In two recent papers, Jaegwon KimFootnote 1 offers an account of action explanation that is agent-centered and reasons-based, as opposed to causal and nomological. The account, which is inspired by William Dray,Footnote 2 focuses on simulating the way the agent saw her own action in the light of her reasons and principles of action. Kim’s proposal is interesting in itself, but what is really surprising is that the simulation model appears to imply a pluralist account of explanation that undermines his own highly influential principle of explanatory exclusion, or at least his use of it against nonreductive physicalism. At times, Kim gives the impression that the simulation model is either orthogonal to the exclusion problem or, if related to the issue, has no significant bearing on it. I disagree and will show in the discussion that Kim’s treatment of the simulation model opens the door to a solution to the problem of explanatory exclusion and suggests an avenue of escape from his more recent supervenience/exclusion argument.
My discussion is divided into five sections. In §1, I briefly sketch out Kim’s support for, and application of, the principle of explanatory exclusion. In §2, I outline Kim’s simulation model and how he thinks it differs from the causal-nomological model. In §3, I provide a brief characterization of dual explanandum replies to the principle of explanatory exclusion and explain how Kim’s simulation model lends itself to a variation on the dual explanandum strategy. My support for this point will focus primarily on Kim’s treatment of the explanatory relation in his account of the simulation model and his characterization of first- and third-person “stances.” In §4, I consider and respond to the ways in which Kim tries to resist the conclusion drawn in §3. Finally, in §5, I explain how the general view sketched out and defended in this paper provides an avenue for nonreductive physicalists to escape Kim’s more recent supervenience argument. Thus, my goal is to show not only that Kim’s recent work on first person accounts of action explanation undermines his argument from explanatory exclusion, but also his supervenience argument.
§1. The Principle of Explanatory Exclusion
Kim adopts what he calls “explanatory realism,” which claims that a proposition C is an explanans for proposition E in virtue of there being an objective relation R holding between the events the propositions are about (c and e). Kim therefore takes R to be the explanatory relation that “grounds” the explanans relation between the propositions that constitute an explanation. Since the most prominent species of explanation is causal explanation, Kim claims that the most likely candidate to fulfill the role of R is the causal relation itself in such cases. Hence, he thinks that explanatory realism requires causal realism, the view that causal relations are mind-independent relations between events in the world.Footnote 3 Bringing explanatory realism and causal realism together, Kim defines having an explanation in terms of the possession of causal knowledge: “To ‘have an explanation’ of event e in terms of event c is to know, or somehow represent, that c caused e.”Footnote 4
Kim’s account of causal realism and explanatory realism together yield a criterion of individuation for explanations that plays a central role in his support for the principle of explanatory exclusion. He locates the individuating role of explanations in the explanatory relation that grounds explanations:
Explanatory realism yields a natural way of individuating explanations: explanations are individuated in terms of the events related by the explanatory relation (the causal relation, for explanations of events). For on realism it is the objective relationship between events that ultimately grounds explanations and constitutes their objective content. This provides us with a basis for regarding explanations that appeal to the same events standing in the same relation as giving, or stating, one explanation, not two—just as two inequivalent descriptions can represent the same fact.Footnote 5
This is a fully extensional criterion of individuation for explanations. Logically distinct descriptions of the cause or the effect in explanatory claims will state the same explanation if they are “grounded” in the same metaphysical relation R between the same pair of events. If two explanations refer to distinct pairs of events or potentially even different relations between the same pair of events, it seems reasonable to treat the explanations as distinct. Alternatively, if we are referring to the same relation between the same two events in two putative explanations, it is not implausible to treat the explanations as equivalent. This provides prima facie support for Kim’s principle of explanatory exclusion, which states, “there can be no more than a single complete and independent explanation of any one event.”Footnote 6 In cases where we seem to have more than one explanation because, for example, the cause can be picked out under multiple descriptions, it turns out that we, in fact, have only one explanation since the “alternative” explanations are grounded in the same causal relation.
