Harvard Journal of Law and Public Policy

Summer 1994

Annual Institute for Humane Studies Law and Philosophy Issue

Symposium on Presumptions and Burdens of Proof

Presumptions and Transcendentalism

*691 YOU PROVE IT! WHY SHOULD I?

Lawrence B. Solum [FNa]

Solum

This Article explores ideas developed in Richard Gaskins' provocative book Burdens of Proof in Modern Discourse. [FN1] One idea, familiar and powerful, is that the conceptual device of a burden of proof is functionally related to the problem of uncertainty. A second idea, less familiar and more controversial, is that modern discourse is characterized by the use of burden shifting arguments that are premised on a special sort of ignorance -- ignorance of a hidden or transcendental realm of values. This Article advances two claims in relation to these ideas. Part I argues that an adequate understanding of the function of burdens of proof requires attention to the distinction between two sorts of uncertainty -- that is, to the difference between risk and ignorance. Part II contends, pace Gaskins, that much of contemporary discourse (in particular, the work of Rawls and Dworkin) can best be understood as based on an approach to moral theory that avoids transcendental commitments.

I. Functions of the Burden of Proof

This Part explores the functions served by the device we call "the burden of proof." At the outset, observe that the burden of proof is a complex idea that represents at least two different notions. Using the legal context to illustrate this point, call the requirement that a party raise an issue "the burden of production." This requirement can be distinguished from a second requirement, the requirement that one party satisfy a standard of proof on an issue after it has been raised. Call this requirement "the burden of persuasion." The burden of persuasion has two components, the risk of nonpersuasion and the standard of proof. The risk of nonpersuasion refers to the consequence that flows if a burden of persuasion is not met. The party that bears the burden of persuasion carries the risk of nonpersuasion: if the standard *692 of proof is not met, the issue is decided against the party that bears the burden. The standard of proof refers to the quality of convincingness: "beyond resonable doubt," "by clear and convincing evidence," and "by the preponderance of the evidence" are all standards of proof. When we use the phrase "burden of proof," we sometimes refer to a complex standard incorporating a burden of production and a burden of persuasion: for example, the prosecution has the burden of production on each of the elements of a crime, and to meet that burden, the state must meet a burden of persuasion beyond reasonable doubt. At other times, we refer to one of these ideas (that is, the burden of production, the risk of nonpersuasion, or the standard of proof), as the "burden of proof." With these preliminary points about terminology out of the way, we can address the function of a burden of proof.

A. Uncertainty and the Burden of Proof

Why do we have burdens of proof? What functions do such burdens serve? Richard Gaskins suggests that part of the answer to these questions may be found in our need to cope with what he calls "ignorance." Under the subheading, "The Varieties of Ignorance," he writes:

Everyday legal proceedings are supposed to reach definite conclusions about individual entitlements and responsibilities, notwithstanding any limits of information. Courts are routinely expected to meet this need, even when vitally relevant information is either too costly or simply unavailable. Moreover, judges often reject potential evidence that ordinary observers might otherwise consider, based on the assumption that it could not be weighed fairly by a neutral party. Thus the courts deliberately make themselves ignorant of self- serving statements, of third-hand accounts of what witnesses may have said, and of the layperson's conjectures on technical matters. Formal legal rules about the burden of proof compensate for precisely these occasions of routine or practical ignorance. [FN2] Gaskins goes on to describe other forms of ignorance, but at this point, let us pause and consider the variety that he terms "practical." Gaskins uses the phrase "practical ignorance" to describe the lack of information for which legal burdens of proof compensate. For reasons that will become clear below, I substitute the term "uncertainty" for "ignorance." With the substitution in *693 place, we can briefly explore Gaskins' point, bearing in mind the distinctions between the burdens of production and persuasion.

The function of a burden of proof, according to Gaskins, is to compensate for uncertainty. [FN3] Take a simple illustrative case. Someone, P, has suffered an injury, alleging that it has been caused by the negligent actions of another person, D. P brings a lawsuit against D. Let us suppose a jury trial has been conducted and all of the evidence has been heard. The judge instructs the jury that P bears the risk of nonpersuasion on the issue of causation and that the standard of proof is the preponderance of the evidence. What is the function of the burden of persuasion in this situation? First, note that if there is no uncertainty at all, the burden of proof drops out as a factor in the decision process. If the jury is certain that D's actions caused P's injury, then D would be entitled to prevail irrespective of the burden of persuasion or the standard of proof. Second, the burden of persuasion provides a rule for deciding the case in the event of complete uncertainty. (For reasons provided below, I call a state of complete uncertainty "ignorance"). In the unlikely event that the jury believed that the evidence told us nothing at all about the question whether D's injuries had been caused by P's actions, that is, that the jury had no relevant information, [FN4] then D should prevail on this issue because P has the burden of persuasion. Third and finally, the standard of proof tells the jury what to do if it believes that the evidence leaves it somewhere between certainty and ignorance. For example, if the jury believes that there is a 50/50 or smaller chance that P's action caused D's injuries, then D is entitled to prevail, because the standard of proof is the preponderance of the evidence.

