On the idea of Tolerable Outcomes (Epistemic Arguments for Conservatism V)

What does it mean for an institution to be associated with “tolerable” outcomes over a period of time? The question is more subtle than I thought at first; under prompting from a friend who commented on the paper I am writing, here’s a stab. (For an introduction to this series, see here; for all the other posts, click here; the idea of “tolerable” or "ok" outcomes is used here, here and here).

The first problem is to determine the sense in which we might say that some outcome (or some set of circumstances) is “tolerable.” One promising idea identifies tolerable outcomes with those outcomes that do not involve “clear evils.” By a “clear evil” I mean the sort of thing that all (reasonable?) individuals could identify as an evil: slavery, genocide, etc. (Though then, of course, we have the problem of sorting out the reasonable from the unreasonable; see here Estlund’s Democratic Authority). Some evils are not clear in this sense: reasonable individuals (in the Rawlsian sense of the term) might disagree about their importance, or their identification as an evil, given their (differing) beliefs about justice and the good.

A more problematic, but more substantive sense of “tolerable,” identifies tolerable outcomes with those outcomes that are above some threshold of badness on some substantive scale. Here the idea is not that some evils are necessarily clear in the sense discussed above, but that the determination of which evils are tolerable and which are not is an “easier” problem than the determination of which goods make a society optimal or fair or just, for example. Even if reasonable people disagree about whether, for example, persistent poverty is a tolerable evil, the conservative can still argue that determining whether persistent poverty is a tolerable evil is an “easy” problem relative to, for example, determining whether an egalitarian society is justified. (Perhaps the majority of people believe that poverty is a tolerable evil, while slavery is not; if we assume that the majority of people have some warrant for these beliefs, then the belief that persistent poverty is a tolerable evil might be epistemically justified, even if some reasonable individuals disagree). 

Taking some criterion of “tolerability” as given, a second problem emerges: institutions are associated with outcomes over time. Should a conservative discard any institution that is associated with even a single intolerable outcome? Or should the conservative somehow “average” these outcomes over time, or “discount” past outcomes at a specific rate?

For an example, consider the basic institutions of liberal democracy. If we look, say, at the institutions of the Netherlands or Sweden since 1960, we could easily agree that these institutions have been associated with tolerable outcomes since then, in the sense that they do not seem to have been associated (or produced, though by assumption we cannot tell whether outcomes associated with these institutions have been produced by them) with clear evils.

But now consider the entire history of relatively liberal institutions in the USA since the late 18^th century.  These institutions were not always associated with tolerable outcomes; they were in fact associated with slavery and ethnic cleansing, which count as clear evils if anything does, and with many other evils besides (aggressive war and colonialism among them). But at the time they were also not the same institutions as today; there has been a great deal of institutional change in the USA. Though the basic structure of the institutions, as specified in the US constitution, has not changed that much – e.g., we still have competitive elections, two legislative chambers with specific responsibilities, an executive, a relatively independent judiciary, a bill of rights, etc. – the actual workings of these institutions, the associated circumstances under which they operate, and the expectations that shape their use have changed quite a bit. Suffrage was extended to all adult males; then it was extended to women in the early 20^th century. Slavery was abolished. The regulatory powers of the Federal government expanded. The country industrialized. And so on. Since (by assumption) we do not know which aspects of American institutions and circumstances produced clear evils and which aspects and circumstances did not, we cannot in general answer the question of whether liberal institutions in the USA have produced tolerable outcomes in all past circumstances; at best, we can say that American institutions that are in some ways similar to existing institutions were associated with intolerable (not ok) outcomes in the past.

What might a conservative say to this? One possibility would be for the conservative to have a particular “discount rate” for the past: the further back in the past an outcome is associated with an institution, the less it is to “count” towards an evaluation of whether the institution is to be preserved, on the assumption that the further back in time we go, the less we are talking about the same institutions. Early nineteenth century American institutions were only superficially similar to modern American institutions, on this view; and so the outcomes associated with them should be discounted when we consider whether or not American institutions should have “epistemic authority.”

The problem with this is that, the smaller the discount rate is, the more intolerable outcomes it will “catch,” so that the conservative is forced to discard almost all institutions. With a small discount rate, the conservative is forced to argue that American institutions should not, in general be given the benefit of the doubt, since they (or similar enough institutions) have produced intolerable outcomes. But with a large discount rate, the conservative can be far less confident that the institutions  in question will be associated with tolerable outcomes in the future, since he has less evidence to go on. So the conservative faces a sort of evidence/discount rate tradeoff: the conservative position is most powerful, the more evidence we have of the association of institutions with tolerable outcomes; but the more evidence we have of outcomes, the more likely it is that some of these will be intolerable, forcing the conservative to argue for changes.

(In more formal terms: consider the series of states of the world {X₁,..X_i…X_n}, associated with the institution {I₁,…I_n}. For each X_i, we know whether it represents a tolerable or an intolerable outcome, and we know that it was associated with I_i, though we do not know whether I_i produced it. Suppose all intolerable outcomes are found in the past (i.e., in the series {X₁…X_i}, where i is less than n). Suppose also that our confidence that institution I_n (today’s incarnation of the institution) is similar enough to institution I_i decreases according to some discount rate d. The larger the d, the smaller the series of states that can serve as evidence that I_n will be associated with tolerable outcomes in the future; but the smaller the d, the more likely it is that the evidential series of states will include some states in the series {X₁…X_i}).

What do people think?

