sign in sign up
<<
Home , Inquiry, Irrationality, Rationalism, Rationality, Scientific essentialism

Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 282

282 , COMPLEXITY, COLLECTIVITY 16

Rationality and Rationalist Approaches in the Social

David Henderson

FRAMEWORK associated with (the claim that some were a priori knowable). But, Rationalism may be understood as the they hastened to add that whatever was so philosophical position asserting a certain dis- knowable would be something on the order tinctive epistemic status for certain classes of of definitional truths—claims that there claims—that asserts or supposes that there were necessarily true as a consequence of are a priori knowable truths. On this under- the character of, and relations between, the standing, one is a rationalist if one holds that or concepts employed in those claims. there are certain necessary truths that can All unicorns are mammals—necessarily, be justifiably believed (and that would then since our of a unicorn is of a rather par- count as ) independent of empiri- ticular horsey thing, and our idea/concept of cal for their . This is a some- a horse is the idea of a particular sort of what minimalist understanding of rationalism mammal (or so the rather plausible story (although these days there are those who goes). Nothing would count as a horse, and would count themselves as rationalist merely thus as a unicorn, were it not a mammal. by of embracing this much). But, they insisted, this of itself does not Rationalism has traditionally been under- guarantee that there are any unicorns, or stood as making a stronger claim. In the horses, in . For that, field work, or modern period, empiricist philosophers such trips through the country with one’s eyes as Locke and Hume sought to debunk what open, would be needed. they believed to be the pretensions of ratio- While one might know by reflection the nalist thinkers such as Descartes and relations between our own ideas, said Hume, Leibniz. The empiricists would have granted it is a wholly different question whether there there were some claims that satisfied the is, in , anything in the world satisfying rather minimal characterization just now those ideas or concepts. Matters of fact could Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 283

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 283

only be justifiably believed, could only be This picture was challenged by a line of known, empirically. In opposition, the ratio- thought developed by Kripke (1972), Putnam nalists insisted that there were at least some (1975a), and Burge (1979, 1992). They argued things beyond the creations of our own idea- that what made for, or constituted, the concepts craft that could be known a priori—they in play could include elements of the individ- insisted that one could know substantive ual’s social and physical environment—and truths about the world by reflection, without was thus not wholly internal to the individ- reliance on experiential evidence. Descartes, ual: thus the idea of an externalist seman- for example, thought that we could know tics). On this view, at least some elements of that we were non-material souls, that God the semantics of (at least some important) existed, and that material objects were concepts are not (or need not be) accessible (Euclidean) three-dimensional extended to agents employing those concepts. As a things. (Perhaps there should be a three- result, there could be claims whose truth are strikes rule applied to .) guaranteed by the semantics of the concepts In any case, were we to be fully faithful to featured in them, but which could not be the terms of this venerable debate, we would appreciated by those merely by need to refine our characterization of ratio- their drawing on whatever makes for an indi- nalism: we would need to understand ratio- vidual’s possession of the relevant concepts. nalism as the view that there are certain Perhaps all a priori knowable truths are con- substantive claims that are both necessary and ceptual truths, it might then be said, but, if can justifiably be believed (and thus known) externalism is correct, not all conceptual independent of for their truths are a priori knowable (even by those truth—where a claim is substantive if it is not who count as conceptual adepts at a given ‘merely definitional’, or ‘analytic’ or guaran- time). Even more significantly, when con- teed by the content of its featured concepts. cepts (semantic entities, meanings, and the I mention the traditional and more robust like) come to be conceived as rather more understanding of rationalism only to explic- than ideas in heads, the suggestion itly lay it to the side. There are multiple that a priori truths might turn out to be con- for focusing on the minimal under- ceptual truths does not seem as threatening to standing in this contribution. Several reasons their significance as was once readily sup- have to do with the state of play in contem- posed. The empiricist idea that a priori truths porary . might be limited to conceptual truths no longer First, the idea of a truth that is guaranteed lends itself to the deflationary rephrasing, by the semantics of its elements no longer ‘mere conceptual truths’. seems to be what it used to be—and the Second, perhaps influenced by such con- changes have significantly complicated the siderations, those with avowedly rationalist philosophical landscape. A central develop- inclinations have comes to think largely in ment has come with the advent of what is terms of conceptual truths without termed ‘externalist semantics’. Up until the insignificant (for example, Bealer, 1987; 1970s, almost all thought about the Chalmers, 1998, 2002a, 2002b; Jackson, (ideas, meanings, intensions, con- 1998; Peacocke, 1992). Conceptual truths cepts, semantics) supposed that these things pack some punch, at least if contemporary were settled by what went on inside a given understandings are roughly correct. individual. While such things as meanings or Now that conceptual truths have come to be concepts might be abstract entities, whether a thought of as ‘more muscular’ or substantive given individual entertained or deployed a as a class, the fan of a priori knowledge has given meaning or concept in a given stretch come to face a new challenge: to explain how of thinking was thought to depend on what it can be that those who are relatively profi- occurred within the skin (or perhaps head) of cient with the concept have, by virtue of that that individual.1 conceptual competence, access to powerful Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 284