Although Kim intended the principle of explanatory exclusion to have a broad application, he has utilized it primarily to place conceptual pressure on nonreductive physicalism. The argument is elegant in its simplicity. If we take a bit of intentional human behaviour, such as George’s rising from the couch (B), to be a physical movement, it must have a complete explanation in physical, perhaps neurobiological terms. However, it seems as though we can also explain this behaviour by appealing to George’s psychological states, such as his desire for a beer. Hence, we have one causal explanation that appeals to the occurrence of a neurobiological property, and another causal explanation that appeals to the occurrence of a psychological (content) property. In Kim’s own words,
Thus, we have two causal claims about a single event, B: (1) c’s content property R caused B and (2) c*’s having neural property N caused B… When these two claims are viewed together, we should find the situation perplexing and somewhat unsettling. Why did George get up from the couch? What caused it? We are offered two causal explanations: ‘Because c occurred and had R’ and ‘Because c* occurred and had N.’ But what is the relationship between c and c*, or between c’s having R and c*’s having N? George’s reason, his desire for beer and his belief about where he could find some, is offered as what made him get up; but then a certain neurophysiological event of kind N is also offered as what made him get up. We want to ask: ‘Which really did it? What’s the real story?’ Footnote 7
For Kim, as one might expect, given his commitment to explanatory realism, the real story is the one that is grounded in the objective relation between the identified events: in this case the causal relation. Given the assumption of physicalism and causal closure, we have little reason to doubt the veracity or the completeness of the explanation that appeals to the neurobiological property, in which case this explanation threatens to exclude the alternative causal explanation that appeals to the content property.Footnote 8 This tension could perhaps be remedied if it could be shown that c’s being R = c*’s being N, but, according to Kim, this would require us to accept that property R = property N, which is just what nonreductive physicalists deny.Footnote 9 Thus, although a reductionist can account for the efficacy of the content property by identifying it with the neurobiological property, the nonreductive physicalist seems to be stuck with type epiphenomenalism: the inefficacy of mental properties. This is unfortunate since there seems to be little reason to maintain the distinctness of mental properties if they have no causal role to play in our behaviour.
§2. Kim’s Simulation Model of Action Explanation
In “Reasons and the First Person” and “Taking the Agent’s Point of View Seriously in Action Explanation,” Kim sketches out an account of how agents understand their own actions and then offers an approach to action explanation that draws on this model. In both papers Kim contrasts the agent-centered “first-person” approach, which is influenced by William Dray, with what he calls the “third-person” approach of philosophers like Hempel and Davidson. The fundamental difference between the two is that third-person approaches are predictive about what the agent will do whereas the first-person approach emphasizes the way the agent understands her own behaviour in the context of choice and deliberation, not prediction. Interestingly, both views make reference to what Davidson called the agent’s primary reason for acting.
The most influential third-person approach is what Kim calls the nomological version of the causal account. This view takes the explanatory relation between an agent’s primary reason and action to be a matter of their joint subsumption under contingent empirical laws. Hence, to explain an agent’s action by appealing to her primary reason involves showing how the action is entailed by the presence of the reason and strict or statistical laws. In the former case, action explanation takes the form of a deductive nomological explanation, whereas in the latter it takes the form of an inductive statistical explanation. In Hempel’s version, the laws are explicit and are formulated in the same general terms as our initial description of the agent’s behaviour, whereas in Davidson’s case identification of the relevant laws requires shifting from mental to physical vocabulary since he denies there are strict psychological (or psychophysical) laws. Either way, though, Kim maintains that the purpose of such explanations is to give us enough information to predict the explanandum event (i.e., the action).