The Robustness or Resilience Argument in Practice: Noah Millmann vs. Jim Manzi (Epistemic Arguments for Conservatism IV.55)

Noah Millmann and Jim Manzi over at The American Scene (and Karl Smith at Modeled Behavior) have been debating the degree of deference we should give to economic science when considering what governments should do about a recession. Manzi emphasizes the large degree of uncertainty and difficulty attendant on any attempt to determine whether a particular policy actually works, and he is right about this: we do not know very well whether any policy intervention actually works (or worked), given the enormous number of potentially confounding variables. Lots of econometric ink is spilled trying to figure out this problem, but the problem is intrinsically hard, given the information available. By contrast, knowledge in physics or chemistry is far more certain, since it can be established by means of randomized experiments that are easily replicated. So, Manzi argues, we should give less deference to economists than we do to physicists when making decisions. Millmann sensibly points out that the relevant analogy is not to physics or chemistry but to something like medicine. The knowledge produced by medical science is hard to apply in practice, and doctors base their treatment decisions on a combination of customary practice, experience, and some limited experimental and observational evidence. In particular cases, then, medical practice offers at best an informed guess about the causes of a disease and the best course of action. But Millmann argues that this does not undermine the epistemic authority of medicine: in case of sickness, we should attend to the advice of doctors, and not to the advice of nonexperts.

I think Manzi’s argument would be more compelling if it were put as a robustness or resilience argument (discussed previously here and here). Consider first the case of medicine. If we get sick, we have three basic options for what to do: heed the advice of doctors, heed the advice of non-experts, and do nothing. It seems clear that heeding the advice of non-experts should (normally) be inferior to heeding the advice of doctors. But is heeding the advice of doctors always epistemically preferable to doing nothing? (Or, more realistically, to discounting the advice of doctors based on one’s own experience and information about one’s body). The answer to this question depends on our estimation of the potential costs of medical error vs doing nothing. Because medical knowledge is hard, doctors may sometimes come up (by mistake) with treatments that are actively harmful; in the 18^th century, for example, people used “bleeding” as a treatment for various diseases, which may have been appropriate for some things (apparently bleeding with leeches is used effectively for some problems), but probably served to weaken most sick people further. At any rate, we may not know whether a treatment works or not any better than the doctor; all we know is that people treated by doctors sometimes die. If our estimate of medical knowledge is sufficiently low (e.g., if we think that in some area of medical practice medical knowledge is severely limited), our estimate of the potential costs of medical error sufficiently high (we could die), and our experience of what happens when we do nothing sufficiently typical (most illness goes away on its own, after all: the human immune system is a fabulously powerful thing, perfected to a high degree by millions of years of evolution!) it may well be the case that we are better off discounting medical advice for the sake of doing nothing. Of course, atypical circumstances may result in us dying from lack of treatment; that is one of the perversities to which this sort of argument may give rise. But given our epistemic limitations (and the epistemic limitations of medicine), there may be circumstances where “doing something” is equivalent to doing something randomly (because the limitations on our medical knowledge are so severe), and so we may be (prospectively) better off doing nothing (i.e., tolerating some bad outcomes that we hope are temporary, since our bodies have proven to be resilient in the past).

Consider now the case of a government that is trying to decide on what to do with respect to a moderately severe recession. Here the government can do nothing (or rather, rely on common sense, tradition, custom and the like: i.e., do what non-experts would do), heed the advice of professional economists (who disagree about the optimal policy), or heed the advice of some selected non-economists (or the advice of some mixture of economists and non-economists). When is “heeding the advice of economists” better than “doing nothing,” given our epistemic limitations? And when is “heeding the advice of non-economists” better than “heeding the advice of economists”?

We know that the current architecture of the economic system produces recessions with some frequency, some of which seem amenable to treatment via monetary policy (whenever certain interest rates are not too close to zero), some of which appear to be less so (these are atypical), but in general produces long-run outcomes that seem tolerable (not fair, or right, or just: merely tolerable) for the majority of people (there are possible distributional concerns that I am ignoring: maybe the outcomes are not tolerable for some people). The system is robust for some threshold of outcomes and some unknown range of circumstances: it tends to be associated with increasing wealth over the long run, though it is also associated with certain bad outcomes, and we do not know if it is indefinitely sustainable into the future (due to environmental and other concerns). We also know that there is some disagreement among economists about what is the optimal policy in an atypical recession (which suggests that there are limits to their knowledge, if nothing else). If we think that the limits on economic knowledge are especially severe for some area of policy (e.g., what to do in atypical recessions), historical evidence suggests that sometimes economists may prescribe measures that are associated with intolerable outcomes (e.g., massive unemployment, hyperinflation, etc.), and we think that most recessions eventually go away on their own, we may be justified in doing nothing on epistemic grounds. In other words, if we think that for some area of policy economists’ guesses about optimal policy are not likely to be better than random, and carry a significant risk of producing intolerable outcomes, then conservatism about economic policy is justified (doing what custom, tradition, etc. recommend, and heavily discounting the advice of economists).

But these are big ifs. Suppose that the epistemic limitations of economic science are such that most policy interventions recommended by professional economists have a net effect of zero in the long run; that is, economists recommend things more or less randomly, some good, and some bad, but in general tend not to recommend things that are very bad for an economy (or very good for it). (Historical evidence may support this; “Gononomics” is something of an achievement, not necessarily something common). In that case, we are probably better off heeding the advice of economists (and gaining the experience of the results) than doing nothing (and not gaining this experience); there may not be exceedingly large costs from heeding economic advice, but there may not be very large benefits either, and the result will still be “tolerable.” (At the limit, this sort of argument suggests that we ought to be indifferent about almost any policy intervention, so long as we have reasonable expectations that the outcomes will still be tolerable). Moreover, distributional concerns may dominate in these circumstances; doing nothing has a distributional cost that is passed to some particular group of people (e.g., the unemployed), so we may have reason to be concerned more about distribution than about long-run economic performance. And much depends on our estimates of the epistemic limitations of economic science: sure, economics is not like physics, but is it more like 20^th century medicine, or more like 17^th century medicine? (And the answer to this question may be different for different areas – different for macroeconomics than for microeconomics, for example).

Epistemic Arguments for Conservatism IV.5: An Addendum on Resilience

Rereading the long post below, it occurred to me that I didn’t mention why the argument I describe there should be called a “resilience” argument. Here’s what I had in mind. Institutions that have lasted for a long time have presumably endured in diverse circumstances while still producing tolerable outcomes, so we may think that there is a reasonable probability that they will still do ok in many unknown future circumstances: their endurance can be taken as evidence of resilience. If the potential costs of error in trying to find the optimal set of institutions are very high (e.g., getting a really bad political system, like the mixture of feudalism and Stalinism they have in the DPRK), and the “optimal” set of institutions for a given set of circumstances is very hard to find (if, for example, nobody knows with any certainty what the optimal political system would be for that set of circumstances, and the system would have to be changed anyway as they change), then it would make sense to stick with institutions that are correlated with ok outcomes over long periods of time and tolerate their occasional inefficiencies and annoyances. Resilient institutions are better than optimal institutions, given our epistemic limitations.

