any global body (such as a national legislature) can have about the circumstances attendant to their decisions. Essential to the writing of any legislation is knowledge of its probable effects. But if this cannot reliably be forecasted, because of limits on information accessible to the legislators, this too generates a strong presumption in favor of the devolution of legislative and enforcement powers to localities as small as feasible. Sometimes this will be a national legislature. Other times, it will be a county. Still others, it may be the individual level which is vested with such authority.
Section 3 -- The Big Picture
What vision of society emerges from these considerations? It's one of widespread devolution of the authority to legislate and enforce legislation to small-scale bodies. It bears significant resemblance to a Nozickean meta-utopia, in that much of the decision-making about forms of life and changes to them rests with individual communities, or perhaps even with individuals.17 Devolution in the abstract is nothing new. Many countries enjoy a federal system of government. But this is a much more radical proposal than that, in that the individual federations may be as small as a family or an individual, depending on the case.
I don't spell out here any specifics as to which decisions are left to which level. This is something best discovered, not by reasoning a priori, but by experimentation and experience generally. (I'll have more to say about this too in a later essay.) But it does specify a rule for the management of legislation and its enforcement: Whenever feasible, localize.
This has anarchist overtones, but is not itself anarchist. Or at least, not yet. The presentation above says nothing about whether the level of devolution must end at the individual or voluntary associations. (The association with a state is, practically, not voluntary.) More specific content would need to be given to make this into an anarchist vision. So I remain open to the idea that the proper level for all legislation and enforcement is that of the individual and voluntary associations. But, as I said above, this is discovered by experimentation and entering into these forms of life, not by some abstract reasoning about which rights people possess in a way detached from the forms of life into which people enter. Experience is the school of mankind, and he will learn in no other.
Hayek, Friedrich A. “The Use of Knowledge in Society.” The American Economic Review, vol. 35, no. 4, 1945. p. 520. Henceforth cited as UKS.
UKS, p. 521
UKS, p. 523
UKS, p. 523
Beer, Stafford. Brain of the Firm. 2nd ed., John Wiley & Sons, 1981. p. 257
Ironically, the greatest success of Cybersyn was to circumvent a labor strike; see Medina, Eden. Cybernetic Revolutionaries: Technology and Politics in Allende's Chile. The MIT Press. 2011. Chapter 5; Medina, Eden. “Designing Freedom, Regulating a Nation: Socialist Cybernetics in Allende’s Chile.” Journal of Latin American Studies, vol. 38. 2006. pp. 593-4
Medina 2006, pp. 603-4.
Beer 1981, p. 323.
UKS, p. 519.
Hayek, Friedrich A. “Rules, Perception, and Intelligibility.” Studies in Philosophy, Politics, and Economics, 1st ed., Taylor & Francis Books Ltd, New York, NY, 1967. p. 62.
Hayek, Friedrich A. The Sensory Order: an Inquiry into the Foundations of Theoretical Psychology. The University of Chicago Press, 1952. p. 39.
Hayek 1967, p. 61.
Except perhaps if, by coincidence, the designers were fortunate enough to create a neural network or some such which acted as if according to the same rules encoded in human psychology. But a problem can hardly be said to have a systematic solution if one may only solve it by sheer dumb luck.
There's a potential objection here: What if there's a statistical or algorithmic or other formal method which allows the allocators to disregard individual variations in favor of large-scale trends, psychohistory or the like? I won't deny that such technology is conceivable. I merely doubt its feasibility. Produce just such a statistical or formal technique, and I'll consider it
There are exceptions, or rather mitigating factors. While the abolition of transatlantic slavery might well have been accomplished in a gradualist way (according to, say, the plan outlined by Edmund Burke), as a moral matter it was too heinous a horror to be allowed to be phased out gradually. I'll also have more to say about how we tell which is which in a later essay.
These laws, and this situation, are fictional, but not a mere fiction. See Horne v. Department of Agriculture, here.
See Nozick, Robert. Anarchy, State, and Utopia, Basic Books, New York, 1974, chapter 10.