In Kim’s view, although this approach can be very successful at yielding accurate predictions of human behaviour, this is not how self-understanding works. When we try to make sense of our own actions, we are not trying to predict what we will do, but to decide what we will do. Furthermore, Kim argues that Hempel’s and Davidson’s accounts involve an implausible epistemology of action explanation. With Hempel’s approach, we need access to behavioural-psychological laws to make rational sense of our own actions, yet it seems that we do not have ready access to such laws; on Davidson’s account we need access to underlying strict neurological laws, and it is clear that we don’t have access to those.Footnote 10
Kim’s alternative first-person approach to self-understanding involves the following general claim:
To know, or be able to reconstruct, the practical deliberation that led me to decide to do something (and actually do it) is a fully satisfactory way of understanding why I decided, and acted, as I did.Footnote 11
Mimicking Davidson’s claim that placing an event in the context of its cause explains the event, Kim claims that placing an action in the context of its deliberation explains the action in a sense of “explain” that we understand as well as any. All we require for understanding our own actions is a variety of what Kim calls “subjective intelligibility”, which is achieved by reconstructing our practical deliberations in accordance with the following schema:
I am in circumstances of kind C
In circumstances of kind C, the appropriate thing to do is X
So I ought to do X
(And you do X)Footnote 12
This, of course, requires that the agent actually knows she is in circumstances of kind C, but this is much less onerous than having to know behavioural-psychological laws or neurological laws. This also treats the agent’s reasons for acting as “internal reasons”, since they must be accessible to the agent. Kim fleshes out this model of self-understanding with the following example. Suppose you want to know why you are carrying your umbrella on a sunny morning. All you need to do is to reconstruct your practical reasoning in accordance with the above schema:
It’s going to rain this afternoon
Unless I have an umbrella with me, I will get wet
I don’t want to get wet
In this situation, the right thing to do is to take an umbrella
Therefore, I ought to take an umbrella
(And you do take an umbrella)Footnote 13
Once you have reconstructed your deliberation in this way, you have a perfectly acceptable explanation for what you are doing.
In “Taking the Agent’s Point of View Seriously in Action Explanation,” Kim claims that the above way of explaining is not restricted to attempts to understand our own behaviour, but can be applied to the behaviour of other agents as well. This is simply a matter of putting one’s self in the agent’s place and running through the above process “off-line.” In Kim’s own words, “… I simulate the agent’s practical deliberation by imagining myself in the agent’s situation and asking myself what is the appropriate thing to do—appropriate for me to do.”Footnote 14 For this to succeed, one must come out close to where the agent did, and, although one must bring one’s own principles of action to bear on the situation, in order for the deliberation to come to the same conclusion as the agent’s, there must be broadly shared principles of action. When these conditions are met, the understanding arises from one’s own simulation. Hence, Kim thinks there is a distinct way of explaining what an agent does that is agent-centered and takes the first-person perspective seriously in action explanation.
§3. The Dual Explanandum Strategy
One popular response to the problem of explanatory exclusion is to relieve the tension between the neurobiological explanation and the psychological explanation by proposing that they explain distinct things. According to this approach, the tension identified in the principle of explanatory exclusion is genuine only if the explanations are competing to explain the same explanandum. If the neurobiological explanation seeks to explain a different explanandum than the psychological explanation, there is no tension and the threat of exclusion is averted. Thus, one promising way to resolve the problem of explanatory exclusion is to propose that when we talk about explaining human behaviour we are often not as circumspect as we should be. If we are more precise, we will see that there are two explananda in need of explaining, in which case the psychological and neurobiological explanations each have their own distinct jobs to do.
There have been two influential versions of the dual explanandum reply, each of which divides the explanandum in a different way. The first, developed by Fred Dretske,Footnote 15 divides the explanandum event into two distinct events, each of which is explained by appealing to a different kind of cause. Dretske distinguishes between triggering causes on the one hand and structuring causes on the other. A triggering cause is what most of us have in mind when we think about causes: an event that triggers or brings about an effect. Structuring causes, on the other hand, account for what caused the triggering cause to cause what it did “rather than something else.”Footnote 16 Dretske thinks explanations appealing to mental events employ structuring causes while physiological explanations appeal to triggering causes. In this case, to return to our example, George’s rising from the couch qua physiological occurrence is, on Dretske’s view, a distinct event from his intentionally rising from the couch. Because these are distinct events the neurobiological explanation of George’s rising from the couch does not threaten to preempt the explanation of his getting up that appeals to a structuring cause, such as his desire for a beer. Each cause explains a distinct occurrence.