The argument also seems to imply that we ought to be indifferent about different sets of “ok” institutions. For example, there are a variety of democratic institutions in use today: some countries have parliamentary forms of government, some presidential; some have bicameral legislatures, others unicameral; some have FPP electoral systems, others use MMP; some countries use rules mandating “constructive” no confidence votes, others use other rules. But though we have some (statistically not especially good) evidence that some of these combinations work better than others (in some sets of circumstances: e.g., unitary parliamentary systems with list PR seem to produce better long-run outcomes than federal presidential government with nonproportional systems, at least on average, though I would not put too much stress on this finding), for the most part they all work ok, and we cannot tell with reasonable certainty whether some particular combination would be much better for us given foreseeable (and unforeseeable) circumstances. Perhaps switching back to FPP in New Zealand, for example, would produce better economic performance or induce better protection of civil liberties, but the best estimate of the effect of switching to FPP (or retaining MMP) on long run economic performance or the average level of protection of civil liberties is basically zero. (We might have reason to retain MMP or switch to FPP, but these will probably have more to do with normative concerns about representation and ideas about how easy it is for citizens to punish a government they dislike for whatever reason than with any special ability of MMP to deliver better economic performance). So we should not be bothered overmuch about these details of institutional design; given our epistemic limitations, on this view, it is unlikely that we would achieve even marginal improvements in our institutions that are sustained over the long run.

This does assume that gradual tinkering cannot at least serve to mitigate the effects of a changing “fitness landscape” (to use the terminology of the previous post), a controversial assumption. (It might be better to constantly tinker with our institutions than to let them just be, even if the tinkering is unlikely to lead to sustained improvements: we are just trying to stay on a local peak of the fitness landscape). And it also assumes that this landscape is very rugged for all the heuristics available: either minor changes just take you to another set of "ok" institutions (another variety of democracy, with some other combination of electoral system, relationships between executive and legislative powers, veto points, etc., and producing basically the same average long-run benefits), or they mostly throw you down a deep chasm if you try something new and radical (you get communist feudalism, or some variety of kleptocracy, and so on). I'm not sure this assumption makes sense for most problem domains, however: perhaps gradual tinkering in some cases does lead to better long-run outcomes, pace my previous argument against gradualism. But I have to think some more about this problem.

Epistemic Arguments for Conservatism IV: The Resilience Argument and the “Not Dead Yet” Criterion

(Fourth in the series promised here. Usual disclaimers apply: this is work in progress and so it is still a bit muddled, though comments are welcome).

One of the more promising epistemic arguments for conservatism is the argument from resilience. The general idea is that we owe deference to certain institutions (and so should not change them) not because they are “optimal” for the circumstances in which we find ourselves, but because they have survived the test of time in a variety of circumstances without killing us or otherwise making us worse off than most relevant alternatives. This argument might be used, for example, to justify constitutional immobility in the USA: even if the US constitution is not optimal for every imaginable circumstance, it is tolerable in most (“we’re not dead yet”); after all, it has lasted more than 200 years with relatively minor changes to its basic structure (save for the treatment of slavery, of course; but let us focus only on the basic structure of constitutional government); and if we have no good reason to think that changes to the constitution would improve it (because the effects of any change are exceedingly difficult to predict, and would interact in very complicated ways with all sorts of other factors, a caveat that would not necessarily apply to the treatment of slavery in the original constitutional text, which we may take as an obvious wrong), and some reason to think that the costs of ill-advised changes would be large (“we could die,” or at the very least unchain a dynamic leading to tyranny, oppression, and economic collapse), we are better off not changing it at all and putting up with its occasional inefficiencies.

The oldest and in some ways the most powerful version of this argument can be found in Plato’s Statesman (from around 292b to 302b). There the Eleatic Stranger (the main character in this dialogue) argues for a very strict form of legal conservatism, suggesting that we owe nearly absolute deference to current legal rules in the absence of genuine political experts who have the necessary knowledge to change them for good. This might seem extreme (indeed, it has seemed extreme to many interpreters), but given the assumptions the Stranger makes, the argument seems rather compelling. 

The basic logic is as follows. (For those interested in a “chapter and verse” interpretation of the relevant passages, see my paper here [ungated], especially the second half; it’s my attempt to make sense of Platonic conservatism.) In a changing environment, policy has to constantly adjust to circumstances; the optimal policy is extremely “nonconservative.” But perfect adjustment would require knowledge (both empirical and normative) that we don’t have. In Platonic terminology, you would need a genuine statesman with very good (if not perfect) knowledge of the forms of order (the just, the noble, and the good) and very good (if not perfect) knowledge of how specific interventions cause desired outcomes; in modern terminology, you would need much better social science than we actually have and a much higher degree of confidence in the rightness of our normative judgments than the “burdens of judgment” warrant. Worse, in general we cannot distinguish the people who have the necessary knowledge from those who do not; if unscrupulous power-hungry and ignorant sophists can always mimic the appearance of genuine statesmen, then the problem of selecting the right leaders (those who actually know how to adjust the policy and are properly motivated to do so) is as hard as the problem of determining the appropriate policy for changing circumstances.

