IS THERE AN >US= IN >JUSTIFICATION=?
I. INTRODUCTION
A critical question for epistemologists is whether there are any inter-subjective requirements for having a justified belief C whether there is an >us= in >justification=. One recent epistemologist that has addressed this issue is Keith Lehrer. In Knowledge, Lehrer presents a theory of justification which he describes as subjective.1 More recently, however, Lehrer has defended the view that rationality is partially intersubjective. He claims,
Rationality, whether practical or intellectual, intersects with consensus in a central way. Expectation is a function of probability. On the account of epistemic utility I have articulated, utility is measured in terms of probability. The subjective theory of probability affirms that such probabilities can be extrapolated from preferences. Coherent subjective probability assignments, though difficult to attain, do not seem sufficient to avoid irrationality. For a person may be completely coherent in his probability assignment, and yet be demented or dogmatic. We may insist, consequently, that he is irrational. A person who maintains that some supernatural being is dedicated to his torment, or who claims that the earth is flat and evidence to the contrary fabricated, may be guilty of neither inconsistency or incoherence. We insist, nonetheless, that he is unreasonable. In terms of his probability assignments, he may be maximizing epistemic utility in what he accepts. His unreasonableness must lie within his probability assignment ...
The solution is, I suggest, contained in the editorial use of we. We think their assignment is unreasonable though coherent. From our point of view they are unreasonable... .A person who contravenes reasonable consensus, though he may yet be right, is in one sense unreasonable. Given an account of reasonable consensus, we shall be able to explain what we mean when we say that some individual, though perfectly coherent in his probability assignment, is unreasonable.2
Lebrer presents in this passage an argument that some sort of intersubjectivity is necessary for justification. For he claims that coherent probability assignments are necessary, but not sufficient, for justification. If a person violates Lehrer=s intersubjectivity requirement, his belief is thereby rendered unreasonable. We shall see later the precise nature of this intersubjectivity requirement and the conditions under which intersubjectivity is necessary for justification. At this point, though, it is sufficient to see that Lehrer affirms some sort of intersubjectivity requirement as necessary for justified belief, and thus that the subjectivity of his earlier theory has been abandoned in his more recent work.
I wish to argue that this rejection of subjectivity fails and that the way in which it fails suggests that intersubjectivity is not very important in justification. Let us begin the task of defending these claims by investigating the nature of Lehrer=s intersubjectivity requirement.
II. INTERSUBJECTIVITY REQUIREMENTS
A problem for some intersubjectivity requirements is that different members of the same community can have different information. When this difference in information occurs, any sort of intersubjectivity requirement will fail to be necessary for justification. For example, a member of a community could have conducted experiments which no one else in his community has, and thereby have a rational belief on the basis of information not shared by anyone else in his community. Lehrer=s intersubjectivity requirement is intended to block these sorts of objections. His requirement is understood in terms of violating rational consensus. But one can only violate rational consensus once dissensus has been reached, i.e., when each member of the group is aware of all the information taken to be relevant to the truth of the issue at hand by all the other members of the group. When dissensus has been reached, the discussion is at a point of dialectical equilibrium.
Lehrer claims that once equilibrium of belief is reached by the members of a community, each member ought to weight the relative expertise of the other members of the community with respect to the issue at hand, and alter one=s original probability assignment by combining his original assignment with the results obtained by the competency estimates concerning the other members of the group. The exact mathematical function by which a new probability assignment is reached on this basis need not concern us, for the important point is simply that once equilibrium of belief is reached, this sort of aggregation by members of the community is required for rational belief. Lehrer defends the aggregation move by claiming that
... to refuse to revise a probability assignment is equivalent to assigning everyone else a weight of zero, assigning a weight of unity to oneself and averaging by the prescribed method. Assuming that one does not assign a weight of zero to other members of the group, one must, to be consistent, not refuse to follow the implications of the weights one assigns.3
Lehrer claims that the nonviolation of rational consensus is not a necessary condition for S=s having a justified belief, if the consensus view is not known by S or if the consensus is not calculated at the state of equilibrium. But if the consensus view is accessible to S and is calculated in a state of dialectical equilibrium, then the nonviolation of rational consensus is a necessary condition for S=s having a justified belief.4 Further, one violates rational consensus only after an equilibrium of dissensus has been reached, i.e., the members of the community have shared their views with each other and the disagreement that remains will not be solved by further discussion. After this equilibrium has been achieved, one violates rational consensus when one refuses to aggregate by altering one=s original probability assignments in accordance with the expertise that one takes each of the other members of the group to have.