Ausonio MarrasFootnote 17 has also formulated an influential version of the dual explanandum solution. Instead of dividing the event of George’s rising from the couch into two distinct events, Marras proposes there is only one event here, but two explanandum facts. An explanandum fact corresponds to the tokening of a property type by the event in question. Marras assumes that a single event can token more than one property at a time, and accordingly, proposes that this can generate more than one fact that is in need of explanation. On this view, George’s rising from the couch exemplifies two (or more) properties: one is a physiological property and the other is an action-theoretic property. The former is explained by the occurrence of the neurobiological property and the latter by the content property (i.e., George’s desire for a beer). Since each explanation targets a distinct fact, there is no competition between them. Marras also argues that the explanations are distinct from one another on the basis of the claim that the properties highlighted in each explanans differ with respect to their causal relevance. This approach requires that we accept the assumption that explanandum facts, and not events, are the proper objects of causal explanation, but as Marras points out, this seems to be born out by the fact that explanation is an intensional relation.
Both of these accounts provide plausible avenues of escape from the problem of explanatory exclusion, but neither is without its difficulties. Dretske’s version is susceptible to the complaint that either there must be a physical explanation for the event explained by the structuring cause, in which case exclusion pressures arise once again, or else he is committed to an implausible form of dualism.Footnote 18 Although Marras’s approach avoids this difficulty, his account relies on a way of distinguishing the causal relevance of the properties tokened by a cause that has been ably criticized. Marras proposes the following necessary condition for causal relevance: “Where c causes e, where c is F and e is G, c’s being F is causally relevant to e’s being G only if the counterfactual ‘¬Fc Ò ¬Ge’ holds.”Footnote 19 He then demonstrates that the neurobiological and content properties of the cause of an intentional action differ in their causal relevance since they entail distinct counterfactuals. Fuhrmann and Mendonça,Footnote 20 however, identify a number of counterexamples to Marras’s account of causal relevance, which undermines the grounds he provides for thinking of the proffered explanations as distinct. In addition, Marras’s approach depends on a model of co-instantiation that might be problematic.Footnote 21 In the remainder of this section, I explore an alternative response that exploits Kim’s simulation model of action explanation.
We just saw that dual explanandum solutions try to alleviate the tension created by having more than one explanation for a single event by dividing what needs explaining so that each explanation performs a distinct explanatory task. Although Marras and Dretske offer very different versions of this solution, they share the assumption that both explanations are causal explanations and so try to divide the explanandum in two in a way that ensures the explanations do not compete with one another. There is, however, another way to approach this issue that takes its lead from the dual explanandum response, but develops the objection in quite a different way. Instead of dividing the explanandum in some way, what if we were to distinguish the explanations by the way they explain?
We can call this a “pluralist” model of explanation. Kim himself seems to provide a clear way of articulating this pluralism by contrasting the third-person causal-subsumptive model of explanation (associated with Hempel and Davidson) and the first-person simulation model described in the previous section. One could maintain that, when we explain an action by identifying the agent’s primary reason for acting in the attempt to reconstruct the agent’s process of deliberation, we employ the simulation approach and glean one variety of understanding. Alternatively, when a neurobiological property is identified as a cause of an agent’s behaviour, we employ the third-person causal-nomological model and regard the agent’s behaviour as the lawful or statistical product of the relevant antecedent conditions and laws of nature. Kim treats these as “exclusionary stances” in the sense that one cannot simultaneously deliberate and predict, but there seems to be no reason to think that a causal-nomological explanation precludes the possibility of adopting the alternative stance and understanding what was done in terms of the agent’s deliberation. Such an approach would seem to require that each explanation be grounded in a distinct explanatory relation. The third-person causal-nomological explanation seems pretty clearly to be grounded in the causal relation between the identified events, but what shall we say about the explanatory relation at work in the simulation model? Does Kim leave room for an alternative explanatory relation?
The answer to the second question is clearly, yes. Even in his earlier work on the principle of explanatory exclusion, Kim claims that other kinds of relations can serve as the explanatory relation and, what is more, such relations can provide a different criterion of individuation for explanations than the causal relation.