If the first best option of policy perfectly tailored to circumstances is impossible, then (the Eleatic Stranger argues) the second best option is to find those policies that were correlated in the past with relative success according to some clear and widely shared criterion (the “not dead yet” or “could be worse” criterion), and stick to them. Note that the idea is not that these policies are right or optimal because they have survived the test of time (in contrast to some modern Hayek-type “selection” arguments for conservatism), or even that we know if or why they “work” (in the sense that they haven’t killed us yet). On the contrary, the Stranger actually assumes that these policies are wrong (inefficient, non-optimal, unjust); they are just not wrong enough to kill us yet (or, more precisely, not wrong enough for us to bear the risk of trying something different), even if we happen to live in the highly competitive environment of fourth century Greece. (The true right policy can only be known to the possessor of genuine knowledge; but ex hypothesi there is no such person, or s/he cannot be identified). And he also assumes that we do not know if the reason we are not dead yet has to do with these past policies; correlation is not causation, and the Stranger is very clear that by sticking to past policies we run large risks if circumstances change enough to render them dangerous. But the alternative, in his view, is not a world in which we can simply figure out which policies would work as circumstances change with some degree of confidence, but rather a world in which proposals are randomly made without any knowledge at all of whether they would work or not, and where the costs of getting the wrong policy are potentially very high (including potentially state death). If these conditions hold, sticking to policies that were correlated with relative “success” (by the “not dead yet” or “could be worse” criterion) is then rational. (There are some complications; the Stranger’s position is not as absolute as I’m making it seem here, as I describe in my paper, and Plato’s final position seems to be that you can update policy on the basis of observing sufficient correlation between policies and reasonable levels of flourishing and survival even in the absence of perfect knowledge).

Does this argument work? In order to understand the circumstances under which it might work, let us recast the argument in the terminology of a “fitness landscape.” Let us assume that, in some problem domain, we have some good reason to believe that the “fitness landscape” of potential solutions (potential policies) has many deep valleys (bad policies with large costs), some local but not very high “peaks” (ok policies) and only one very high peak (optimal policies). Assume further that this “fitness landscape” is changing, sometimes slowly, sometimes quickly; a reasonably ok policy in some set of circumstances may not remain reasonably ok in others. Under these circumstances, an agent stuck in one of the local peaks has very little reason to optimize and lots of reason to stick to their current policy if it has reason to think that its heuristics for traversing the fitness landscape are not powerful enough to consistently avoid the “deeps.” Conservatism is then rational, unless your local “peak” starts to sink below some acceptable threshold of fitness (in which case you may be dead whether or not you stick to the policy).

For a concrete example, consider the space of possible political systems. The vast majority of imaginable political systems may be correlated with some very bad outcomes, sometimes very bad –oppression, economic collapse, slavery, loss of political independence, even physical death. A smaller set – including liberal democratic systems, but potentially including other systems, are reasonably ok; they are correlated with a measure of stability and other good things, though (let us assume) we have no good way to know if they actually cause those good outcomes or if the correlation occurs by chance, and we have no reason to assume that these are the best possible outcomes that can be achieved, or that these good outcomes will be forever associated with these political systems. Finally, let us assume that there exists some utopian political system which would induce the best possible outcomes (however defined) for current circumstances (imagine, for the sake of argument, that this is some form of communism that had solved the calculation problem and the democracy problem plaguing “real existing communism”), but that we do not have enough knowledge (neither our social science or our theory of justice is advanced enough) to describe it with any certainty. Does it make sense to try to optimize, i.e., to attempt to find and implement the best political system, in these circumstances? I would think not; at best, we may be justified in tinkering a bit around the edges. Both the uncertainty about which political system is best and the potential costs of error are enormous, and circumstances change too quickly for the “best” system to be easily identified via exhaustive search. Hence conservatism about the basic liberal democratic institutions might be justified. (Note that this does not necessarily apply to specific laws or policies: here the costs are not nearly as large, and the uncertainty about the optimal policy might be smaller, or our heuristics more powerful. So constitutional conservatism is compatible with non-conservatism about non-constitutional policies).  

On the other hand, it is important to stress that the “not dead yet” criterion is compatible with slow death or sudden destruction, and somehow seems highly unsatisfactory as a justification for conservatism in many cases. Consider a couple of real-life examples. First, take the case of a population in the island of St Kilda, off the coast of Scotland, described in Russell Hardin’s book How do you Know? The Economics of Ordinary Knowledge. According to Hardin, this population collapsed over the course of the 19^th century in great part due to a strange norm of infant care:

It is believed that a mixture of Fulmar oil and dung was spread on the wound where the umbilical cord was cut loose. The infants commonly died of tetanus soon afterwards. The first known tetanus death was in 1798, the last in 1891. Around the middle of the nineteenth century, eight of every tenth children died of tetanus. By the time this perverse pragmatic norm was understood and antiseptic practices were introduced, the population could not recover from the loss of its children (p. 115, citing McLean 1980, pp. 121-124)

Though this norm was bound to decimate the population eventually, it worked its malign power over the course of a whole century, slowly enough that it may have been hard to connect the norm with the results. And, perversely, it seems that the conservatism of the St Kilda’s was perfectly rational by the argument above: a population that has that kind of infant mortality rate is probably well advised not to try anything that might push them over the edge even quicker, especially as they had no rational basis to think that the Fulmar oil mixed with dung was the root cause of their troubles (rather than, for example, the judgment of god or something of the sort).

Or consider the example of the Norse settlers of Greenland described in Jared Diamond’s Collapse. Living in a tough place to begin with, they were reluctant to change their diet or pastoral practices as the climate turned colder and their livelihood turned ever more precarious, despite having some awareness of alternative practices that could have helped them (the fishing practices of the Inuit native peoples, for example). So they eventually starved and died out. Yet their conservatism was not irrational: given their tough ecological circumstances, changes in subsistence routines were as likely to have proved fatal to them as not, and they could have little certainty that alternative practices would work for them. (Though it is worth noting that part of the problem here was less epistemic than cultural: the Greenland Norse probably defined themselves against the Inuit, and hence could not easily learn from them).

In sum, the resilience argument for conservatism seems most likely to “work” when we are very uncertain about which policies would constitute an improvement on our current circumstances; the potential costs of error are large (we have reason to think that the distribution of risk is “fat tailed” on the “bad” side, to use the economic jargon); and current policies have survived previous changes in circumstances well enough (for appropriate values of “enough”). This does not ensure that such policies are “optimal”; only that they are correlated with not being dead yet (even if we cannot be sure that they caused our survival). And in some circumstances, that seems like a remarkable achievement. 