Burdens of proof, then, can be seen as decision rules that provide direction in cases of uncertainty. To explore this conception of the function of the burden of proof further, one might consider a more general question: How can one make rational choices under conditions of uncertainty? Answering that question *694 requires us to take a brief detour and roughly sketch the contours of rational choice theory.

B. Foundations: A Sketch of Decision Theory

Rational choice theory (or decision theory) attempts to give a formal account of how rational beings can make decisions under a variety of conditions. [FN5] The basic story is familiar folk psychology; that is, rational choice theory formalizes common-sense assumptions about how people do (and ought to) make choices, at least in some circumstances. Rational choice theory makes some simplifying assumptions, which may prove false in real choice situations, but we will leave consideration of those assumptions aside to get to the point quickly.

We will begin with an example from outside the law: an actor is faced with making a choice between two actions -- for example, to eat ice cream or to eat celery. Each action has an associated consequence; call the consequences of an action its "payoff." Eating ice cream produces a very pleasant sensation, provides calories (which in turn may add to one's weight problem), and so on; eating celery will produce a mildly pleasant sensation, provides virtually no calories, and provides roughage. The rational choice is the choice with the best consequences, that is with the most valuable payoff. Let us add a formal, idealizing assumption to the story: assume that there is a single quantum scale on which each actor assigns values to actions. Call the units on this scale "utiles" and the payoff for an action its "utility." This notion of utility allows us to represent all the consequences of an action with a single value. In our simple example (Table 1), the rational choice is Action One, which has a payoff of 5 utiles, as opposed to Action Two, with a payoff of 2 utiles.

Utility

Action One: Eat ice cream. 5

Action Two: Eat celery. 2

Table 1

The discussion that follows does not argue that the above version of rational choice theory is true; rather, we will assume that rational choice theory is correct, so that we may focus clearly on *695 the problem that uncertainty poses for our understanding of how people make choices and how they ought to do so.

1. The Idea of Risk

In our simple story, we have not yet asked any questions about how we know what the payoff is. Life is not always certain. Perhaps one will not like the ice cream -- it might be too cold and cause a headache -- but perhaps it will be delicious. We can add uncertainty to our story and get the picture shown in Table 2.

Utility (U) of Action Given

State

State One: Ice cream State Two: Ice cream is just

is too cold right

Choice One: Eat ice 1 5

cream.

Choice Two: Eat celery. 2 2

Table 2

Utility (U) of Action Given

State

State One: Ice cream State Two: Ice cream is Expected

is too cold. (p = just right. (p = .5) Utility S(p

.5) * U)

Action One: 1 * .5 = .5 5 * .5 = 2.5 .5 + 2.5 = 3

Eat ice

cream

Action Two: 2 * .5 = 1 2 * .5 = 1 1 + 1 = 2

Eat celery

Table 3

What has just been illustrated is one kind of uncertainty: the kind for which it is possible to assign probabilities to the various *696 possible states. Let us refer to this sort of uncertainty as "risk." Under conditions of risk, it is rational to maximize expected utilities. At least it is rational in the sense that this strategy will, if applied to many choice situations, result in a greater sum of payoffs in the long run than will any alternative strategy. A caveat: this is a very simple version of the rational choice story; for example, it omits any discussion of risk preference or aversion, avoiding issues that must be faced in a more complete version of the theory.

2. The Idea of Ignorance

But what if we do not know what the chances of the ice cream being too cold are? Suppose that we have no basis for assessing the probability that the ice cream is too cold; we have no prior experience, no way to sample the ice cream, no information other than the knowledge that the ice cream will either be too cold or not too cold. The theory will no longer be able to identify the rational choice. Returning to the ice cream and celery example, as illustrated in Table 4, one now cannot act rationally by maximizing expected utility. By stipulation, we have no information about the probabilities of State One and State Two, and hence we cannot discount the utility of each payoff by the likelihood that it will occur.

Utility (U) of Action Given

State

State One: Ice cream State Two: Ice cream is just Expected

is too cold. (p = right. (p = ?) Utility

.?) S(p * U)

Action One: 1 * ? = ? 5 * ? = ? ?

Eat ice

cream

Action Two: 2 * ? = ? 2 * ? = ? ?