To the explanatory irrealist, [the extensional] … way of individuating explanations would be unmotivated: explanations would be more appropriately individuated in terms of descriptions or propositions and their internal logical, conceptual, and epistemic relationships. Nothing needs to prevent the explanatory realist from accepting this “internal” individuation criterion as well, as defining another useful sense in which we can count explanations. The point is only that explanatory realism motivates an “objective” individuation of explanations, which is both intuitively plausible and well suited for the exclusion principle.Footnote 22
The simulation model involves the criterion of individuation that Kim describes above as “irrealist,” for the reconstruction of a practical deliberation is clearly a matter of logical, conceptual, and epistemic relations involving principles of action and assumptions about rationality. Does this somehow undermine the legitimacy of this relation as an explanatory relation? Does Kim suppose that the explanatory relation is exclusively grounded in an objective metaphysical relation between events? Kim’s earlier work on the principle of explanatory exclusion certainly gives this impression, but his more recent treatment of the simulation model tells a different story. Here Kim seems to be comfortable with the idea that the explanatory relation need not be a metaphysical relation, but can be entirely logical, epistemic, or conceptual as the simulation model would require. If he thought that these relations were too flimsy or “internalist” to provide genuine explanations his simulation model clearly wouldn’t get off the ground. The key, perhaps, is that Kim would claim the simulation model does not seek an “objective” explanation since it focuses squarely on a (broadly speaking) subjective matter: how the agent understands her own action. As the above quotation illustrates, the explanatory realist can accept these as genuine explanations, even if they are not “objective.”
For the sake of convenience we can call third-person explanations “causal explanations” and the kind of explanations involved in the simulation approach “reason explanations” or “rationalizations” since they require the rational reconstruction of an agent’s practical deliberation. The suggested response to the problem of explanatory exclusion, then, says that these are distinct types of explanation that are grounded in distinct explanatory relations between the same events. Since the explanations generate understanding in different ways and are grounded in distinct explanatory relations, there is no competition between them. To return to our earlier example, we can say that the appeal to a neurobiological property offers a causal explanation for George’s rising from the couch, whereas the appeal to the content property provides a rationalization of George’s rising from the couch. Since the first, but not the second, is a causal explanation, there is no tension between the explanations; they do not refer to competing causes. In that case, having a causal (neurobiological) explanation does not threaten to preempt or undermine the legitimacy of the rationalization. This kind of approach can be generalized to all intentional actions, and so we have a clear way to escape the problem of explanatory exclusion—at least in the context of action explanation. This approach appears to escape the difficulties that beset Marras’s and Dretske’s versions of the dual explanandum reply. Since the central claim here is that reason explanations and causal explanations explain in distinct ways, there is no need to distinguish the explananda as distinct events (thus avoiding Dretske’s apparent commitment to a fundamental dualism of events), nor is there a need to differentiate the neurobiological from the content property in terms of its causal relevance (thus avoiding Marras’s troubles with relevance).
§4. Kim’s Potential Responses
Kim has two obvious lines of reply to the version of the pluralist account of explanation proposed in the previous section. Both can be traced back to his earlier work on the exclusion principle, though it would appear Kim no longer stands behind at least one (perhaps even both) of them. The first argues that the notion of cause is essential to rationalizations, and hence, returns us to the causal model and brings the exclusion principle back in full force. The second line of reply, which Kim reiterates in his more recent work,Footnote 23 claims that the simulation model doesn’t explain behaviour, but only explains the agent’s intention.
One way to construe the pluralist account of explanation just described is as a denial of the claim that rationalizations are causal explanations. In an earlier paperFootnote 24, Kim considers this way of avoiding the problem of explanatory exclusion but is in the end dismissive. The main difficulty is that he thinks this undermines our ability to distinguish between having a reason for acting and acting for a reason.