284 RATIONALITY, COMPLEXITY, COLLECTIVITY

elements of that semantics. This has been a putative a priori truth—that actions have matter of identifying a component of the rational antecedents and , or semantics of the concept that are accessible at something along this line—would both least to those who count as ‘possessors’ of the inform and constrain work in the social concepts involved.2 But, we not now sciences. It would inform a kind of explana- detail the lines of the contemporary debate tory practice—and one apparently ‘on the over the of the a priori. cheap’. After all, one would not have to These philosophical preliminaries do develop well-evidenced generalizations or serve to indicate why it is that contemporary descriptive of ; one would rationalists do not seem much concerned not need empirical theories of human cogni- with what their modern ancestors would have tion that underwrite the explanatory practice thought crucial—why they commonly are in question. Instead, one’s own not much concerned to show that there are a principles of reasoning would supposedly priori knowable truths that are not ‘merely turn the trick of informing and supporting conceptual truths’. They also serve to explain explanations. One’s own normative principles to my readers why my discussion of ‘ratio- or reasoning competence, representing or nalist approaches to the social sciences’ will tracking support relations between contents, focus on positions regarding the subjects of would structure explanations underwritten by social-scientific thought (on positions the very concepts of action and reasons. Such regarding beliefs, thought, actions, and the normative principles would also be constrain- like) that might be thought to be conceptual ing (again a priori constraining)—normative in their foundation. principles would need to figure in explana- In connection with the social sciences, the tion, thereby limiting the kind and character central matter on which a priori truths have of one could employ in the been sought or sensed by those with rationalist human sciences. Such, in outline, is the inclinations has been the role of rationality in central rationalist approach to the social the explanation of action. Put starkly and sciences. In all variations of this generic overly simply, it is said to be a priori that approach, some significant degree of rational- finding one’s subjects to have beliefs and ity in , desire and action is thought to be desires that makes their actions rational a priori conceptually guaranteed. explains their actions, while failing to do so leaves their action unintelligible and unex- plained. This is said to follow from the con- THE RATIONALIST BRIEF cept of an action—which is said to involve the idea of a behavior engendered by rea- Why embrace some variant on the basic line sons. The concept of a is said to of thought sketched above? In surveying the involve the idea of a contentful state that rationalist case, one can begin by reflecting bears a (normatively approvable, i.e., ratio- on the everyday practice of explaining an nal) support relation to some other contentful action by attributing reasons to the agent state (that for which it is a reason). Putting or agents. This is pretty pervasive. One these thoughts together, it is said to be a con- encounters it in discussions of friends, ene- ceptual truth that actions are the sort of thing mies and acquaintances. One encounters it engendered by contentful states (promi- in histories, in newspapers, in meetings and nently, beliefs and desires) that rationally just about anywhere that actions are up for support the decision to undertake that action. explanation. It is reasonably taken to reflect Doubtless, such points would need to be some of our must fundamental understand- sharpened and qualified in certain ways (I ings of what it is to act—and very central will try for more nuance below). At this junc- elements of this understanding might rea- ture, it is useful to state the general thrust of sonably be taken to reflect our concept of an the rationalist position. One then sees that the action,3 and so to be grist for the rationalist Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 285