There are two possible views here that are distinguished by what I shall call the positive value objection. That objection claims that it is not the case that one ought always to assign some positive value to the other members of the group, i.e., that there are some cases where it is reasonable to assign a zero to all the other members of the group and hence not reasonable to aggregate.
Lehrer grants that the positive value objection is sound; in fact, his position is formulated to account for that objection. For Lehrer claims that there are cases in which it is reasonable to assign everyone else in the group a value of zero. He says,
[The] defense of scientific and social iconoclasm can be dealt with by our model. The defense of iconoclasm might be treated as an exhortation to discount the opinions of others to zero at some level in order to avoid being driven from one=s iconoclastic outlook. We have not said that it is always unreasonable to assign others a weight of zero.5
It is important to pause here briefly and reflect on the nature of the intersubjective view of rationality that Lehrer is accepting. As we have seen, one version of an intersubjective account of rationality suffers from the problem of different information. A second intersubjective account suffers frQm the problem raised by the positive value objectionCthat account requires that one assign some positive value to others in the group and that one aggregate on the basis of these universally assigned positive values, once dissensus has been reached. It is important to notice here that it is subjective considerations that move one from the first intersubjectivity position to the second, and it is equally important to notice the role that a subjective notion of rationality plays in moving Lehrer from the second intersubjectivity position to the intersubjectivity position that he accepts. For it is considerations of what is going on in the head of the person in question that makes the problem of different information so critical, and it is also these same intra-mental considerations that grant plausibility to the positive value objection. For the problem of different information refers to evidence that only some isolated subject is aware of, and the positive value objection claims that some isolated subject can reasonably assign all other members of the group a zero value. Thus, what we have here are three levels of intersubjectivity requirements, specifiable by how much each position is infiltrated with subjective considerations.
In order to make discussion clearer, let us name the three positions. The first I shall call a completely objective intersubjectivity requirement (COIR), and the second position I shall call a partially objective intersubjectivity requirement (POIR). Lehrer=s own version 1 shall call a subjective intersubjectivity requirement (SIR). Now, it seems obvious to me that the problem of different information is a decisive objection to any COIR, but I shall not defend that here. In the next section 1 shall first defend the positive value objection. This defense supports Lehrer=s view in one sense, but in a deeper sense it also undermines his position. For there is another objection, which I shall call the anti-aggregation objection, which bears close resemblance to the version of the positive value objection which I shall develop. Thus, I shall argue that a defense of the move from a POIR to an SIR contains within it the seed for the destruction of that SIR.
III. SUBJECTIVE INTRUSION INTO INTERSUBJECTIVITY
We can approach the version of the positive value objection I wish to develop by considering the following purported objection to Lehrer=s views:
Lehrer=s view is that I should aggregate when dissensus has been reached. But suppose that it is perfectly rational for me to believe that I am in a better epistemic situation6 than the other members of my group with respect to the issue concerning which we disagree. In such a case, it would be quite irrational to alter my original estimate concerning the truth of this issue since I reasonably believe that I am so situated that I have better access to the truth than the other members do. I reasonably believe that I have better access to the truth because, it one=s epistemic situation is superior to another=s, then it is more likely that the first will believe the truth than the second, if they both have reasonable beliefs. But then, if I reasonably believe that my epistemic situation is superior to the othcr members of the group, I should not aggregate with thcm. I should not aggregate because my reasonably believing that my epistemic situation is supcrior makes it reasonable for me to bclieve that aggregating would only make me less likely to believe the truth.
Now, Lehrer=s response to this objection is to construe it as a version of the positive value objection C that it is sometimes reasonable to assign everyone else in one=s group a zero weight and to assign a unit weight to oneself.7 The objection, under this construal, it not then an objection to Lehrer=s SIR; but rather an objection to any POIR.
