Simple Reliabilism and Agent Reliabilism

John Greco offers a version of virtue epistemology which attempts to include the good features of reliabilism and yet avoid the problems facing that theory. The most intriguing element of his approach is its theoretical unity. Whereas other theorists have claimed that both objective and subjective justification are needed to avoid the problems of simple reliabilism, none have proposed a unified theory of both notions that is generally reliabilist in spirit. Greco proposes to do just this by offering a theory that gives a unified, character-based account of both objective and subjective justification. This theory is reliabilist in spirit while maintaining that subjective justification is required for knowledge. Such a proposal is unique and ingenious, and deserves careful study because of it.

Though I find the project significant and unprecedented in this way, I am not convinced that it is entirely successful, and I will try to explain here the grounds of my concern. We can begin with Greco’s list of requirements for an adequate theory of knowledge, and the relationship he sees between simple reliabilism and his own theory, agent reliabilism.

Greco desires a theory of knowledge with five features. (1) It needs to imply that not all evidential relations are inferential, and in particular (2) why sensory evidence can be non-inferential. (3) It must also be a foundational theory of knowledge. Finally, (4) it must reveal how inferences that are only contingently reliable can yield knowledge, and (5) how knowers can be sensitive to the reliability of their inferences (p. 164).

Greco believes that simple reliabilism, the view that identifies knowledge with beliefs produced by reliable methods or processes, explains adequately the first four claims. He holds, however, that agent reliabilism is needed in order to account for the last. These two central claims are at the heart of my worries about Greco’s project.

On the first point, I think Greco misleadingly endorses simple reliabilism as an explanation of (1)-(4) because it implies all of them. It is a foundational theory of knowledge, at least in its formal structure (some beliefs are known apart from being based on other beliefs), and this foundational knowledge can be sensory knowledge that is non-inferential. It also implies that knowledge can arise from contingently reliable processes or methods.

We might be willing to grant that no problems arise with the reliabilist explanation of (1)-(3), but we shouldn’t grant the adequacy of the explanation of (4). For simple reliabilism does not provide a theoretical explanation of (4) unless knowledge is present in a number of cases where we know it isn’t. An explanation is not adequate simply because it implies a number of things that it ought to imply. If that were true, we could theoretically explain everything by citing a contradiction. An adequate theoretical explanation must not only imply things that are true, but also fail to imply things that are false. So simple reliabilism is an adequate explanation of (4) only if knowledge is present in BonJour’s clairvoyance case, in Plantinga’s serendipitous lesion case, and in a number of other cases that constitute counterexamples to simple reliabilism.

This problem will be innocuous enough if Greco’s way of supplementing simple reliabilism so as to yield agent reliabilism eliminates these counterexamples. Perhaps this is the way Greco conceives the relationship between simple and agent reliabilism; at the very least, this way of conceiving the relationship would explain the perplexing dialectic in Greco’s discussion. Greco’s official account of the relationship between (1)-(5), and simple and agent reliabilism is as follows: simple reliabilism explains (1)-(4), and agent reliabilism adds something to simple reliabilism to enable an explanation of (5) as well. Yet, Greco introduces agent reliabilism by considering the problem of “strange and fleeting” processes, processes such as Plantinga’s serendipitous lesion that happens to cause a true belief that such a lesion is present or BonJour’s clairvoyant who forms (true) beliefs based on clairvoyance even though he knows better. When Greco finally turns to (5), the explanatory work is done by a distinction between subjective and objective justification, which I will comment on below. Yet, if simple reliabilism adequately accounted for (1)-(4), why doesn’t Greco simply jump to the agent reliabilist’s explanation of (5) in terms of the distinction between subjective and objective justification?

The answer may be that Greco knows that simple reliabilism’s explanation of (4) is inadequate, and that the first thing agent reliabilism must do is to offer a better explanation of (4), one lacking the untoward implications of simple reliabilism regarding strange and fleeting processes such as those involved in the Plantinga and BonJour cases. If my suspicions are correct that it is this awareness that explains Greco’s dialectic, then Greco sees that two tasks, and not just one, are demanded of agent reliabilism: to correct the explanation of (4) given by simple reliabilism, and to explain (5).