Eat celery

Table 4

One may call the condition in which we are truly in the dark about probabilities of possible states "ignorance." Ignorance and risk are both forms of uncertainty, but in a condition of risk, one can assign probabilities (or at least qualitative assessments) to the possible states, whereas in a condition of ignorance, one lacks information on which to base such an assignment.

Is there an alternative to the method of expected utilities that will allow us to make a rational choice under conditions of ignorance? Consider some possible methods for coping with ignorance:

.*697 . Guess: for example, assume that each state is equiprobable.

. Play it safe: choose the action with the best worst-case outcome.

. Go for it: choose the action with the best best-case outcome.

There are many other possible strategies for coping with ignorance, but there is no knock-down argument that any such strategy is more rational than any of the others as a general rule for deciding how to act under conditions of ignorance. Each of the strategies has serious problems.

Consider, for example, some of the problems with guessing. Assuming that each state is equiprobable has the advantage of allowing the calculation of expected utilities, but there is a price to be paid. Why is it any more rational to assume the states are equally probable than to make some other guess, for example that the probability of State One is 0.3 and the probability of State Two is 0.7? Moreover, the equiprobability assumption makes the way one slices up the possible states of the world very important. Why not three states, for example ice cream is too cold, ice cream is almost too cold, ice cream is just right? Suddenly, the probability of too cold becomes 0.33 instead of the 0.5 that was assigned when there were two equiprobable states!

The alternative strategies, playing it safe and going for it, also have problems. Playing it safe under conditions of ignorance has a certain appeal. By choosing the action with best worst-case payoff -- call this the "maximin strategy" [FN6] -- one can be sure to avoid the worst disaster. No headaches from eating cold ice cream! But why is it rational to be so conservative? Why deprive oneself of the pleasure of eating ice cream because of an unknown risk of a headache? Why not go for it? If one fails to choose the action with the best best-case outcome -- call this the "maximax strategy" -- one will never know how good it could have been. I will refrain from further rhetorical questioning, but I hope my informal argument has conveyed the sense of the more rigorous debate in decision theory. Although the debate may not yet be finally settled, there simply does not seem to be any generally applicable strategy for rational choice under conditions of ignorance. The lack of a generally applicable strategy does not, however, mean that we are utterly lost when confronted with ignorance. We do make choices when we are truly in the dark, *698 and in different situations we may employ each of the strategies outlined above.

To sum up: rational choice theory allows us to formalize some common-sense ideas. People choose the action that will produce the best consequences. Some choices must be made despite a lack of certainty about what the consequences will be. Two sorts of uncertainty can be distinguished. The first, risk, obtains when one knows or can estimate the probabilities of the various possible consequences. The second, ignorance, obtains when one does not know and cannot estimate these probabilities. Under conditions of risk, the rational choice is the one that maximizes expected utilities. Under conditions of ignorance, it is not clear whether any strategy for choosing is uniquely rational.

C. Decision Theory Applied to the Case of Destruction of Evidence

Consider the implications of this conceptual apparatus for a problem the law faces in allocating the burden of proof. [FN7] This problem arises in the context of the destruction of evidence; such destruction usually becomes an issue when done by a party to a civil or criminal case. [FN8] Destruction of evidence may create uncertainty. [FN9] Documents are shredded; a tape is erased; the murder weapon has vanished. As a result, uncertainty increases. Was there a smoking gun with the fingerprints of the defendant? Or someone else? How does the law cope with this uncertainty?

Consider two hypothetical cases. In the first, "the case of the burned will," Jane has recently died, leaving two children, Tom and Bill. Before her death, Jane wrote a will and told her friends that she left all her money to one of her two sons. After Jane's death, her servant Walter, hiding behind a curtain, sees Tom take the will from the safe, read it, and then burn it. Under the *699 law of intestate succession, Tom and Bill would share equally in her estate. [FN10] What will the law do? In situations akin to the case of the burned will, the classic legal solution has been to infer -- from the fact that Tom burned the will -- the conclusion that the will would have been unfavorable to him; in this case, the inference is that the will left everything to Bill. The inference is sometimes called the "spoliation inference," deriving from the Latin maxim omnia praesumuntur contra spoliatorem. [FN11] Everything is presumed against a wrongdoer. The assumption behind the inference is that Tom would only destroy the will if it named Bill as the sole heir, because only in that situation would destruction of the will be in Tom's self interest.

In the second case, "the case of the shredded files," the Acme Manufacturing Company, the nation's leading maker of widgets, has been involved in a series of bitter disputes with its rival, Zenith Widgets Incorporated. Zenith has been making noises about filing an antitrust action against Acme. A few days before the suit is filed, a secretary at Acme shreds all of the documents that could have sorted out the truth of Zenith's claims. [FN12] The documents were destroyed pursuant to a policy under which these files are routinely shredded once every twenty-four months.