Epistemic Arguments for Conservatism III: Computational Arguments

(Another one in the series promised here; I’m writing a paper. This is still somewhat muddled, so read at your own risk, though comments are greatly appreciated if you find this of interest.)

Many problems of social life have “solutions” that can be correct or incorrect. Determining whether someone is guilty or innocent of a violation of legal rules; allocating goods and services to their best uses; policing a neighbourhood so that potential violators of community norms are deterred; all of these things can be characterized as problems with solutions that can be said to be at least partially correct or incorrect, assuming there is broad agreement on the values a solution should promote. Different institutional solutions to these problems can thus be evaluated with respect to their epistemic power: the degree to which, on average, a given institution is able to reach the “correct answer” to the problem in question. A “computational” argument for a particular institutional solution to a problem is simply an argument that, given the incentives the institution provides for gathering or revealing information or knowledge important to the problem, the average capacities of the relevant decisionmakers to process this information or use this knowledge, and the rules or mechanisms it uses for aggregating their judgments, the institution in question has greater epistemic power than the alternatives (relative, of course, to a particular problem domain).

Consider, for example, the institution of trial by jury. Though jury trials have more than one justification, their epistemic power to determine violations of legal norms is often invoked as an argument for their use vis à vis, say, trials by ordeal or combat. A trial by jury is expected to be a better “truth tracker” than trials by ordeal or trials by combat in the sense that it is expected to better discriminate between violators and non-violators of legal norms (more precisely, it is expected to better identify non-violators, even at the expense of failing to identify violators) because, among other things, it may provide good incentives for the revelation of both inculpatory and exculpatory evidence via adversarial proceedings and may allow the judgments of individual jurors to be combined in ways that amplify their epistemic power (see, e.g., Condorcet’s jury theorem). By contrast, trials by ordeal or combat are supposed to be lousy discriminators between violators and non-violators of legal norms (or lousy ways of identifying non-violators), because they provide bad incentives for the revelation of the relevant information (though see Peter Leeson’s work for some interesting ideas on how trials by ordeal might have exploited widespread beliefs in “god’s judgment” to discriminate accurately between the guilty and the innocent). To be sure, even if the computational argument for jury trials is correct, we may still not want to use them to determine every violation of legal norms: considerations of cost, fairness, lack of suitable jurors, or speed, may reasonably limit their use. But the epistemic power of jury trials would surely be one important consideration for using them rather than other mechanisms in trying to figure out whether a particular person has violated a legal norm.

Now, the idea that some institutions are better than others at “social cognition” or “social information processing” is not  inherently conservative, as the example of jury trials indicates. “Computational” or “social cognition” arguments have been deployed in defence of a wide variety of institutions, from democracy to Wikipedia, and from markets to the common law, without necessarily bolstering a “conservative” position in politics, however conceived. (For a good discussion of the concepts of social cognition and social information processing, as well as a review of some of the research that attempts to untangle when and how social cognition is possible, see this short paper by Cosma Shalizi). But there is a set of arguments for “conservatism,” broadly understood, that argues for the epistemic power of some “default” solution to a social problem and against the epistemic power of an “explicit” intervention on computational grounds. The same contrast is sometimes expressed differently – e.g., in terms of decentralized vs. centralized institutions, “unplanned” vs. “planned” social interaction, or customary vs. explicit rules – but it always indicates something like the idea that some institutional solutions to a problem need not be explicitly produced by a single, identifiable agent like a government. A computational argument for conservatism thus makes the (implicit or explicit) claim that we can “conserve” on reason by relying on the computational services of such patterns of interaction or institutions to determine the solution to a problem of social life rather than attempting to explicitly compute the solution ourselves.

This can get confusing, for it is not always clear what would count as a “default” solution to a social problem, and “restoring” (or even implementing) the default solution may entail robust changes to a social system (amounting even to revolution in some cases). So bear with me while I engage in a classificatory exercise. Three options for defining the default seem possible: the pattern of interaction that has historically been the case (“custom” or “tradition”); the pattern of interaction that would have prevailed in the absence of explicit planning or design by one or more powerful actors (e.g., “the free market” as opposed to a system of economic allocation involving various degrees of centralized planning); and the pattern of interaction that “computes” the solution to the social problem implicitly rather than explicitly (compare, for example, a “cap and trade” market for carbon emissions with an administrative regulation setting a price for carbon emissions: both are designed, but the former computes the solution to the problem of the appropriate price for carbon implicitly rather than explicitly). We might call these options the “Burkean,” “Hayekian,” and (for lack of a better word) “Neoliberal” understandings of the relevant “default” pattern of interaction, which in turn define epistemic arguments for, respectively, the superiority of tradition over innovation, the superiority of spontaneous order over planned social orders, and the superiority of implicit systems of “parallel” social computation over explicit centralized systems of social computation. But what reasons do we have to think that any of these “default” patterns of interaction have greater epistemic power than the relevant alternatives, or rather, under what conditions are they computationally better than the relevant alternatives? 

Let us start with the last (“Neoliberal”) position, since it seems to me the easiest to analyze and at any rate is the farthest from conservatism in the usual sense of the term (the “Burkean” position is the closest to conservatism, while the “Hayekian” sits more or less in the middle; I'm leaving the analysis of "Burkean" arguments to another post). Here the relevant comparison is between two designed institutional solutions to a problem, one that aims to determine the solution to a social problem by setting the rules of interaction and letting the solution emerge from the interaction itself, and another that aims to induce a set of actors to consciously and intentionally produce the solution to the problem. Thus, for example, a “cap and trade” market in carbon emissions aims ultimately to efficiently allocate resources in an economy on the assumption that the economy should emit less than X amount of carbon into the atmosphere, but it does so by setting a cap on the amount of carbon that may be produced by all actors in the market and letting actors trade with one another, not by asking a set of people to directly calculate what the best allocation of resources would be (or even by directly setting the price of carbon). We might compare this solution to a sort of parallel computation: given a target amount of carbon emissions, the relevant computation concerning the proper allocation of resources is to be carried out in a decentralized fashion by economic actors in possession of private and sometimes difficult to articulate knowledge about their needs, production processes, and the like, who communicate with one another the essential information necessary to coordinate their allocation plans via the messaging system of “prices.” 