I don’t think agent reliabilism succeeds on either score.

According to Greco, agent reliabilism claims that “A belief p has positive epistemic status for a person S just in case S’s believing p results from stable and reliable dispositions that make up S’s cognitive character.” (p. 177) The account involves two ideas not present in simple reliabilism. First, the disposition that results in belief must be part of S’s cognitive character. Second, in addition to being a reliable disposition, the disposition must be “stable.”

What is it for a disposition to be stable? Greco introduces agent reliabilism to solve the problem for simple reliabilism raised by “strange and fleeting” processes, one of which is Plantinga’s serendipitous lesion example. Stability of a disposition does not seem incompatible with strangeness (and Plantinga’s lesion is certainly strange), but it is to be contrasted with fleetingness. Yet, the problem of strange and fleeting processes is not centrally a problem about what period of time the process is operative. Plantinga’s lesion might have been present from birth; BonJour’s clairvoyant might have formed clairvoyant beliefs most of his or her life. Making the dispositions in question more permanent in the cognitive lives of the persons in question does nothing to relieve the force of the counterexamples. So adding stability doesn’t seem to help.

Moreover, stability doesn’t seem required either. Suppose God is in a playful mood when creating Adam, and suppose a disposition for belief is stable only if it lasts for n units of time. God’s playfulness leads him to consider the array of sets of dispositions toward belief, and to change the set Adam has over the course of his life. How often does he do this? For some m, such that 0_m_n, God changes the set every n-m units of time. He begins by giving Adam all the ordinary dispositions toward belief that we all have. So Adam looks at a ripe pear on a tree and believes That is yellow. Adam has perceptual knowledge of the color of the pear, but Greco’s account implies that Adam doesn’t have such knowledge, because God is being too playful with his creation. There may be possibilities of playfulness that are incompatible with knowledge, but the sort in question here isn’t one of them. Those familiar with the inevitable degradation of our cognitive equipment, whether graceful or otherwise, will see any appeal to stability of dispositions toward belief as a red herring. For some humans the degradation occurs over decades, for others much more quickly. The facts about present knowledge, however, don’t seem connected at all with how long we will maintain the integrity of our cognitive mechanisms.

The second difference between Greco’s agent reliabilism and simple reliabilism appealed to the person’s cognitive character, but this condition is of no help here. For Adam’s cognitive character is constituted by the set of dispositions during the time he has those dispositions. The only problem is that Adam doesn’t have those dispositions long enough for Greco’s theory to allow him perceptual knowledge of a very simple sort.

It is interesting to note in this regard that Greco introduces examples of “strange and fleeting” processes of belief formation to motivate his appeal to stability, but never uses that concept to explain why knowledge isn’t present in the examples of BonJour and Plantinga. I think that fact is telling, because lack of stability is not what is missing in those cases. What goes wrong is, rather, a matter of lack of coherence between the understanding a person has of the reliability of various ways of forming and holding beliefs and how the beliefs are formed in these cases. What goes wrong, in the language of Sosa, for example, is that there is no reflective knowledge in such cases.

Greco’s explanation of (5) touches on this very point. So we might hope that his explanation of (5) will simultaneously solve the problem simple reliabilism has in explaining (4). Greco’s explanation of (5) centers on his distinction between subjective and objective justification. According to Greco, objective justification amounts to a belief being the result of dispositions that make a person reliable regarding that belief in the conditions in question, and subjective justification involves a belief being “the result of dispositions that S manifests when S is thinking conscientiously” (p. 218). The hope is that this combination of positions will block the difficulties that confront other versions of reliabilism.

To see if the attempt is successful, we need to look more carefully at Greco’s remarks about the dispositions involved in conscientious thinking. First, Greco identifies conscientious thinking with the “default mode” of being motivated to get to the truth (p. 191). The contrast is thus between thinking honestly and being motivated by non-alethic factors such as greed, prestige, comfort, and the like.