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 285

mill. What then does the rationalist find for so acting. Indeed, Davidson (1982: 299) when reflecting on the practice of explaining insists that such rationalizing explanation actions by citing reasons in the form of provides ‘the only clear pattern of explana- beliefs, desires, and the like? tion that applies to action’. It is a pattern that When explaining an action by citing some has been elaborated in significant ways of an agent’s beliefs and desires, it seems within , for example. But, what- important that the (typically small) constella- ever we ultimately want to make of these elab- tion of the agent’s contentful states mentioned orations (and that is something to be be such as would count as a reason for so act- considered below), the core rationalist idea ing. If the cited beliefs and desires seem seems to be that this ‘only clear pattern’—this unrelated to the apparent action, then no practice of narrow rationalizing explanation— explanation has yet been provided. To say represents a deep element of our concepts that an agent had such and such a reason for of action and reason. This is to say that some action seems to require that the beliefs action and reason are coordinate concepts— and/or desires here mentioned rationally concepts made for each other—and that it is favor that action. This is not to say that they then a priori necessary that actions be ratio- need make the action rational for the agent all nal at some significant level. things considered (that is, given the full range Suppose that an individual agent under- of the agent’s beliefs and desires). It is to say takes some action that can be given a that, considered just of themselves, they must relatively uncontroversial characterization. rationally count in favor of doing such an Commonly, there will be several constella- action. This is crucial for ratio- tions of the agent’s beliefs and desires that nalists, and is taken to reveal the most funda- could have prompted such an action. That is, mental outlines of the coordinate concepts of within the agent’s full array of beliefs and an action and a reason. An action is the sort desires, there may be several small constella- of thing undertaken for a reason. A reason for tions that are each made up of beliefs and an action is a contentful state, or perhaps desires so related as to present that action small constellation of such states, that pre- (under the uncontroversial understanding) as sents that action, that undertaking, as ratio- a good thing. Understanding and explaining nally a good thing. Because this much only the action will then depend on identifying supposes that the reasons cited make the what—from among the various reasons that action narrowly rational, rather than all things the agent might have had (even did have) for considered rational, the commonsensical so acting—was the reason that the agent did practice of explaining by citing reasons may the action. For example, the agent may have be termed narrow rationalizing explanation. given a gift. Is that to be explained by a Donald Davidson is perhaps the most strong altruistic desire to meet a need of the influential of contemporary commentators receiver, or by a desire to be remembered in on rationalizing explanation, and the ratio- an upcoming decision process (or in a will), nalist brief here recounted is indebted to his or by a nagging guilt over the giver’s own writings. He argues that rationalizing expla- role in a recent decision, or by some combi- nation in terms of reasons turns on the idea of nation of these? Perhaps the agent had all mental states as rational causes. Thus, it is these reasons for giving a gift. taken to be central to our concept of an To explain the action—the giving of a action, and to our most rudimentary explana- gift—one must appreciate which reasons tory practice with respect to actions, that were operative or dominant. To correctly actions are events with (in at least some min- explain the action, one must be able to say imal sense) a rational cause. To explain an that the agent did the action because the action of an agent, the beliefs and desires agent desired such and such, and believed attributed to the agent must ‘make the act this and that, and sometimes this requires intelligible’, must amount to rational reasons saying that the agent did the action for these Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 286