And it seems clear that the positive value objection shows that no POIR is necessary for justification, for there are many cases of this sort in which persons proceed perfectly rationally and yet violate a POIR simply because proceeding in that perfectly rational manner yields a perfectly reasonable belief that everyone else in their group is in an inferior epistemic situation with respect to the issue at hand. One may reasonably believe that they are in an inferior epistemic situation because one reasonably believes that their capacities are limited so that they cannot see the truth in this case, for example. But, if this were so, and one had no information other than their performance in this particular case, one would have no reason to assign a positive value to the other members of the group. After all, they are clearly mistaken in this case, and this case is all the experience one has to ~() on in assigning a value to them.
Now, in order to avoid this objection, a defender of a POIR must define what it is to be at the point of equilibrium with regard to p so that it is not possible that one be at the point of equilibrium and yet it be reasonable to believe that one is in a better epistemic position than the other members of the group with regard to p. One obviously inadequate definition of equilibrium is:
El: Group G is at the point of equilibrium concerning p = df. No further discussion would alter the probability assignments that the members of G make concerning p.
El is obviously inadequate since equilibrium is achieved on this definition whenever discussion will not resolve disagreement. Thus, equilibrium can be reached without any discussion at all, and the aggregation move is required at that point. But then a POIR faces the problem of different infomation which reveals the inadequacy of any COIR. Thus, El is inadequate.
Suppose then we try:
E2: Group 6 is at the point of equilibrium concerning p = df. Each member of 6 is completely aware of all the evidence had by all the other members of 6 concerning p.
But E2 is inadequate too. For suppose that a group of philosophers sits down to discuss whether wants are produced by the same forces of production that satisfy them. Suppose that we each share every reason that we have for the view that we think is correct. It need not be the case that we are all in the same epistemic situation with respect to the dependency of wants on the forces of production at this point, for it could be that everyone in the group except me weighs these reasons incorrectly and I know that they weigh these reasons incorrectly. In other words, the epistemic principles8 used by the other members of the group are inadequate and I know that they are inadequate. But, all of us are completely aware of the evidence each of us has. And it is nonetheless rational for me to believe that I am in a superior epistemic situation than the rest with respect to the issue at hand.
What we need, to repeat, is an understanding of equilibrium such that it is not possible that it be reasonable to believe that one=s epistemic situation is superior to the other members of the group once equilibrium is reached. Our final attempt is:
E3: Group 6 is at the point of equilibrium concerning p = df. Each member of 6 is aware of all the evidence had by the other members of G, and each member is aware of the epistemic principles used by the other members and why the other members of the group take the remaining members= epistemic principles to be inadequate.
This sort of situation would occur when each member offers what he takes to be obvious counterexamples to the epistemic principles used by the other members of the group, and cannot understand how what the other membe=rs take to be counterexamples to his principles are actually counterexamples. Does this understanding of equilibrium rule out the possibility of one rationally believing that one=s epistemic situation is superior?
It seems to me that it does not. After all, being in such a state of equilibrium entails being unconvinced by any objections to one s own epistemic principles and having others fail to see that what one thinks are decisive objections to their principles undermines those principles. Consider a case where each is offering purported counterexamples to the others. The other members of the group fail to accept that one=s proffered examples are effective in demonstrating the unacceptability of an epistemic principle. And further, no amount of discussion has changed or will change their stance. Might it not be the appropriate belief in this situation is that they either have a quite irrational belief, or that they suffer from some cognitive deficiency which has shown up in this case? It seems to me that this is not only possible, it is the actual and approprite conclusion that most of us would come to in such a case. After all, understanding the force of a counterexample is one of the most fundamental abilities relevant to ascertaining truth.
Thus, it would seem that there is enough subjective intrusion to force one to move from a POIR to an SIR, for one cannot define equilibrium so that it is never rational to completely discount others views. Now the distinctive intersubjective dimension of Lehrer=s SIR is its insistence that if one both assigns a positive value to other members of the group at equilibrium and refuses to aggregate, one=s belief is thereby irrational.
We are now in a position to see the connection between the version of the positive value objection which I developed at the beginning of this section and the objection, which I call the anti-aggregation objection, which I shall develop against Lehrer=s SIR. For that earlier formulation can be read as claiming that it is possible to assign others a positive value at the point of equilibrium, and yet it not be reasonable to aggregate. For it is possible that, although one does take the others members to have some expertise, the fact that it is reasonable to believe that one=s epistemic situation is superior to their=s allows it to be rational not to aggregate on the basis of that estimate of their expertise.