Second, Greco does not identify subjective justification with such proper epistemic motivation; in fact, his definition does not even require proper motivation. For his definition only requires the activity of the dispositions that are present when one is properly motivated, and those dispositions might be active both when one is properly motivated and when one is not. I think this may be too weak a connection between subjective justification and proper epistemic motivation. For example, if the same disposition can accompany both well-motivated and ill-motivated belief, it will be possible to display that disposition ill-motivatedly while rationally believing that an ill-motivated display on this occasion is highly unlikely to get one to the truth. Such a situation strikes me as paradigmatic for lack of subjective justification, rather than one in which subjective justification is present, as Greco’s theory implies.

Moreover, Greco’s appeal to subjective justification does not explain away the standard counterexamples to simple reliabilism. In BonJour’s clairvoyant case, Greco needs to charge the clairvoyant with not manifesting the dispositions that he manifests when trying for the truth. I don’t see why the clairvoyant has to be guilty of this charge. The clairvoyant knows better than to trust claivoyance, but that doesn’t imply that he is manifesting dispositions different from the ones manifested for him when aiming for truth. The clairvoyant’s failure is that he doesn’t take into account possessed defeating information, but one can fail to be do this, and even fail to be disposed to do this, and nonetheless manifest the dispositions one normally does when proceeding honestly.

Most of us do not behave this way cognitively, but that is a contingent fact about us. As we improve cognitively, we learn to monitor for defeating information and we learn to withhold belief when we learn of the presence of defeating information. Even so, in the process of so improving, we often think honestly and display the dispositions that we ordinarily display when honestly trying for the truth without monitoring for defeating information and without withholding belief in the presence of known defeaters. Greco’s response to the clairvoyance case requires that the only explanation for retaining a belief in the presence of known defeaters is that we are being moved by dispositions other than those operative when thinking honestly, but there are other options. Habits are often overly general, displaying themselves even where not especially useful or desirable. Transparently honest people sometimes hurt others’ feelings by unthinkingly displaying such honesty, and belief formation can exemplify this same feature. The motivations present when one is honestly trying for the truth might be unthinkingly displayed when attention to the presence of known defeaters would have prevented the display of these motivations.

This point is not merely a trifling failure of detail, but rather a particular instantiation of a more general weakness of reliabilism. On the face of it, justification is a function of the information or evidence we possess, but reliabilism wishes to talk in terms of belief-forming mechanisms and so must try to mimic evidential relations with these mechanisms (say, by individuating mechanisms or character traits in such a way that the reliability of these mechanisms coincides with the intuitive idea of information or evidence possessed). I have argued repeatedly that the prospects for successful mimicry are hopeless, and my point here is that the above difficulty is simply another example of the difficulty. The concept of defeating information is, intuitively, one concerning the epistemic relationships between semantic contents, and the hope of any version of reliabilism is to mimic these relationships by appeal to the right kinds of mechanisms or character traits. Greco’s attempt on this point is not entirely successful, I think, for there are no grounds for thinking that in order to display dispositions involved in trying for the truth, one must be disposed always to avoid belief when defeating information is present. So there is no reason to think that the clairvoyant case is explained away by an appeal to such dispositions.

Let me hasten to add that I think Greco’s approach to these counterexamples to simple reliabilism is on the right track. In these cases, the persons involved have beliefs that are subjectively unjustified, preventing their beliefs from being known to be true. The problem is that Greco’s account of subjective justification doesn’t yield this implication. A more general point is that it is highly unlikely that any disposition-based proposal can yield an adequate response to the counterexamples to simple reliabilism. Dispositions, by their very nature, can be overly general, and hence will be able to be displayed while failing to take into account the presence of defeating information. Not even an appeal to the disposition to take into account such information can help, for one needn’t have that disposition in order to have subjectively justified beliefs–sensitivity to such information is learned behavior, behavior that can be learned by forming subjectively justified beliefs which turn out to be false.

If I’m right in concluding that no disposition-based account of subjective justification can be adequate, Greco will have to abandon one of the most attractive features of his account. Greco’s hope is to use features of character to elucidate the concept of subjective justification, thereby unifying his account of objective and subjective justification in terms of an appeal to cognitive character. If I am correct, however, he’ll have to sacrifice theoretical unity in order to find an adequate account of subjective justification.