Epistemology

The literature on the Epistemology of Disagreement has been primarily characterized by two opposing views: the steadfast and conciliatory. The steadfast view proposes that, in the context of peer disagreement, certain factors allow one to hold one’s ground. That is to say, the evidence provided by peer disagreement is weak—not something that would generally require belief revision. The conciliatory position, on the other hand, places great weight on the evidence provided by the fact that two epistemic peers disagree. The conciliationist bases the particular weight given to peer disagreement on the principle that if epistemic parity is indeed maintained, one should have no reason to discount one’s peer. Provided this, the question remains as to how one might certify an epistemic peer in such a way as to prevent question-begging downgrades based on the content of the disagreement itself. Given this problem, the conciliationist requires a regulatory principle: independence. This principle certifies that the evaluation of an epistemic peer should take place antecedently to the disagreement in order to prevent such downgrades of parity. It will be the purpose of this paper to argue that the current conciliationist resolution to certain hard cases of disagreement either violates or trivializes this principle. To establish these results, I will (1) outline the principle in full detail, (2) provide an example case in which independence becomes problematic, (3) detail the arguments provided to resolve this case in which independence is purportedly maintained, (4) argue that these results either violate independence outright or force modifications to the principle in such a way that it becomes epistemically vacuous, and (5) develop some possible alterations available to the conciliationist that might salvage the independence principle.

The most recent concilliationist formulation of independence is provided by David Christensen (draft, 2008), the principle is stated thus:

Independence. In evaluating the epistemic credentials of another’s expressed belief about P, in order to determine how (or whether) to modify my own belief about P, I should do so in a way that doesn’t rely on the reasoning behind my initial belief about P.¹

It is important to note here that independence as stated provides a necessary condition on epistemic parity: parity must be established independently of one’s reasoning about the content of the disagreement. In order to establish independence of this nature, Adam Elga (2006) provides necessary and sufficient conditions on epistemic parity thus:

A counts B as an epistemic peer with respect to an about-to-be judged claim if and only if S thinks that, conditional on A and B disagreeing about the claim, A and B are equally likely to be mistaken.²

These conditions specify that parity must be established antecedently to the disagreement, and such parity is represented by the equality:³

(EP) Pr(A is right whether p | A and B disagree whether p) = Pr(B is right whether p | A and B disagree whether p)⁴

Given this characterization, nothing has yet been said regarding the factors upon which such a parity evaluation must stand. Christensen, however, has recently provided two tentative requirements:

(A)  One’s dispute-independent assessment of her epistemic credentials yields high estimates for:

1. The likelihood that her expressed disagreement is sincere,
2. Her degree of informedness, and
3. The likelihood of her having reasoned correctly from the evidence she has;

and,

(B) The reasons for one’s assessments of (a)-(c) are strong.⁵

Given these requirements, (PH) can be reformulated thus:

(EP’) Pr(A is right whether p | A is sincere & A is well informed with respect to p & A is basing her reasoning on her evidence) = Pr(B is right whether p | B is sincere & B is well informed with respect to p & B is basing her reasoning on her evidence)

So, if one has antecedently judged that one’s interlocutor satisfies (PH’), and such a judgment satisfies condition (B), which is just to say that one is doxastically justified in one’s assessment of (EP’), one is then in a position to split-the-difference when a disagreement with respect to p arises⁶.

Given the preceding characterization of the conciliatory position, I will now outline a (very familiar) case that serves to problematize the position:

Splitting the Check. Suppose that I am going to dinner with my friend. We often dine together, and she has never given me reason to doubt her abilities to perform mathematical calculations on the fly. We often split the bill, and on this occasion we have decided to do just that. On previous occasions, we have come to agreement over the price of our shares, and my friend has a very good track record with respect to splitting the check. On this occasion, however, I come to the conclusion that the split is $43, and she comes to the conclusion that the split is $430.⁷

Given this case, and provided that there are no antecedent consideration relative to the context of disagreement that would cause me to downgrade my compatriots status as an epistemic peer, such as the consumption of a few bottles of wine, it would appear that (PH’) would easily be established pre-disagreement. Further, given the track record we have relative to this kind of occasion, it is conceivable that I am in fact justified in my assessment. But this seems to be a case where my compatriot is clearly in error. And we can fill out the background information of this case to extenuate this error: consider, for example, that no item on the menu is priced higher than $25, and we have not been overindulgent and ordered multiple meals. Given these considerations, it would seem that the conciliationist would require me to lower my credence in the proposition that the split is $43. Neither Elga nor Christensen sit comfortably with this result.