In the case of the burned will, the destruction of the will creates uncertainty about its contents. Nonetheless, the inference that we can draw from Tom's act of destruction allows us to deal with that uncertainty. If we had the will itself, we could know what it said with absolute certainty; without the will, we have the inference that it is more likely than not that the will favored Bill. Although Tom may protest that he destroyed the will to protect his brother from Jane's hurtful action, we are inclined to think that version of events is less likely than the one in which Tom destroyed the will to get half the estate rather than none of it. The law copes with the form of uncertainty that we are calling "risk" by allowing Bill to meet his burden of persuasion by the requisite quantum of proof, that is a preponderance of the evidence. The spoliation inference is sufficient to shift the burden *700 of persuasion to Tom, who now must prove that the will did not favor Bill. [FN13]

The case of the shredded files concerns a different sort of uncertainty. The spoliation inference does not seem to fit this case, because the files were destroyed pursuant to a routine policy and were not destroyed in direct response to the filing of a complaint by Zenith against Acme. Without the inference from intentional destruction, we really do not know what the files would have proved; in the terminology that we have adopted, we now face that kind of uncertainty we call "ignorance." But the law can deal with this ignorance. For example, the courts could raise an presumption of law that evidence that is unreasonably destroyed would have been unfavorable to the party that destroyed it. In the case of the shredded files, that presumption would result in a shift of the burden of persuasion. Acme would now carry the burden of proving with a preponderance of the evidence that it did not commit an antitrust violation. Deprived of its own files that bear on the question, Acme might be unable to produce evidence sufficient to meet that burden of persuasion with a preponderance of the evidence.

The two cases differ from each other in a very important way. The case of the burned will involves uncertainty as risk; the court could make a decision that was rational (in these sense that expected utility would be maximized) by making the decision most likely to be correct. The case of the shredded files involves uncertainty as ignorance. We cannot say which decision is most likely to be correct, because we do not have evidence that establishes the probability that the files would have proved Zenith's claim against Acme. We are truly in the dark, but the case must be decided anyway. The courts have had a difficult time deciding cases that are analogous to our hypothetical case of the shredded files. In such cases, the creation of a presumption and hence of a corollary burden of persuasion would frequently decide the dispute. If we presume the evidence was damaging to the destroyer, then the party that destroyed the evidence is likely to lose. If we place the burden on the other party to prove that the destroyed material would have favored its case, then it is unlikely to be able to meet that burden. No matter where we place the burden, we cannot know the likelihood that the decision will result in error.

*701 The discomfort elicited by the case of the shredded files suggests that when we deal with ignorance, our decision about where to place the burden of proof must be made on some ground other than maximizing the likelihood that the decision will be accurate (meaning correct in the sense that it is the decision that would have been reached had there been no uncertainty). If we do not look to accuracy, then to what should we look? One possibility is that we should assign the burden of proof to create incentives that will reduce future uncertainty; we erect a presumption against destroyers of evidence in order to deter future acts of destruction. Another possibility is that we should assign the burden of proof to prevent unfairness; as between Acme, which allowed its employees to destroy evidence even though a major legal dispute was on the horizon, and Zenith, which did not act in this unreasonable fashion, we might conclude that it is more fair that Acme should bear the burden of proof and the corresponding risk that an erroneous decision will be made.

D. Two Functions for the Burden of Proof

In sum, the main idea of this Part of the Article is that the notion of a burden of proof has two different but related functions, with each function corresponding to one of the two varieties of uncertainty.

1. Dealing with Risk

Sometimes, we deal with risk by assigning a burden of proof. We know something about the probabilities (either in a general class of cases or in a particular case). That knowledge underwrites a burden of persuasion. Thus, in a legal case, allocation of the burden may simply be a method or procedure for producing the best outcomes in the long run, or (in the language of decision theory) maximizing the expected utility of legal proceedings. Usually, this will be the most accurate outcome, but not always. For example, the burden of persuasion in a criminal case may be proof beyond a reasonable doubt, because the disutility of convicting an innocent person far exceeds the disutility of finding a guilty person to be not guilty: better that ten guilty persons go free than one innocent person be convicted.

*702 2. Coping with Ignorance

In other cases, however, we use the idea of a burden of proof not to deal with our knowledge of risks but to cope with our ignorance. Ignorance poses a challenge to the courts that is, in a sense, much more fundamental than the challenge posed by risk. When we allocate a burden of proof to cope with ignorance, we cannot say that this minimizes the chances of an erroneous decision; we do not even know what those chances are. It might be nice if we could somehow avoid this problem, but that is not an option. The burden of proof must be assigned to one party or the other. Someone must win and someone must lose.