286 RATIONALITY, COMPLEXITY, COLLECTIVITY

reasons rather than because of certain other The upshot of these few quick paragraphs, constellations of beliefs and desires that the focusing on the coordinate concepts of agent might (even did) have. As Donald beliefs and desires as reasons, and of actions Davidson (1980a) argued, there is really no as caused by reasons, has been to present the account to be given of the ‘because’ here, most rudimentary prima facie case for the without invoking a causal relation. It will not rationalist idea that it is a priori that actions do to say that the beliefs and desires simply and reasons must be understood in terms of ‘make intelligible’ the action—for any of processes that are preponderantly rational, at various constellations would do that—and of least narrowly so. Thus, subject to qualifica- only one (more or less sprawling) constella- tions, we can announce a general, if crude, tion of the agent’s actual reasons will it be result: correct to say that the agent did the action because of those reasons. Only prominent Restricted Rationality Principle Concerning members of this constellation will do as Actions (RRP–a) explanations (or will serve as the core of an Actions are behaviors with rational causes. To be explanation) of the action in question. So, the an action is to be a bit of behavior which is (at least rationalist insists, it is a priori that actions are narrowly) rationalized by a constellation of beliefs caused by constellations of beliefs and and desires that cause it. desires that together rationally indicate the desirability of the action. Not only must the The parenthetical qualification is important. agent have reasons for his or her action, but For the above paragraphs have not estab- some constellation of such reasons must have lished or defended any sweeping rationality been central to the processes that causally principle—no principle to the effect that produced that undertaking. actions are rational in virtue of the full of Many a piece of behavior—say, the giving the agent’s beliefs and desires, or that the of a gift—might be motivated by various agents’ beliefs and desires are themselves beliefs and desires. Indeed, the piece of preponderantly rational. (The extent to which behavior might then be understood as one of one might reasonably seek to extend these several actions. A given stretch of behavior general lines of thought to provide for such a might be understood as a selfless gift, an stronger rationalist principle will be the con- obsequious bit of pandering, a response to felt cern in much of the discussion below.) In its guilt, and so on. In important respects, the limited form, it is plausible that RRP repre- character of the action, as action, itself sents something central to, something guar- depends on the reasons that caused it. (This is anteed by the concepts of action, belief, and so, even though we clearly allow some lati- desire. tude for various descriptions of a given action, RRP as depicted above focuses on and some of these do not suppose that the actions—saying that they are behaviors with agent was motivated to undertake an action so rational causes. (Thus, I designated it the described. For example, we allow actions to Restricted Rationality Principle Concerning be described in terms of their unintended con- Actions.) But parallel lines of thought sug- sequences. So the present point would need to gest a parallel limited rationality principle be understood in a fashion consistent with the with application to beliefs and desires. The possibility of such re-descriptions.) Insofar as result would be conceptually grounded ratio- the character of an action depends on the nality principles that apply to any given engendering reasons, insofar as this much is a belief (RRP–b) and to any given desire deep element of the concept of an action, it is (RRP–d)—indicating that these must also be conceptually mandated that actions are at least narrowly rational in their interrela- caused by reasons—by contentful states that tions with generating and sustaining beliefs rationally (perhaps narrowly rationally) pre- and desires. Of course, beliefs and desires are sent them in a favorable light. not undertakings, while actions are. So the Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 287

