It's time to respond to Avery Archer's ardent comment to my Hornsby review, which he posted on The Space of Reasons blog last week (see his commentary to my review below). To begin with let me just briefly delineate the issues I want to bring up in this reply to his response. Avery's main challenge to my review seems to be his rejection of knowledge as an factor that has an explanatory role in action as opposed to a mere justificationary role. The difference between these roles Avery explains best himself and it goes as follows:
Explanatory reasons (or what I sometimes prefer to call “motive-giving explanations”) have to do with our attempt to make sense of or explain the purposive activity of intentional agents (rational and non-rational alike). By contrast, justificatory reasons have to do with our attempt to ascertain whether or not the actions of an intentional agent are rationally recommended (and is therefore limited to agents with rational agents). Both explanatory and justificatory reasons are normative (since they both allow for the possibility of error); but while the former answers to a why question (vis-a-vis the actions of any intentional agent) the latter answers to a should question (vis-a-vis the actions of a rational agent).
Thus, for Avery there is two senses in which something can be a reason for action: (1) the reason that explains why someone F-s and (2) the reason that justifies F-ing. Obviously, he thinks knowledge is a fair constraint and necessary requirement on the latter whereas he is opposed to admitting knowledge any explanatory force. Roughly speaking, he positions himself among the orthodox and thinks that what is responsible for the causal workings in the agent can always be explained in terms of more kosher mental states like beliefs and desires. Good, that means we can agree about justification and that we're halfway through to the promised land. Of course, Avery wants to jump off halfway but we'll see about that. For Avery the knower can retreat from the world of causality to linger in a world of rationality and justification. In other words, in the explanatory world knowledge does not belong. This claim about the explanatory futility of knowledge is what I want to focus on in this reply and, of couse, I seek to refute it.
Let me start by granting that Hornsby's paper does not make the case for knowledge's explanatory role to the extent that I would have wanted her to. So her argument may or may not be liable to Avery's criticism that the argument for knowledge's role in action confuses a false claim about its explanatory role from true and obvous claims about its justificatory role in action. Maybe so. But the claim that knowledge has such an explanatory role - and thus the claim that knowledge operates in the agent's mind as a psychologically relevant factor as opposed to some mere rationalistic chimera or ineffective ought - is central if one wants to buttress the view that knowledge sometimes must figure in an explanation of an agent's action. Here Avery and I are in agreement about what goes together with Hornsby's view then; we disagree whether we think her view is falsely bringing these things together. Now let's briefly see why Hornsby needs to commit to the stronger claim that knowledge is an explanatory factor as well.
Hornsby's claim is that in any case where an agent acts for an objective reason to F she must know that this is a reason to be F-ing. Hornsby therefore takes it as a substantial psychological claim that agents sometimes do act out of knowledge. That is to say, in some cases knowledge is psychologically relevant and responsible for the events initiated by the agent. Thus, Hornsby's claim implicates that knowledge makes - at least in some situations, namely those where one acts for objective reasons - a significant contribution to those events. If Avery's right she would have needed to supplement her story, though, since her original story would only concern how the agent is justified in F-ing. What is needed to explain what the agent does is a proper explanation and a psychological story about how the actual F-ing came about. And Avery thinks that knowledge plays no role in the latter.
For people like Hornsby and I, however, justification - or, better, what justifies - sometimes do have explanatory value. In other words, these features sometimes go together. The best way to argue for this intermingling I think one can find in that other source I mentioned in the review for thinking that knowledge is relevant in action explanations, namely Timothy Williamson (2000). What Williamson claims is that knowledge is sometimes relevant in action explanation because the attribution of knowledge to another agent provides one with the relevant generalization to explain the particular action (as well as other cases relevantly similar to this particular one but I'll return to this below). Basically, what is going is that Williamson wants to say that an agent F-es presicely because she knew that P was a reason to F and that this knowledge is essential for giving you the proper generalization for the case at hand. The important thing to notice is namely that in order to explain something - say, someone's action - one must also take account of the case's modale profile. When someone F-es because P there are certain events that could have happened that comprise the modale space surrounding the actual F-ing; and what a proper explanation seeks to do is to generalize and be sensitive to this modal profile. So suppose A would F if she believed that P is a reason to F. Likewise that she would F if she knew that P is a reason to F. What determines the attribution we ouht to make is the case at hand; that is to say, the modal profile we want to get a proper grasp of determines whether we need a belief attribution or a knowledge attribution to explain the case at hand. Williamson's claim is that in some cases, like in the burglar case I mentioned in the previous posts, only a knowledge attribution would do the job.
Let us see if we can get a firmer grasp of this at a more abstract level first. Start by noticing that there is a different modale profile function connected with the attribution of Bp and Kp. With this I mean that belief and knowledge have different properties and thus contributes differently to certain cases. Knowledge is, for instance, factive whereas beliefs admit of being mistaken. That difference is enough to give you a different function for knowledge than for belief, a function that would take you from a given knowledge attribution to a set of values or a structure in modale space for the case at hand; and, analogously, from a given belief attribution to a possibly different set of values.
Sometimes those functions will, of course, deliver the same modale space as a value to a situation. In these situations it would not make a difference whether one attributes Bp or Kp; so maybe conservatives would prefer to attribute Bp in those cases. However, in most cases Kp gives you a different modale profile than Bp. For one thing, Kp, besides from being factive, comes along with a certain required reliability relation: knowledge would not obtain unless all the epistemic alternatives for the agent are p-worlds. That is not the case with Bp. This difference will obviously influence the space of close worlds in a case at hand and thus influence the modal profile one can get by attributing Kp or Bp. If not for oher reasons so for the reason that the kind of reliabiliy associated with knowledge requires a certain ammount of epistemic credibility in terms of evidence. When one attributes Kp to an agent this required evidence follows suit, as it were, and their presence may make a difference to what goes on in this case as opposed to a Bp case.