This sort of pattern of interaction will be computationally superior to a centralized computation of the solution to the same problem whenever the relevant knowledge and information to determine the solution is dispersed, poorly articulated, time-sensitive, and expensive or otherwise difficult to gather by centralized bodies (perhaps because actors have incentives not to truthfully disclose such information to centralized bodies), yet essential features of such knowledge are nevertheless communicable to other actors via decentralized and asynchronous message passing (like prices). (Hayek’s famous argument for markets and against central planning basically boils down to a similar claim). The problem can thus be decomposed into separate tasks that individual actors can easily solve on their own while providing enough information to other actors within appropriate time frames so that an overall solution can emerge.

But these conditions do not always hold. Consider, for example, the problem of designing an appropriate “cap and trade” market. Here the relevant knowledge is not dispersed, poorly articulated, and time-sensitive but is instead highly specialized and articulated (e.g., knowledge of “mechanism design” or “auction theory” in economics), is not as obviously time-sensitive, and cannot easily be divided. (Though the problem of discovering the truth about mechanism design or auctions might itself be best tackled in a decentralized manner).  We might perhaps learn here from computer science proper: some problems can be tackled by easily “parallelized” algorithms (algorithms that can be broken down into little tasks that can run in a decentralized fashion in thousands of different computers), but some problems cannot (the best available algorithm needs to run in a single processor, or the problem can only be broken down into steps that need to run sequentially, like the algorithms for calculating pi); in fact there seems to be an entire research programme trying to figure out which classes of problems can be parallelized and which cannot. (And this seems to be a deep and difficult question). Or we might speak here of “epistemic bottlenecks” that limit the degree to which a problem can be broken down into tasks that can be solved via a division of epistemic labor; the problem of designing an appropriate division of epistemic labor for a specific purpose might be one of these.

The “computational” argument for implicit over explicit computation depends on the identification of an epistemic bottleneck in explicit mechanisms of computation that are not present in the implicit mechanism. But it does not depend on a contrast between designed and undesigned solutions to a problem: both a carbon market and an administrative regulation are equally designed solutions to the same problem. In order to make the computational case for spontaneous order (as against “planned” order), one has to argue not only that there are epistemic bottlenecks in the explicit mechanism of computation, but that the problem of designing an order for computing the solution to the problem is itself subject to the epistemic bottlenecks that render explicit solutions unfeasible; and here, I am not sure that Hayek or anyone else has given a convincing argument yet. (One could, of course, give “selection” arguments for preferring spontaneous to designed orders; but that is a subject for another post).

Changes small and large: Epistemic Arguments for Conservatism II

One epistemic argument for conservatism (or rather, gradualism) goes more or less like this. We have grave epistemic limitations that prevent us from understanding the consequences of institutional change, and the bigger the change, the more these limitations apply. All things considered, it is harder to predict the consequences of major institutional change than of minor institutional tinkering. Moreover, even when we are in error about the consequences of minor change, the potential damage to our social life from such errors is also likely to be minor and easily contained, whereas when we are in error about major changes, the damages may be devastating. Hence we should try to avoid wholesale change to our institutions.

The key idea is that ex ante we have a better idea of the risks of small changes than we do of the risks of large changes, and ex post we can contain the damage of small changes better than the damage of large changes. Given reasonable loss aversion in the face of genuine uncertainty, this implies that we should be more careful about larger than about smaller changes to our institutions. Note that the argument is not that major changes will have bad effects, but that we should be more uncertain about the effects of major changes than about the effects of minor changes, and that this asymmetric uncertainty should make us more cautious about undertaking these larger changes.

There is a whole question here about how to define “small” and “large,” but let’s let that slide for the moment, since there are clearly some cases where the argument seems to make sense. Something like it forms part of Burke’s case against revolution: there are many potential bad consequences from trying to completely change an entire form of life, but given our epistemic limitations we can hardly know which major changes will be on balance good and which will be on balance bad. And today we might see this argument deployed to justify less action than some economists might like on restarting the global economy, or less action on climate change than some environmentalists might like, though whether the argument works in these cases depends in part on our estimates of the costs of inaction, estimates that may also be subject to our epistemic limitations. (A full formal investigation of the problems here would presumably use Bayesian analysis. But here I reveal my on epistemic limitations – I do not know enough to use it).

This is a sort of “selection” argument for conservatism, though it does not look like one at first sight. In order for this argument to justify gradualism, however, rather than rigid immobility, we have to assume that small deleterious changes are either eliminated quickly or else that they cannot accumulate and interact in collectively very harmful ways. For example, a firm that produces bad products should go bankrupt without affecting the ability of other firms to operate too much, and small regulatory changes that do not work must be quickly identified and eliminated through some explicit mechanism (e.g., litigation, as in some defences of the common law). Whenever there is no selection mechanism to get rid of such minor but potentially harmful changes (and there may not be one, especially if the changes are only truly harmful in connection with many other changes), they can accumulate until they reach a kind of threshold (when they become really bad). The point is that if our epistemic limitations are binding with respect to the consequences of large “revolutionary” changes, they are also binding with respect to the consequences of collections of small changes that interact in complex ways with one another. Thus, in the absence of knowledge about the interactions of many small changes, the argument would seem to justify rigid immobility, not gradualism.

Indeed, an awareness of this problem seems to have led most classical political thinkers to a position that is in a sense the reverse of the Burkean “gradualist” position, i.e., willing to contemplate major changes to institutions under some limited circumstances (when the legislator has good knowledge) but extremely wary of small changes that might accumulate and interact in unexpected ways (since the presence of knowledgeable individuals capable of understanding their effects in the long run cannot be assumed). Plato, for example, while quite willing to recommend radical institutional innovation in cities (e.g., equal education for men and women, major censorship, new religious institutions, etc.) insisted that a city must be very careful about changing its “educational” laws, in part because the deleterious effects of any minor changes would not be visible until many years later, and they could easily accumulate in destructive ways. Similarly, Aristotle was not shy about recommending large institutional changes to political communities, but nevertheless urged cities to be vigilant about institutional drift, the minor changes that seem harmless individually but may collectively lead to the destruction of a politeia.