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 287

supporting these additional But, again, as with RRP–a, it is important to rationality principles will need to be some- keep in that the rationality indicated in what different. Still, it seems central to these lines of thought should be understood beliefs and desires that they must be such as as narrow rationality: a matter of not par- could be rendered intelligible in terms of cer- ticularly subtle rationality involving some tain of the agent’s (narrowly) associated limited range of beliefs and desires that are beliefs and desires. The cumulative effect is salient to the agent in those processes that an understanding of the various items— issue in the beliefs, desires and actions beliefs and desires, as well as actions— explained. There does not yet seem to be a according which these are generated and conceptual demand for a subtle and far- sustained in a web of contentful supports that reaching holistic rationality involving vast are (at least narrowly) rational. ranges of a given individual’s beliefs and The most parallel line of thought here desires and turning on sophisticated support would take the form of reflection on how one relations. explains a person’s having a belief that that A related will begin to push person apparently holds (or a desire that one towards the stronger principles that most an agent apparently possesses). A common rationalists regarding rationality in the explanatory move here is to show that the human and social sciences would also belief was (fairly obviously) rationally indi- embrace. It comes into view when pursuing a cated by at least some set of the agent’s question only touched upon above: how can (salient) other beliefs. Similarly, one com- one determine which of the possible beliefs monly explains an agent’s desire for some and desires that might rationalize an action state or object by appeal to apparent are ‘the agent’s reasons for’ so acting? This antecedent desires salient to the agent, along is best thought of as a complex question, with beliefs that seem rationally to jointly one that can be decomposed into at least two indicate that the state or object would facili- questions, each of which might motivate tate the satisfaction of these antecedent rationalist argument. desires. Here again, one explains by noticing One question is: which of the beliefs and the agent’s reasons for so believing (or desir- desires that might help to rationalize the ing). But, here, one need not suppose that the action are possessed by the agent? Action is belief (or desire) so explained is rationally the joint product of interacting beliefs and chosen, as this would involve an implausible desires, as already noted. In paradigmatic with respect to belief and desire. cases, one can take an action as reflecting a Nevertheless, reasons are readily in question choice, and look for sets of beliefs and when explaining beliefs and desires—and desires that might have (rationally) produced showing the belief (or desire) in question to it. But which of these are possessed by the be reasonable makes the agent and the belief agent? Further, this way of putting the ques- (or desire) ‘intelligible’. Here also, one only tion obscures the magnitude of the interpre- succeeds in explaining the agent’s holding tive task, for the action undertaken itself can the belief (or desire) in question when the only be appreciated or understood against antecedent beliefs (and desires) alluded to an understanding of the beliefs and desires are (or feature in) ‘the reasons’ that the agent engendering it. Thus, as Davidson (1984c, (or desires) as he or she does. Again, 1980c) argued, there seem to be three causal dependencies are at issue. unknowns that must be sorted out on the Generalizing we arrive at a restricted ratio- basis of the behaviors presented to an nality principle with respect to belief and observer/interpreter: beliefs, desires and desire: actions/intentions. Here we confront directly what has been termed charity in interpreta- RRP–b/d: Beliefs and de\sires are states that inter- tion. If we presuppose rationality on the part act in significantly contentfully appropriate ways. of the agent, about any two of Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 288