II. Burdens Without Transcendentalism

This Part takes up Gaskins' discussion of two more catagories of what he calls ignorance but I call uncertainty. The connection between this Part and the last will not be immediately apparent. The earlier discussion of decision theory will be brought to bear in the analysis of Gaskins' critique of Ronald Dworkin's legal theory and John Rawls's political philosophy; Gaskins' failure to grasp the difference between risk and ignorance is related to his criticism of what he calls the "transcendental" basis of arguments made by Rawls and Dworkin. But to elucidate that relationship, a number of points must be explored.

A. Consensual and Transcendental Ignorance

After identifying "practical ignorance" (or "practical uncertainty"), Gaskins discusses two additional sources of ignorance. The first variety Gaskins calls "cultural ignorance." Gaskins writes:

The language of proof could misleadingly suggest that factual evidence alone marks the difference between ignorance and knowledge in public discussion. In legal proceedings, as in wider rhetorical contexts, however, the basic norms or rules for weighing evidence often contribute more fundamentally to the condition of ignorance. These rules are the reigning presumptions used by the legal system to interpret and to draw conclusions from factual evidence. When uncertainty arises about the meaning or legitimacy of these presumptions, a more profound level of ignorance casts its shadow over the strategic course of argument. This source of ignorance is not *703 merely informational, but largely consensual or cultural in origin. [FN14]

In addition, Gaskins identifies a second variety of ignorance, which I will call "transcendental ignorance." Gaskins writes, "finally, in addition to practical and consensual indeterminacies, we can identify an even more fundamental level of public ignorance. This distinctive notion is ultimately derived from philosophical theories about hidden, or transcendental, layers of possible knowledge." [FN15] He continues, "as commonly applied to the duality of facts and values, for example, transcendental theory projects a realm of potential knowledge (the realm of values) while denying that human beings have direct sensory access to it." [FN16] One observation about Gaskins' notion of transcendental theory is appropriate at this point: We lack direct sensory access to a great deal of knowledge, ranging from knowledge of scientific entities like atoms and molecules, which we cannot perceive directly, to knowledge of events like our parents' wedding, which most of us could not attend. On the one hand, if lack of direct sensory access is the criterion of the transcendental, then the transcendental realm is quite large. On the other hand, if the transcendental realm is hidden in some more profound sense, then we might ask what the precise criterion of the transcendental is.

Gaskins' exploration of the role of burden shifting arguments in relationship to these two varieties of ignorance is rich and complex, and this Article cannot do it justice. What it can do is to isolate particular examples of that exploration, that is, Gaskins' claim to have identified transcendental assumptions in the work of Dworkin and Rawls.

*704 B. Is Hercules a Transcendentalist?

The basic outline of Dworkin's normative and descriptive theory of judicial decisionmaking is, by now, likely to be quite familiar to the readers of this Symposium, and I will not be repeat it here. [FN17] Instead, I turn directly to Gaskins' interpretation of that argument, which he summarizes as resting on the following moves:

1. We can presume there are correct answers to questions about legal rights, even though we may not know (or be able to prove) what they are.

2. We can presume there is a complex procedure appropriate for discovering rights in controversial situations, even though that procedure cannot itself be formulated as a series of rules.

3. We can presume that judges are uniquely placed to perform this procedure, even though no one else is in a position to measure their success against independent criteria. [FN18]

Consider the first of these three points as an example of Gaskins' argument. Does Dworkin argue that we can presume the truth of the right-answer thesis, even though we cannot know what they are? Gaskins offers as evidence the following passage from Taking Rights Seriously:

Some readers may object that, if no procedure exists, even in principle, for demonstrating what legal rights the parties have in hard cases, it follows that they have none. That objection presupposes a controversial thesis of general philosophy, which is that no proposition can be true unless it can, at least it principle, be demonstrated to be true. There is no reason to accept that thesis as part of a general theory of truth, and good reason to reject its specific application to propositions about legal rights. [FN19] If this were the only argument that Dworkin gave in defense of the right-answer thesis, Gaskins might have a point. [FN20] But this passage is not Dworkin's primary argument in favor of the thesis; indeed, as is apparent from the text itself, Dworkin's aim here is to answer an argument that might be raised against the idea that *705 there can be rights at all. This argument that there are no rights at all is based on the notion that no proposition can be true unless there is a method for demonstrating that it is true. Dworkin then argues that this argument is false for two reasons. The first reason, which Dworkin does not elaborate, is that there are many propositions that are true but cannot be demonstrated to be true. Indeed, Gaskins is quite familiar with this fact; the problem of ignorance is central to his work. There are many garden-variety factual propositions that are true, but which will never be provable simply because we lack the information to prove them.