The anti-aggregation objection is a plausible objection, and its plausibility can be seen by distinguishing between primary and secondary evidence. ~ prime example of this distinction occurs between testimonial evidence and perceptual evidence. But there are other examples as well, in which the secondary evidence is nontestimonial. One such example uses a deck of cards. Suppose one wished to determine whether a deck was a fair deck, i.e., that it had four suites with the appropriate cards in each suite and no duplication of cards. An indirect way of determining whether the deck is fair is by drawing one card at a time, replacing it after each draw, and keeping records of the results. Thus, if the deck is fair, one would expect roughly the same number of occurrences for each member of the four suites.
But a simpler, more direct procedure would be simply to turn the deck over and organize the deck into suites to see if the deck is fair. Now suppose the evidence provided by the two procedures conflicted; i.e., suppose out of 1,000 draws, the queen of hearts was drawn 400 times, but that, on organizing the deck into suites, it was clear that the deck was fair. In such a case, one=s direct examination of the deck justifies for one that the evidence provided by randomly drawing cards from the deck was misleading evidence concerning the fairness of this deck.
There seems to be a straightforward analogy here between the two types of evidence in the card case and the two types of information available to one at the point of equilibrium. The more direct, primary evidence is the information and the epistemic principles which have been under discussion; the information which results from the values assigned to the other members of the group is secondary. And thus, just as in the card case, isn=t it obvious that it is possible that the primary evidence justifies for one that the secondary evidence is misleading? It should be clearly noted that to say that some evidence is misleading is quite different from saying that there is no such evidence C in the latter case, one would be assigning the other members of the group a value of zero; but in the former case, one does assign them some positive value, one just fails to aggregate on the basis of that assignment. And, it seems to me, such a failure to aggregate is perfectly reasonable in some cases.
To see more clearly that it can be reasonable to believe that the evidence provided by the value one assigns to the other members of the group is misleading, consider the possible bases on which to assign a value to the other members of the group. One can make either a type-expertise assignment or a token-expertise assignment. In the creation of the wants case discussion earlier, one may assign a value to the other members of the group on the basis of their expertise about political philosophy, i.e., on the basis of their expertise about a certain type of issue C those in political philosophy. There are other possible type-expertise assignments in this case as well C perhaps one assigns a value on their expertise about socialism in general, for example. Both of these approaches assign a value to the other members of the group on the basis of their expertise on a certain type of issue; thus the term >type-expertise=. Token-expertise assignments, on the other hand, cannot be assigned on the basis of anything other than the particular issue presently before the group, for otherwise the assignment will, by definition, no longer be a token-expertise assignment. A token-expertise assignment could not, by definition, be inductively supported by reference to a member=s performance in other cases like the present one; for a token-expertise assignment is an assignment based only on the present case and nothing else.
Now, how is one to make assignments of value to thc other members of the group? If one were to make a token-expertise assignment, the fact that the other members of the group disagree with one would seem to indicate that one ought to assign all of them a value of zero. For, at least in the cases I have described, it is reasonable to believe that they are quite mistaken in the way they handle the evidence, and thus, on the basis of this particular issue alone, it is obvious that their testimony concerning what they take the truth to be is not much good. So, one ought to assign a value to them only on the basis of whether they agree with one or not, since any other possible assignment would be completely groundless as a token-expertise assignment.
Lehrer can save his version of an SIR by requiring that token-expertise assignments be made. For, if such assignments are made, there will be no cases in which one assigns a positive value to others and yet one is not rationally committed to aggregation. However, Lehrer insists that this sort of assignment is improper. He says,
It is important to note that there is more than one way to assign a weight to another. One may assign the weight in either an egoistic or a disinterested manner. A person assigns a weight to another egoistically when the weight assigned simply represents how nearly the other person agrees with him ... The disinterested weighter will assign weights to others in terms of how expert and reliable they are in the subject at hand rather than in terms of how closely they agree with him... We require that the weights be assigned in a disinterested manner.9
Thus, since token-expertise assignments can result in the sort of assignments that Lehrer means to rule out in the above quote, it would seem that Lehrer is committed to type-expertise assignments.
But it is precisely on such type-expertise assignments that the anti-aggregation objection is strongest. For, even if another is fairly reliable about a certain type of issue, isn=t it clear that it is possible to rationally believe that the evidence provided by that person=s general reliability is quite misleading in the particular case at hand=?