In hard cases such as this, the conciliationist appeals to counterfactual considerations in order to break the peerhood symmetry. The basic line is that, while the original probability function in the right-hand-side of (EP’) provides an equality, there is a possible alternative probability function Pr_C(B is right whether p | B is sincere & B is well informed with respect to p & B is basing her reasoning on her evidence & B is malfunctioning cognitively) that does not satisfy this equality. Given that were I to conditionalize in this way, I would not consider B my epistemic peer, it is clear that in cases where my compatriot’s answer is in fact crazy, I should downgrade her peerhood status.⁸ Christensen makes this move by an appeal to personal information, a concept introduced by Lackey (forthcoming), where it is observed that one has access to additional evidence about “the normal functioning of one’s own cognitive faculties.”⁹ Christensen further extends this kind of information to include information regarding your estimation of the sincerity of your judgement. These considerations lead to the ability to asses the various aspects of the conditionalization relative to the contextual information available after the disagreement has occurred, and given these considerations, were these circumstances to arise, I would again be able to downgrade the peerhood status of my compatriot by instantiating a different probability function than that of my original parity judgment.¹⁰

Both Elga and Christensen claim that this kind of downgrading is independent of the content of the disagreement. This is an odd claim. Looking diachronically at the Splitting the Check case, it would seem that we have the following temporal ordering of events:

t₀: A and B agree to split the check, and calculate their respective answers to the question whether p. Given the information available, A and B then make a parity judgment relative to (EP’).

t₁: A proposes that the split is $43, and B proposes that the split is $430.

t₂: Given the new information available, and realizing that something has gone horribly awry, A makes a judgment to the effect that B’s peerhood status with respect to A is downgraded.

Now, given that t₁ represents the content of the disagreement, it is difficult to see how a judgment made by A at t₂ is independent of the said content. For without the new information introduced by t₁, there would be no reason to change B’s parity status. Claiming that the information upon which A made a parity judgment is purely contextual, and thus not relevant to the content, does not help either, for any relevant contextual pre-disagreement information would be exhausted in the initial parity judgment. Any additional information would have to be inferred from the content of the disagreement, and an inference based on the content of the disagreement fails to satisfy the condition in Independence such that one’s reasoning be independent of the reasoning behind one’s initial belief that p. For the content of the disagreement is based on one’s initial reasoning about p, and the reasoning involved in t₃ is based on the content of disagreement. It is simply one step removed.

Even if we were to grant that independence is not violated in this case, what would prevent one from, by parity of reasoning, making a counterfactual judgment such that an alternative, equality violating, probability function is introduced: Pr_QB(B is right whether p | B is sincere & B is well informed with respect to p & B is basing her reasoning on her evidence & B is mistaken whether p). It would seem that, in any case of disagreement, one could simply appeal post-disagreement to a probability function such as this in order to downgrade the status of one’s peer. If the purpose of Independence were to prevent question-begging maneuvers such as this, it would seem that it has ceased to function as a regulatory principle. Thus, it seems that either the conciliationist has violated independence by a chain of reasoning based indirectly on the initial reasoning about the question whether p, or the regulatory function of the principle itself has been trivialized.¹¹

How, then, should we consider the claim made both by Christensen and Elga that the line of reasoning adopted in the above case is independent of the reasoning about the disagreement? It looks initially like this kind of counterfactual consideration could possibly be some kind of safety or sensitivity constraint on one’s epistemic parity judgment, and, if so, taking into consideration the above constraints, we might have some necessary and sufficient conditions for judgments of epistemic parity like this:

(JEP) S judges S’ an epistemic peer iff

(1) (EP’) is satisfied,

(2) S is justified in (1),

(3) S bases S’s (1)-Judgment on (2), and

(4) If S’ were not an epistemic peer of S, S would not judge S’ to be an epistemic peer.

Now, in this case (4) is a sensitivity requirement. In other words, S’s judgment of epistemic parity must align with what is in fact the case regarding S’. So, in the Splitting the Check case, S would violate (4) if S did not pick up on some symmetry breaking contextual information that violates (1) prior to the disagreement occurring. (JEP) looks like it accomplishes that which the conciliationist is attempting with his response to the Splitting the Check case, but I think it is going to end out too strong. The force of the previous case is just that there is no antecedent information, contextual or otherwise, that breaks the symmetry. It is only post-disagreement that one finds that there is some problem with S’. Thus, (4) looks to be too easy to violate. In any normal case of disagreement, where (1), (2), and (3) are satisfied, there will be possible worlds in which S’ has some sort of hidden cognitive malfunction which delivers illicit answers to the question under discussion.

Given the preceding considerations, perhaps (JEP) might benefit from a safety condition:

(4′) If S were to judge S’ to be an epistemic peer, S’ would be an epistemic peer of S.

If satisfaction of (EP’) is our definition of epistemic peer, however, then it looks like we run into problems similar to the sensitivity constraint. To be clear, the reasoning in (1) comes antecedently to the disagreement. In all of those cases where there is some hidden form of cognitive malfunction, the state of information previous to the disagreement is such that S would evaluate S’ correctly relative to (EP’). It is only after the judgment that the additional information becomes available, and it is only relative to this information-state that the safety constraint would become substantive. But this is not the point at which the parity judgment takes place.

In conclusion, I have developed some conditions on epistemic parity and epistemic parity judgments amenable to the conciliationist, but it is not initially clear how these judgments successfully navigate the hard cases of disagreement in a way that does not violate or trivialize Independence.