The point is not that they believed that major change was epistemically “safe” while gradual change was not; they thought both major and minor change was epistemically problematic. (Plato’s insistence on the need for knowledge in politics is grounded precisely on an extreme distrust of normal human epistemic capacities; we would only need philosopher kings if we were effectively incompetent in political matters). But they did not think gradual change was any less exempt from the problems caused by human epistemic limitations than revolutionary change, since they did not think that there were appropriate “selection mechanisms” able to weed out deleterious gradual institutional change under normal circumstances (other than “state death,” which is of course not especially ideal from the point of view of the state in question; as Josiah Ober has noted, whereas bankruptcy may be the consequence of “failure” in a market economy, “state death” in the classical world typically implied the death or enslavement of much of the population). From this point of view, major but planned institutional change had a slight “epistemic advantage” over unplanned, gradual change, at least so long as gradual changes could accumulate and interact in deleterious ways that could not be easily identified either at the time or in advance. 

Epistemic Deference and Epistemic Arguments for Conservatism I

A couple of months ago, there was a very interesting debate between Jim Manzi and others on the question of whether social science knowledge was sufficiently well established to warrant radical policy innovations. Manzi argued in a widely blogged article that the methodological limitations of the social sciences – in particular, the impossibility of appropriately measuring the counterfactuals of policy interventions through, for example, randomized experiments – implied that the findings of social science should normally not “outweigh settled practice and even common sense” when we are thinking about changing existing practices or institutions. His critics argued, among other things, that social science can provide evidence that overturns such settled practices, since, after all, the status quo is typically the result of power struggles, not rational discussion; why should we give it more credence than the findings of social science, which emerge from a process that at least approximates rational debate?

We might put the general point thus: Manzi thinks that, where the complexity of social reality makes full scientific knowledge all but impossible, it makes better sense to give “epistemic deference” to settled practice; his critics think that the practices of “social science” as a whole (not necessarily a single random study) often have more “epistemic authority” than settled practice. This sort of division about the locus of epistemic authority also maps into a political division between “conservative” and “non-conservative” positions, with the “conservatives” granting some settled practices or institutions the bulk of epistemic authority, while the “non-conservative” position insists that some other person or institution, usually of more recent vintage, should be given greater epistemic deference than settled practice. Is there any way of adjudicating under what conditions are these positions likely to be right?

The conservative position is typically defended by arguments about the “limits of reason.” The classic statement is in Burke’s Reflections on the Revolution in France:

We are afraid to put men to live and trade each on his own private stock of reason, because we suspect that this stock in each man is small, and that the individuals would do better to avail themselves of the general bank and capital of nations and of ages. Many of our men of speculation, instead of exploding general prejudices, employ their sagacity to discover the latent wisdom which prevails in them. If they find what they seek, and they seldom fail, they think it more wise to continue the prejudice, with the reason involved, than to cast away the coat of prejudice and to leave nothing but the naked reason; because prejudice, with its reason, has a motive to give action to that reason, and an affection which will give it permanence. Prejudice is of ready application in the emergency; it previously engages the mind in a steady course of wisdom and virtue and does not leave the man hesitating in the moment of decision skeptical, puzzled, and unresolved. Prejudice renders a man's virtue his habit, and not a series of unconnected acts. Through just prejudice, his duty becomes a part of his nature.

As Burke puts it, the comparison makes sense; individual reason (or, for that matter, the individual social science study) is highly limited in its epistemic power in comparison to settled social practice. There is typically some reasonable basis for even highly perplexing social practices; and individual reason is likely to be highly misleading in many circumstances. Individually, we suffer from so many cognitive biases and defects that it is a wonder we get up in the morning; and even highly trained experts are often wrong, even in their own fields. (I read somewhere that most peer-reviewed research is eventually overturned). 


But as an argument the comparison is flawed; the relevant comparison should not be that between settled practice and individual reason, but between settled practice and some alternative social practice (e.g., the social practice of science, with its various self-correction mechanisms), or between settled practice and some other collective judgment (e.g., the collective judgment of an assembly or a market). Moreover, the “general bank and capital of ages” is often a “depreciating asset,” as Adrian Vermeule puts it, especially insofar as later people make no contributions to it. From this point of view, Burkean concerns about the limits of individual reason cut much less deeply; the question is about the relative “epistemic power” of different social practices, institutions, or processes. 

Yet I think there is often something to these “epistemic” arguments for conservatism. In later posts, I want to try to investigate under what conditions epistemic arguments for conservatism make sense. (This blog is mostly my scratch research notes, so read at your own risk). As a starting point, I think we might divide these arguments into three kinds: selection arguments (where the epistemic power of settled social practices comes from the operation of a selection filter that is not found in alternative social practices), computational arguments (where the epistemic power of settled social practices comes from the way in which they organize expertise and information), and resilience arguments (where settled social practices are found to be preferable to alternatives because they are adapted to a wide variety of circumstances, even if not optimal for any particular one of them). (Resilience arguments are really a subtype of selection arguments, but I want to think about them separately). More on each of these later.


[update 2:35: edited a few words for clarity].

Footnotes on things I've been reading: Adrian Vermeule's "Law and The Limits of Reason"

Adrian Vermeule, Law and the Limits of Reason (2009).

I read this book for somewhat hazy reasons; perhaps I found the title in a footnote of some paper and it seemed interesting, perhaps I ran across it by doing a search in our library catalog for something else and remembered that someone had once recommended  to me another one of Vermeule’s books. But at any rate, I’m glad I did read it, even though the main topic of the book is something that would normally not interest me very much.