The second reason, which Dworkin elaborates in depth, is that legal propositions in particular can be true even though there is no procedure for proving them. [FN21] It is crucial to distinguish between demonstration (the sort of reason-giving involved in deductive systems such as Euclidean geometry) and reason-giving. Dworkin does not claim that we have legal rights but are incapable of giving good reasons for believing in them. He does argue that we cannot demonstrate their existence. Legal reasoning is not like geometry, but the same is true of many realms of human discourse.

It is fair to ask Dworkin to give an account of legal reasoning that includes the giving of good reasons without demonstrations. Dworkin provides this account in his story of Hercules, an imaginary judge. [FN22] Sometimes Hercules will be faced with a hard case, a case in which a statute is ambiguous or the common law provides support for both sides to a controversy. When Hercules is faced with such a hard case, he uses a procedure of construction to arrive at the right answer in the case before him. The method of construction requires him to achieve coherence between his decision in the case before him and the law as a whole. In particular, the best decision must fit the existing body of law and provide the best justification for it. Hercules constructs a theory of the law that decides the case before him and fits and justifies the law as a whole. [FN23]

Dworkin's Hercules is an idealization of a real judge. If Dworkin is correct and the law is a seamless web, the decision of any hard case potentially requires the judge to develop a theory of *706 the law as a whole. Real judges lack the time to construct such a theory from scratch every time they decide a hard case. Good judges must develop a working theory of the law and may rely on hunches about which lines of argument are likely to prove fruitful. Dworkin's story of Hercules offers a rational reconstruction of the process that real judges approximate.

Does the use of this idealization commit Dworkin to the belief in a transcendental or hidden realm of value? Surely not. The method that Dworkin proposes for legal reasoning is one that can be applied to reasoning about many types of questions. Dworkin himself explores the connections between his approach to legal reasoning and the interpretation of literary texts. But the point can be made more clearly if we consider how we reason about ordinary human affairs. Forexample, we need to interpret the actions of a neighbor to figure out their meaning. Why is our neighbor building a fence? We construct a theory of her motives and beliefs; the ideal theory would fit her action on this occasion into a coherent story about her life as a whole. Unless we are writing a biography, however, we will have to make do with a rough-and-ready sort of theory, based on limited data and arrived at after brief deliberation. We cannot provide a demonstration for our rough theory of our neighbor's behavior; interpretation of human behavior is not like Euclidean geometry.

If asked what method we employed, we might use an idealization to explain what we are doing. If pressed, we might say that when dealing with other people, one should aim at the interpretation of their behavior that a biographer with full access to their life story would give. But using this idealization does not commit us to the proposition that our neighbor's motives and beliefs are to be found in a biography shelved in the central library of some transcendental realm. You cannot see, touch, taste, or smell other people's motives and beliefs, but basing your actions on inferences about them does not involve metaphysical extravagance. [FN24] Dworkin's Hercules possesses extraordinary intellectual virtues, but he does not inhabit Mount Olympus.

*707 C. Does the Veil of Ignorance Shroud a Transcendental Realm?

I assume that the readers of this Symposium will be familiar with the argument of A Theory of Justice. [FN25] From the beginning, however, there has been confusion about the role of the original position in that argument. Consider Gaskins' interpretation of the original position:

Only by constructing this transcendental device could Rawls find a point of view from which to conserve the notion of a self-standing ethical authority independent of the conditioned desires and secular interests that haunted the utilitarian categories of Orwell's style of humanism. Structural principles for the just society were validated in being chosen by individuals behind a carefully constructed "veil of ignorance." Rawls wanted to ask precisely these individuals to make the critical inferences about preferred social structures; their lack of information about the specific consequences of their action, or about the conventions of a specific culture, guaranteed that their selection could not be based on mere temporal calculations. The original position theory was thus a transcendental application of the argument-from-ignorance. It gave an absolute preference to inferences made in a postulated temporal vacuum, where transcendental norms were presumed to be the default criteria. [FN26] This interpretation badly misses the mark, but that does not mean that it is groundless. There are reasons for misreading A Theory of Justice in this way. One reason involves the structure of the book. The original position is introduced very early, and the argument from the original position is quite striking. The original position does rely on a veil of ignorance, and Rawls does borrow the maximin rule from decision theory as a basis for decisions under conditions of ignorance. So far, Gaskins seems to be on track.

Moreover, Section 40, entitled "The Kantian Interpretation of Justice as Fairness," lends support to the notion that the original position involves transcendental metaphysics. [FN27] In particular, consider the following passage: "My suggestion is that we think of the original position as the point of view from which noumenal selves see the world." [FN28] This passage might seem to offer support for Gaskins' interpretation -- the phrase "noumenal selves" suggests *708 beings that actually inhabit a transcendental realm -- but there is another side to the story.