Given that causality is, if not analyzed modaly, so at least sensitive to modal space, it is evident that knowledge as a cause would bring with it something that could turn out to be useful in certain instances of explanation. That is to say, knowledge as a cause could in some cases be what is needed to get a desired generalization or modale profile to a given case. The very existence of such cases is Williamson's crucial claim. Now that we have the general framework at hand we can perhaps better perceive this possibility and be tempted to use it. In any case, whether it is actually relevant or not requires a convincing case and it is here I think that the burglar case suffices.
Before we go on let me note in passing that what Avery's thinks is a point in favour of his dividing explanatory reasons from justificatory reasons is that he thinks knowledge fail to apply to non-linguistic animals who may still be said to act for reasons. However, this point could actually be flipped in Honsby's (or my) favour. Avery seems to think that animals could act for the belief that p whereas knowledge would require linguistic abilities in its possessor. Thus, Kp may be a justificatory requirement, which the animal cannot satisfy, whereas what moves the animal to act and thus explain its behaviour is Bp. I'm aware that there is such a view about knowledge; but one should be aware that there's also such a linguistic view about beliefs (e.g. Davidson, Dummett). None of which I would be inclined to give very much credit. More plausibly, in my view, beliefs do require certain language-like cognitive capacities since beliefs take propositions as their objects, which are object built up and grasped by the proper combination of concepts. They work much the same way as one would determine the meaning of a sentence from the combination of its constituent words; thus we get something like Fodor' postulation of the language of thought. That I think is needed for having a belief; and I hasten to add that animals who act for reasons they have beliefs about must have this capacity. For Hornsby, however, and I think I agree, knowledge is a non-linguistic relation to facts. So there really is no problem for Hornsby or me that animals can act for the objective reason that P; at least as long as this non-linguistic knowledge relation is available to us. On the other hand, the requirement of a language of thought for having beliefs could equally well support the presence of propositional knowledge, in the old-fashioned sense, in an animal. In any form, then, knowledge - either propositional or merely "factive" - could be applied to non-linguistic animals whom we are prone to bestow with a reason-giving explanation. Thus, the presence of these explanations does not pry apart explanatory and justificatory reasons in the way that Avery claims they do.
Back to the talk about modale profiles: I won't do the burglar case again but suffice it to say that Avery fails to pay sufficient attention to the generalization that comes with knowledge attributions in reason-giving explanations; although he poses a hallenge to my kind of view to say something about how one should do the generalization. For in order to prove the necessity of knowledge in action explanations two thoughts must be kept in mind: (1) that the Kp attribution matches the modale profile in the case at hand and (2) that Kp attributions is generalizable to a certain range of cases to serve as a unificatory explanation for those cases. Avery seems to agree that a Kp attribution can match the Burglar Case in the first sense; but then he goes on to provide us with a similar-looking case where it fails to apply since the burglar in this similar-looking case would do the same thing despite her obvious lack of knowledge. However, that knowledge lacks in Avery's case is no reason to think that knowledge is not the cause of the burglar's behaviour in the former. What would be required to show that knowledge fails to be the cause is a case where knowledge obtains without the desired effect and the additional argument that such a case is close enough and relevant to take down the claim of causation. Avery's case is more like saying that "Shakespeare did not write Macbeth since there is a doppelganger case where Macbeth gets written but by Marlowe." And even granted the presence of such a case Shakespeare surely wrote Macbeth and is the cause of Macbeth's existence. Likewise, there could be dozens of cases where the burglar acts in the same way but for other reasons than that he knew P. They say nothing whatsoever about whether Kp is the cause in our case for his behaviour (but maybe I'm missing something here?). And I think it is the cause in our case because it matches up with its modale profile. That is the claim to be challenged if one wants to turn down the burglar case.
One way to do that is to suggest, as Avery mentions, an alternative attitude. Suppose believing P with certainty matches the burglar case's modale profile. Then the question is why we should think of knowledge as the cause as opposed to the mere certain belief. Here one could cite certain rational constraints, of course, but not in order to slide away from explanation and over into a different topic, as Avery says we do, to begin discussing justification. The point is rather that we begin to pay attention to the other horn of our explanatory scheme, namely relevant generalization. We want to capture a certain range of cases by the use of our knowledge attributions. That range or class of cases is determined by considerations of rationality, I admit, but that's clearly different from saying that what we want to talk about is how the agent is justified to act. Rather, we want to explain this group of cases as they ar given to us by considerations about rationality.
So my final point is that Avery's certain belief that P may take on this-and-that particular case but surely would fail others where knowledge that P seems called for, i.e. those cases we get via considerations about rationality. We should, of course, wish for the richer and more general explanation and thus my claim is that knowledge will turn out to be more ammendable to the range of cases we want to explain. In the end this is what makes knowledge our candidate for explainging certain cases, namely all those cases where someone acts for the objective reason that P. Of course, that's a research project and not just for a single case or paper to establish. So we should definitely return to this issue.
The knower belongs in the world, I say. Otherwise there's is no way we can make sense of what we want to make sense of, namely those cases delineated by considerations about rationality where an agent acts for the objective reason that P.