The book is basically an attack on the “constitutional” common law, written primarily for (USA) legal scholars. As I reconstruct it, the basic problem is as follows. 200 and some years ago a few very smart people wrote a constitution for a country. (We’ll leave the name of the country blank, since the problem is pretty general; and I want to move away from any particularities of the American situation). Conditions at the time were very different from conditions today, however; so the problem arises as to whether, and how, to adapt the constitution to current circumstances. Should we change the constitution, and if so, who should change it?

At one extreme we can find the pure originalist position: the constitution should never be changed by anyone. This position is not as crazy as it may look at first glance. Assume that the constitution has survived intact for 200 years (and it has not been ignored during this period). It may then be that the very fact that the constitution has enabled the country to survive for 200 years implies that current constitutional rules are “resilient” – that is, they are broadly adapted to all kinds of circumstances, even if not optimal for any of them. They have been tested by time, with its varied challenges, and come out relatively well. Though different constitutional rules might have produced better outcomes at some points during those 200 years, no other set of rules we could design today would have plausibly produced a better outcome overall. We could perhaps temporarily achieve better outcomes (however defined) by adjusting the rules today, but then we could not be sure that the new rules would be optimal for other, as yet unforeseen circumstances, and we would have to adjust them again and again. And given some assumptions about the likelihood and costs of error, the relative wisdom of the framers versus our current wisdom, and our estimates of the rate of environmental change, it might be that it would never be worth our while to try to change these constitutional rules. We can think of this argument in terms of a trade-off between the benefits of adaptation to current circumstances and the benefits of resilience in changing conditions, given the limits of reason. If our estimates of the benefits of adaptation are low enough, of the resilience of current rules high enough, and of the uncertainty involved in formulating “perfect” rules for current circumstances high enough, perfect conservatism is entirely reasonable. (The earliest version of this argument I know of goes back to Plato’s Statesman, which I discuss extensively here. We also see something like this trade-off in nature: an organism can be too well adapted to its environment to survive drastic changes there. “Generalists” may thrive in changing circumstances, while “specialists” die since they cannot adapt quickly enough to these changes).

But suppose that we do not agree with the pure originalist position. (Few people do). The framers may have been very wise, but they did not foresee everything; and we may have a clearly good idea about how to improve existing constitutional rules (e.g., emancipating the slaves). But the problem still remains, since not every potential change is good: when should constitutional rules be changed, and by whom? The constitutional common lawyers argue that the constitution should (normally) not be changed by legislative assemblies or the “explicit” amendment mechanisms indicated in the constitution itself; instead, it should be changed gradually via the decisions of judges, just as proponents of the “regular” common law argue more generally that law should be changed not via legislative codification but rather through the gradual decisions of judges. Common-law constitutionalism thus tends to be associated with arguments for robust forms of judicial review and a general conservatism on matters of constitutional change, the key assumption being that the limits of reason are less binding on judges than on legislatures or the public, though still quite important.

Vermeule subjects this position to withering (though sometimes repetitive) scrutiny, arguing instead for a position that might be called “Benthamite” or “Thayerian” (as he labels it). On this view, legislatures face fewer epistemic constraints than courts in deciding to adapt constitutional rules to current circumstances, and hence courts should in general defer to them. (Note that this argument does not say much about incentives: it may be that legislatures face fewer epistemic constraints than judges in constitutional matters, but have worse incentives, and it is not clear which constraints would be “binding” in that situation). I am less interested in his ultimate conclusions, however, which are a bit too US-centric (including a plea for including at least one non-lawyer in the Supreme Court for interesting reasons having to do with the virtues of diversity in deliberative groups), than on his general approach to the problem of the limits of reason in politics and law, which is all about trying to figure out how "collective" mechanisms affect those limits.

Vermeule identifies two basic “epistemic” arguments for the constitutional common law position. On one view, the constitutional common law is better than the constitutional law created by legislatures or other explicit mechanisms because it combines the judgments of many judges over time to produce a judgment that is somehow better than the judgments of any individual judge (for reasons having to do with the Condorcet jury theorem, for example, though other mechanisms might be imagined). We can call this the “aggregation” argument. On the second view, which is associated with Hayek and his followers, the constitutional common law is better than the “explicit” constitutional law of legislatures because the process through which it is produced includes a selection “filter” that weeds out bad rules (the litigation and appeals process, for example) in ways that are not possible for explicitly created constitutional rules. We can call this the “selection” argument.

I am not going to discuss in detail all the considerations Vermeule deploys against these arguments. At any rate, it seems to me that the book is written very much like a lawyer’s brief: the constitutional common-law position is attacked from every possible perspective, some of which are less convincing than others. But a couple of points he makes are quite interesting. First, he argues that the “aggregation” arguments tend to suffer from a “nirvana” fallacy, where the benefits of aggregation are assumed to apply only to judge-made law. But in fact laws that are made by legislatures with many people should also benefit from considerations about aggregation (at least assuming equal incentives for information acquisition and sincere voting, which Vermeule discusses in some detail, though not always satisfactorily). And the kind of aggregation over time postulated by common-law constitutionalists may also suffer from informational cascades and epistemic “free-riding”: judges that defer to past decisions at a later time do not add any new information to the law, leaving the first few judges as the real legislators, for example. (“Epistemic free-riding” may also be a problem in legislatures, especially if the legislature is large, since then incentives for information acquisition are smaller. But this is more or less balanced – under some conditions, at least – by the greater epistemic diversity of legislatures, which should improve the quality of their decisions). Second, drawing on some interesting-looking work by Andrei Schleifer, he suggests that there is no plausible “selection” mechanism for the constitutional common law (though he is agnostic on the regular common law) that ensures bad rules get weeded out. Taken together, he is basically arguing that the limits of reason bind judges at least as much, if not more than, as they bind legislators; there are no special advantages to common-law constitutionalism over other forms of constitutionalism, and many apparent disadvantages.

Anyway, this is probably not a work for everyone (even if it is not particularly difficult to read); but I’m rather interested in “epistemic” arguments for conservatism, and this book offered a good overview of how Burkean arguments about the “wisdom of the past” and Hayekian arguments about the wisdom inherent in decentralized systems can be used or misused.