For it is difficult to believe that the rationalizing relation is entirely logical, a relation that can be fully captured in logical terms alone, or quasi-logical and epistemic terms, such as ‘making sense’, ‘being appropriate’, etc. It is difficult to believe that, by using these terms alone, we can capture the difference between being a reason for and being the reason for which, a distinction Davidson has rightly emphasized. Footnote 25
Although this line of thought is also present in another paperFootnote 26, Kim seems to have changed his mind about this problem by the time he proposes the simulation model. In his most recent paper on the subjectFootnote 27, Kim considers a similar issue and it is reasonably clear that his treatment of it can be modified to handle the version mentioned in the above quotation. In “Taking the Agent’s Point of View Seriously in Action Explanation,” Kim describes a problem identified by Thomas NagelFootnote 28 with the proposal that all we need for a satisfactory rationalization is “subjective intelligibility.” The problem is that an agent might have two possible courses of action and reasons for doing both that render them equally reasonable. Nagel says the rationalization “may render the action subjectively intelligible, but it does not explain why this rather than another equally possible and comparably intelligible action was done.”Footnote 29 Kim is not convinced that this is a problem and offers the following response:
It seems to me that there can be a simple answer here: the agent did X rather than X* because she judged, or decided, that X, not X*, was the right thing to do. Why did she judge that way? There could be many kinds of answers here: in comparing R and R* [the respective reasons for each action], perhaps she found that X would have consequences more appealing than the expected consequences of X*, or that X was easier and less costly to do than X*, or that she just felt more like doing X than X*, and so on … What more should we expect to know? … Subjective intelligibility, if it means intelligibility to the agent, is all we should need or want. Nagel goes on to complain that this explanation does not explain why this action, rather than another one was chosen. But that is to complain that the agent’s way of making her action intelligible to herself doesn’t yield an objective, perhaps Hempelian explanation. Well, so be it!Footnote 30
It seems a similar response is available to the Davidsonian problem of distinguishing between a reason for and the reason for which: the reason for which the agent acted is the one she took to be the most relevant, desirable, compelling or appropriate.Footnote 31 Just as Kim resolved the Nagelian problem without making recourse to causal concepts, it seems that he could—and should—resolve the Davidsonian problem in the same way. Mimicking Kim, we can say that the simulation theory doesn’t yield an objective Davidsonian explanation. Well, so be it. That’s not the kind of explanation we are seeking when we reconstruct the agent’s deliberation. Given Kim’s response to Nagel’s problem, then, I fail to see how he could object to the pluralist account of explanation on the above grounds. To do so would be completely at odds with what he says about the Nagelian version of the problem.
Kim’s second line of response to the pluralist view of explanation appears in the same two papers as the first.Footnote 32 There he claims that the denial that rationalizations causally explain fails as a viable response to the problem of explanatory exclusion because rationalizations explain the agent’s intention, not the agent’s action. Kim sums this up nicely in his “autonomy thesis”:
Actions explained by rationalizing explanations are not actions in the sense of overt physical or bodily motions, or in a sense in which such motions are involved. Rather, what is rationalized is an intention or decision.Footnote 33
This undermines the proposed pluralist strategy because that strategy depends on the idea that the same event (George’s rising from the couch) is explained in two distinct ways. If Kim is correct and what is explained by appealing to George’s deliberation is his deciding to get up rather than his getting up, the two explanations are targeting distinct events altogether and there is no exclusion problem to remedy. With Dretske’s approach we at least see the disambiguation of one event into two cotemporaneous events, but here we don’t even have this since the event of George’s deciding or intending to rise from the couch temporally precedes his action.
To reinforce the claim that rationalizations rationalize intentions rather than actions, Kim adds the following:
What actual physical movement issues from our intentions and decisions, whether or not our actions succeed, is beyond the purview of rational intentional psychology. Insofar as commonsense psychology treats bodily actions, it goes beyond “pure” psychology; it is a mixture of folk psychology and folk physics/physiology.Footnote 34
Although Kim confesses that rationalizations can make bodily action “intelligible” in a “derivative” sense, he adds:
But in the same sense we can make sense of bodily actions that are failures—those that fail to implement an intention—in terms of the agent’s calculation and decision and the natural circumstances. How intentions and decisions issue in bodily actions is part of the mind-body problem, not a problem concerning the rationality of agents.Footnote 35
There are actually two lines of argument in the above quoted passages. I’ll take them in reverse order. The second, which appeals to failed actions, is clearly just another incarnation of the Davidsonian problem of distinguishing between a reason for and the reason for which, where a failed action is one for which an agent has a reason but the agent doesn’t act for that reason. As we have already seen, Kim no longer seems to be convinced that this is a problem for the simulation theory, so I will move on to the other line of argument.