Situations of this sort occur quite regularly apart from equilibrium. Each of us often takes the time to discuss matters with certain persons precisely because we take them to be fairly reliable about such matters. And, as is sometimes the case, we see that their general reliability is misleading information in the present case, and so we do not alter our views on the basis of their general reliability.
What I am arguing, then, is that the same sort of situation is possible at the point of dialectical equilibrium. It is possible to assign a positive value to someone and yet reasonably believe that that information is misleading, and it is possible for that situation to occur at the point of equilibrium C thereby implying that aggregation is unreasonable. Since any type-event is multiply-instantiable. any restriction on what types on which one is to make a type-expertise assignment will be compatible with one=s having inductive evidence that the other members of the group are somewhat reliable on that type of issue. This inductive evidence would then justify the assignment of some positive value to these other members. But that assignment of positive value is compatible with recognizing that their general reliability on this type of issue is misleading information when attempting to decide what thc truth is about some particular instance of that type of issue. Vhus, if type-expertise assignments are madc to the other members of the group, the anti-aggregation objection shows that no SIR is necessary for rational belief.
IV.CONCLUSION
I conclude then that Lebrer=s version of an intersubjectivity requirement is no more a necessary requirement for having a justified belief than are the other sorts of intersubjectivity requirements that I discussed. If one is inclined to think that there is some sort of subjective intrusion into an intersubjectivity requirement, i.e., if one agrees that the problem of different information undermines a COIR, and that the positive val=ue objection undermines a POIR; I think one is thereby committed to holding that not even Lebrer=s weaker SIR is a necessary condition for having a justified belief.
There is one intersubjectivity requirement that reflects the final intrusion of subjectivity into an intersubjectivity requirement. That requirement claims that if one is convinced that one is more likely to arrive at the truth by aggregating, then it is necessary that one aggregate in order for one=s belief to be justified. This intersubjectivity requirement is a necessary condition for having a justified belief, but that requirement entails nothing that a purely subjective view of justification does not also entail; and thus, accepting that requirement does not require giving up a subjective theory of justification. What I have been suggesting throughout this paper is that no intersubjectivity requirement that entails different conclusions than a purely subjective theory can be adequate C and thus that subjectivity in justification is irreducible. The >us= in >justification= is no more significant than is the >rat= in >crate= C some crates have rats in them, but crates are still crates without the rats.
NOTES
1. Keith Lebrer, Knowledge (Oxford: Oxford University Press, 1974). chapter 8.
2.
Keith Lehrer, >Keith
Lehrer C A
Self-Profile,= in Keith
Lehrer, ed. by R. I. Bogdan, (Boston: D. Reidel, 1981). The view stated in
this quote is also articulated by Lehrer in other places; cf. >Social Consensus and Rational
Agnoiology,= Syn these
31(1975), pp. 141C160;
>When Rational Disagreement is
Impossible,= Nous 10
(1976), pp. 327C332. His
most recent, and most fully articulated, expression is a joint venture with
Cart Wagner, Rational Consensus in Science and Society (Boston:
D. Reidel, 198 1).
3. Self-Profile,= p. 65.
4. This restriction on when nonviolation of rational consensus is necessary for justification was pointed Out to me by Lebrer in his comments on my paper > Lehrer on Rational Consensus= at the Eastern Division meeting of the APA, December 28, 1982.
5. Lehrer, Rational Consensus, pp. 65C66.
6. The epistemic situation of a person at a time consists (St descriptions concerning that person which are relevant to that person=s having all the only epistemically reasonable beliefs at that time. Epistemic reasonableness is to be clarified roughly in terms of a person=s aiming at all and only true beliefs at a particular time. I think that an epistemic situation is composed of all and only the beliefs of a person at a given time, but my argument here does not assume that construal.
7. Lebrer made this claim in his comments on my paper >Lehrer on Rational Consensus.=
8. Epistemic principles are principles which correlate certain states of affairs with the epistemic status of certain propositions. One example of such a principle is: If S is appeared to redly and S has no ground for doubt that something is red, then it is certain for S that somethihg is red.
9. Rational Consensus, pp. 62C63.
Department of Philosophy
Univ. of Notre Dame
Notre Dame, IN 46554
U.S.A.