The other side of the story begins with two observations about the Kantian interpretation. First, we ought to be very careful in drawing sweeping conclusions about Rawls's view of Kant from the condensed discussion in Section 40 of A Theory of Justice; Rawls made it clear that his views were different from Kant's in several respects but begged our indulgence for his inability to elucidate fully those differences in the compressed discussion in A Theory of Justice. [FN29]

Second, Rawls has clarified the relationship between his political constructivism and Kant's moral constructivism in his recent book Political Liberalism. [FN30] The clarification in the latter work suggests that Rawls rejects the proposition that his constructivism entails transcendental commitments. The passage that follows is taken from a larger discussion of the role of autonomy in Rawls's views as compared to its role in Kant's moral theory. After discussing the way in which political liberalism understands the autonomy of a political view, Rawls writes:

Another and deeper meaning of autonomy says that the order of moral and political values must be made, or itself constructed, by the principles and conceptions of practical reason. Let us refer to this as constituitive autonomy. . . . Constituitive autonomy says that the so-called independent order of values does not constitute itself but is constituted by the activity, actual or ideal, of practical (human) reason itself. I believe this, or something like it, its Kant's view. His constructivism is deeper and goes to the very existence and constitution of the order of values. This is part of his transcendental idealism. The intuitionist's independently given order of values is part of the transcendental realism Kant takes his transcendental idealism to oppose.

Political liberalism must, of course, reject Kant's constitutive autonomy. . . . And certainly political constructivism accepts his view that the principles of practical reason originate, if we insist on saying they originate anywhere, in our moral consciousness as informed by practical reason. They derive from nowhere else. [FN31]

*709 Wrenched from context, this passage is heavy going. For our purposes, two points should be made. First, Rawls is right about Kant: Kantian constructivism does not entail a metaphysical commitment to the reality of an independent realm of transcendental values. Indeed, the point of Kant's enterprise is to deny transcendental realism and to affirm that human reason, not some transcendental realm, is the source of moral values. There may be forms of intuitionism that are committed to the existence of a transcendental order of moral values, but Kant's theory is not a form of intuitionism. Second, Rawls contends that his own view is independent of deep theories like Kant's; Rawls represents that his view is shallow and that it deliberately avoids transcendental commitments. Rawls' political constructivism neither affirms nor denies the existence of a transcendental realm of values.

Why is Rawls's view a shallow one? The answer to this question requires us to back up and consider his theory from a different angle, without reference to the original position until a very late stage in the argument. In his recent work, Rawls has emphasized the idea that political liberalism is a response to the fact of pluralism; the fact that European and North American societies have been characterized by radical disagreements over ultimate questions of religion and morality since the Wars of Religion of the Sixteenth Century. [FN32] Political liberalism responds to this fact; a liberal theory of justice should be a view that can be the subject of an overlapping consensus among a variety of comprehensive moral and religious perspectives. [FN33] Such a theory must be based on public reason -- on common sense and values that can be drawn from the public political culture. [FN34]

Given the fact of pluralism, political liberalism cannot rely on deep premises about metaphysics or the ultimate source of values. Given Rawls's view of the constraints within which his political theory must live to serve its intended purpose, he cannot use a transcendental realm of value as a premise of his arguments. There may be comprehensive moral or religious views that do incorporate belief in such a realm, and these views may participate in an overlapping consensus on a liberal theory of justice. *710 Indeed, it is possible that some moral intuitionists might affirm Rawls's political theory because they believe it is supported by their transcendental views, but their arguments for political liberalism from within their own view are not the arguments of political liberalism itself.

If the argument of A Theory of Justice does not rely on transcendental premises instantiated in the original position, then what is the basis of the argument? Gaskins discusses the key to answering this question: the theory of reflective equilibrium. [FN35] We have already seen how a procedure such as reflective equilibrium can operate without making metaphysical assumptions about the existence of a transcendent realm. The method of reflective equilibrium is analogous to the method employed by Hercules in deciding hard cases. Reflective equilibrium does not promise "to bring (at long last) the transcendental realm into direct contact with everyday experience." [FN36] Putting aside the fancy terminology, Rawls's argumentative strategy is to take the citizens of modern democratic societies as they are, with their differences, such as individual beliefs about religion and morality, and their commonalities, such as shared resources of public reason. There is no other place to begin; there are no transcendental tricks that can convince us to affirm a conception of justice on the basis of arguments that do not take our current beliefs and motivations as a starting point.