The other argument for the idea that rationalizations explain intentions rather than actions seems to be that actions are just the wrong kinds of things to be explained in terms of the items that figure in agents’ deliberations. Actions are, in some sense, impure because they are not wholly constituted by mental content. To the extent that they involve bodily movement, they are not subject to rational explanation since bodily motions themselves lack content that can be rendered rationally intelligible—hence, Kim’s characterization of actions as involving a mixture of folk psychology, folk physics, and folk physiology. The physiological features that (in part) make up an action are simply not the kinds of things that can be rationalized. Intentions, on the other hand, are unambiguously mental, have clearly defined contents, and so are ripe for rationalizing explanations.
I find these considerations unconvincing for two related reasons. First, why should the fact that actions have physical/physiological features count against their candidacy as targets for rationalizing explanations? Unless we are assuming Cartesianism about the mental, it seems to me that the above reasoning ought to apply equally well against Kim’s account of self-understanding and the simulation model. Don’t an agent’s reasons and intentions also have physical/physiological features? Clearly they do, but then by Kim’s own reasoning we should say these aren’t rationally explained either being—as they are—impure hybrids of “pure” psychology and folk physiology. Such a concession would undermine Kim’s entire account of self-understanding, as well as the simulation model, so for the sake of consistency, it seems Kim ought to abandon the above line of argument and agree that actions can be rationalized using the simulation model. It seems more fitting to acknowledge that to the extent that an event has mental properties at all, it is the kind of thing for which there can be a rationalizing explanation. Actions are clearly events that have such properties, for, when we characterize an event as an action, we describe it in terms of its possession of an action-theoretic property, as something that was done intentionally. In the context of action explanation, we are not generally interested in the event’s physiological features. On those occasions when we are interested in such features, we employ a distinct kind of explanatory strategy: we adopt what Kim calls the third-person perspective, but in such cases we are trying to predict the behaviour, not render it rationally intelligible.
The other difficulty with the claim that rationalizations explain intentions rather than actions is that it doesn’t seem to be born out by our explanatory practices. Despite Kim’s advocacy of the autonomy thesis in “Reasons and the First Person,” even he constantly slips into saying that what gets explained on the simulation model is the agent’s action rather than intention. Consider the title of that paper: “Taking the Agent’s Point of View Seriously in Action Explanation.” Clearly, what Kim thinks we are trying to explain when we take the agent’s point of view seriously by employing the simulation model is the agent’s action, not just her intention. After all, the central thesis of Kim’s paper is as follows:
To know, or be able to reconstruct, the practical deliberation that led me to decide to do something (and actually do it) is a fully satisfactory way of understanding why I decided, and acted, as I did.Footnote 36
Had he still believed that the account of self-understanding in the simulation model does not extend beyond the agent’s intention, Kim would not have made the above references to the agent’s action. Indeed, it seems quite plausible to say that the agent’s intention is (the concluding) part of the agent’s deliberation, and that the whole process (including the intention) explains the action. As I said, this seems to be born out by our explanatory practices. When we ask somebody why she did what she did, we are not trying merely to understand her intention (though that is sometimes part of it), we are trying to understand her action. In light of these considerations, it appears that Kim has abandoned the idea that rationalizations only rationalize intentions—and for good reason. In that case, he has yet to identify a compelling reason to resist explanatory pluralism, and so this pluralism can provide relief for those nonreductive physicalists who have felt threatened by Kim’s principle of explanatory exclusion.
§5. Some Broader Considerations
While it is instructive to see how the pluralist account of explanation is implied by Kim’s recent work on first-person explanation and how this pluralism goes some way toward resolving the problem of explanatory exclusion, one might worry that this conclusion is in some sense trivial. After all, Kim’s most recent and influential objection to nonreductive physicalism is no longer cast in terms of explanatory exclusion but the much more troubling problem of causal exclusion as outlined in his supervenience argument.Footnote 37 Hence, one might worry, my observations about how Kim provides the resources for resolving the problem of explanatory exclusion—instructive and interesting as these may be—are not really of any help to the nonreductive physicalist because that position is now threatened by a different guise of the argument.
Such a concern strikes me as both misguided and misinformed. Misguided because it is false to assume that saying something new about an old subject isn’t worthwhile; misinformed because it assumes that what I have said has nothing to do with the newer guise of the problem. The first point is obvious and not worth belabouring. As to the second, while this is not a paper about the supervenience argument, I think it might be instructive to show, in a preliminary way, how explanatory pluralism is relevant to the supervenience argument.
The point of the supervenience argument is to show that the causal powers of supervenient mental properties drain into their physical bases, unless that is, the former can be reduced to the latter. Hence, where we have a mental property instance M putatively causing another mental property instance M*, and M supervenes on physical property instance P (and M* supervenes on P*), it will turn out that the M-M* causal relation is merely apparent; the real causal relation holds between P and P*. A key move in the argument involves introducing the possibility of “downward causation” whereby M causes M* by causing its physical base P*. The reason we are supposed to embrace this possibility is that it allegedly alleviates a tension created by the fact we seem to have two “exclusionary answers” to the question, “Why is M* instantiated on this occasion?” The first answer says that M* is instantiated because it was caused by M; the other says M* is instantiated because its supervenience base P* is instantiated. The reason we should find these two answers to be in tension is that we are supposed to share Jonathan Edwards’ view (Kim calls this “Edwards’ dictum”) that there is a tension between horizontal causation and vertical determination.
This is the point at which it seems to me that a pluralistic account of explanation is clearly relevant. If one is convinced that there can be more than one explanatory relation, it seems to me that one ought to balk at Kim’s insistence that there is a tension here that needs resolving. Why not say that the supervenience relation is a distinct relation from the causal relation and that, as such, it grounds a distinct type of explanation? Surely, we should think there is a tension in need of resolving only if we have two competing causal explanations for the same event, but that is not what we have here.Footnote 38 If this line of argument can be maintained—and I will not here try to demonstrate that it can—this would undercut Kim’s supervenience argument at an early stage.
These brief remarks fall far short of a convincing refutation of the supervenience argument, but it was not my aim to provide one. My goal in this final section is simply to show that considerations about explanatory pluralism—although more obviously relevant to the problem of explanatory exclusion—might also provide valuable ammunition to those nonreductive physicalists who are threatened by the supervenience argument.
I have argued that Kim’s account and endorsement of the simulation model of action explanation opens the door to a pluralistic account of explanation that appears to resolve the problem of explanatory exclusion. The exclusion problem claims that once we have a causal explanation in physical terms for an instance of behaviour, this excludes a rationalizing explanation unless it can be shown that the explanations either fail to be complete or fail to be independent of one another. Drawing on the simulation model, I suggested that rationalizing explanations explain the same events (i.e., actions) as causal explanations, but explain them in a different way. Since rationalizing explanations are not causal explanations, there is no tension between the two, in which case a neurobiological explanation does not threaten to exclude a rationalizing explanation. In his more recent work, Kim seems fully prepared to treat the rationalization of an agent’s action by her reasons as a genuine explanatory relation, and hence, there is little reason to suspect that such explanations are not genuine or are problematically irrealist. This should give some comfort to those nonreductive physicalists who regard the principle of explanatory exclusion as a genuine threat to their position. If I am correct, Kim himself has provided the tools to avert this threat. Of course, it remains to be seen whether the pluralist account of explanation can be successfully defended in its own right. My task has primarily been to show that the view is not only consistent with Kim’s recent work on explanation, but is entailed by it. This is a modest conclusion but is not without interest given the significance of Kim’s work in this area.