Does Rawls's constructivism necessarily involve him in transcendental commitments? No; constructivist reasoning can be simple and straightforward. A parent asks a child who has hurt a sibling to think about how she would feel if the sibling had treated her the same way. This is a piece of constructivist reasoning in its most simple and basic form, and it surely does not entail any extravagant transcendental metaphysics. The question, "How would you feel if you were treated that way?", does not commit the parent to belief in some transcendental realm in which the role reversal actually occurs. [FN37]

Rawls asks us to imagine what sort of society we would create under conditions that would be fair. The features of the original *711 position must be justified independently, by the method of reflective equilibrium, by good arguments from premises that we have reason to accept. Because the original position is only an argumentative construct in political philosophy, the argument from the original position cannot be the only route to justification of principles of justice. The same principles might be justified by any number of argumentative strategies. The original position may provide focus; it allows us to formalize considerations of fairness that might otherwise be difficult to clarify. The original position is laid out as a step in a constructivist argument; it is not a position in transcendental space.

To bring the argument full circle, consider the connection between the method of reflective equilibrium and the problem of uncertainty. Gaskins reads Rawls as if Rawls were offering an argument that proved the truth of his theory of justice with apodictic certainty. Perhaps transcendental arguments promise freedom from doubt and disagreement, but this is not the goal of reflective equilibrium. As Rawls puts it, "The struggle for reflective equilibrium goes on indefinitely." [FN38] Once we put aside the notion that Rawls is after apodictic certainty, then the push to read transcendental commitments into his arguments dissolves. Analogously, we sometimes create inferences, presumptions, and burdens of to deal with uncertainty, but it would be a terrible mistake to mistake a presumption for certain knowledge of the facts presumed.

Conclusion

Is Gaskins right to see transcendental commitments lurking behind the use of burden shifting arguments in contemporary discourse? The starting point for an answer to this question is the distinction between uncertainty as risk and uncertainty as ignorance. Some burden shifting arguments are based on risk (or nonprobabilistic cousins of risk). Sometimes we can produce a good argument for a conclusion that falls short of certainty, but that does make the conclusion more likely than not. In these circumstances, a burden shifting argument is simply rhetorical shorthand: given the balance of reasons, one should believe this conclusion, unless something has been overlooked. No transcendental *712 commitment is entailed by this routine use of burden shifting as a rhetorical technique.

Sometimes a burden of proof is employed in connection with conditions of ignorance. One can imagine burden-shifting arguments that do involve transcendental commitments. For example, "you cannot prove that Angels do not exist because they inhabit a transcendental realm outside of human experience." But it does not follow that every argument which establishes a burden of proof to deal with a condition of ignorance involves a transcendental commitment. Recall the way the law deals with the destruction of evidence. We are faced with our ignorance of the facts, but we must act, and so we make an inference or we establish a presumption or a burden of proof. The establishment of a presumption that destroyed evidence would be unfavorable to the spoliator does not commit the law to the proposition that our hypothetical destroyed files still exist in some transcendental filing cabinet. The burden of proof is simply a practical device for resolving a practical problem. We can argue for such a burden on the basis of common-sense considerations such as fairness or deterrence of future wrongdoing; no transcendental moves are required.

So far, this Article has established that the use of burdens of proof in connection with factual risk or ignorance need not entail transcendental commitments. Gaskins is concerned not with these factual cases, however, but instead with cases involving values. It is moral values, according to Gaskins, that are given transcendental status by arguments from ignorance or burden- shifting arguments. But why is this so? One can give a good moral, political, or legal argument that falls short of certainty but avoids any transcendental commitments. Rawls' use of reflective equilibrium provides one example of how such arguments can proceed, but moral philosophy, starting with Aristotle, is filled with such arguments. Indeed, Gaskins gives no reason for the proposition that the use of arguments from ignorance in moral theory entails a commitment to transcendental premises. Instead, he seems repeatedly to assume that this is the case.

Early in Burdens of Proof in Modern Discourse, Gaskins cites a famous passage from Bishop Whately's Elements of Rhetoric. "There is a 'Presumption' against any thing paradoxical, i.e. contrary to the prevailing opinion: it may be true; but the Burden of proof lies with him who maintains it; since men are not to be expected *713 to abandon the prevailing belief till some reason is shown." [FN39] Most of us do not believe that our practices of argument commit us to belief in a transcendental realm of values. In his rich and intriguing book, Richard Gaskins has not met his burden of proof to convince us otherwise.

[FNa]. Associate Dean for Academic Affairs, Professor of Law and William Rains Fellow, Loyola Law School, Loyola Marymount University. I owe thanks to Randy Barnett, Victor Gold, and David Leonard for their comments on a draft of this Article.

[FN1]. Richard H. Gaskins, Burdens of Proof in Modern Discourse (1992).

[FN2]. Id. at 4.

[FN3]. Uncertainty may result from a lack of evidence or it may result from contradictory evidence. See Ronald J. Allen, Burdens of Proof, Uncertainty, and Ambiguity in Modern Legal Discourse, 17 Harv. J. L. & Pub. Pol'y. 627 (1994).