288 RATIONALITY, COMPLEXITY, COLLECTIVITY

the three (beliefs, desires, actions undertaken) business of interpretation—in order to solve will allow one to determine the third. Given for other factors in the interpretive trinity, any two of the three, and the where solving for the ‘other factors’ involves of rationality, one can ‘solve for the other’. charitable presumptions regarding rational- For example, desires can be determined from ity. Rationalists, of course, insist that that actions undertaken and from beliefs. There charitable presumption of rationality is con- seems to be no alternative but to suppose ceptually mandated—a conceptually man- some measure of rationality, or so it is dated principle significantly stronger than argued. But, says Davidson, one must pre- RRP. sume more. Thus, to tentatively summarize the ratio- The rationality envisioned here as presup- nalist response to the first question: to figure posed or imposed as a part of the interpretive out what beliefs and desires a subject holds, endeavor seems to take one beyond the nar- we are instructed to make two charitable pre- row rationality at issue in the above discus- sumptions in interpretation. We are to pre- sion. Significant light is shed on an agent’s sume some significant degree of correctness ‘standing’ beliefs and desires by looking at a of belief on the part of the subject, at least wider set of actions undertaken by the agent with respect to some significant range of in question. Of course, there might be matters. Further, we are to presume some various interpretations that one might plausi- significant measure of rationality in the bly put on these—but the idea is that one structure and interaction of beliefs and might constrain workable alternative inter- desires—so that the content of a belief or pretations by looking for plausible coherence desire will be understood in terms of its pat- over the standing beliefs and desires attrib- terned relation to other beliefs and desires, uted against the ‘data points’ provided by given their contents. To cut down on alterna- large sets of behaviors. But, as one imposes tive sets of attributable beliefs and desires, a rationality assumption over wider sets of we are to extend the range of observed actions—and their precipitating reasons— choices/undertakings and speech to be one imposes a more holistic and less narrow accounted for.4 This will allow one to rule sort of rationality in interpretation. out some alternatives explanations for a Focus on this issue: how can one get some given action, as the eliminated explanation principled grasp on any one pair—for exam- would be inconsistent with the agent’s pat- ple, beliefs and actions—within the triad of tern of actions and . The crucial unknowns? Again, as Davidson would insist, point is that in casting our interpretive and there is in play a charitable presumption explanatory net more broadly, one must having to do specifically with beliefs—and apparently presume stronger and more holis- beliefs provide a common entering wedge for tic forms of rationality in beliefs, desires and interpretation. Davidson (1980c, 1984b) decisions. holds that we can and must charitably pro- Then there is the second question: which of vide some determinacy to the belief element the constellations of rationalizing reasons to in the triad by assuming that most beliefs be found in the agent represents ‘the reason’ about any given subject are true. (For now, that the agent so acted? Here, the rationalist I am looking to Davidson’s understanding of idea is that, in keeping with what has been charity in interpretation for inspiration in said to this point, the determination of an developing a rationalist brief. Later, I will agent’s reasons for acting in a certain fashion compare this understanding with a more cau- (that is, for determining what reasons were tious or limited charitable approach is sug- the reasons, for determining which were gested in Quine’s writings.) Davidson’s causally salient) is a matter to be settled by a charitable policy with respect to the attribu- kind of to the best explanation, con- tion of beliefs serves as a way of ‘holding strained by the range of that agent’s beliefs belief constant enough’ to get on with the and desires (charitably determined) and by Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 289

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 289

the kind of holistic rationality necessarily perhaps that they are states with the relevant supposed in their determination. content by virtue of this pattern of dependencies. The upshot of all this might now be for- Here, it seems, rationality is partially constitutive mulated as an unrestricted (or much less of beliefs and desires. Again, we have the idea of a cause that is also a reason (RRP–b/d). restricted) Rationality Principle in two 3. To exhibit or reveal the narrow rationality of a clauses. choice or action is to explain it. To exhibit the sig- nificant rationality of an agent’s beliefs or desires Rationality Principle (RP) is to explain them. In both cases, one is exhibit- RP–a: Actions or undertakings are behaviors with ing the choice or to be caused by its rational causes—beliefs and desires cause them, rational antecedents in ‘intelligible’ or ‘reason- and do so by virtue of certain contents that make able’ ways. them reasonable, significantly holistically rational, 4. Overall, to qualify as a belief, desire or action, a and thus intelligible. result requires more than the narrow rationality of choices or actions undertaken, or the narrow RP–b/d: Beliefs and desires (while not choices or vol- rationality of belief and action—as it requires untary in the same limited sense in which actions/ that the choice and its belief and desire parame- undertakings might be) are caused products of each ters themselves exhibit rationality of some signif- other, in a fashion that reflects contents and makes for their being reasonable, significantly holistically icant holistic or extended sort (although certainly rational, and thus intelligible. not perfect holistic rationality) (RP).

Much thought in economics and related dis- Alexander Rosenberg (1985, 1988) provides ciplines is aptly understood as continuous a particularly clear and striking way of with the common thought about beliefs, advancing the rationalist points made here.5 desires and actions that we have been con- He takes note of the ways in which finding sidering. In various ways, economists have rationality is taken to be explanatory—by sought to develop precise mathematical ways making intelligible a choice or undertaking of thinking about choice, ways that attempt by revealing the reasons that motivate it. He to measure strength of desire and degree of is led to give expression to RRP when he for- belief in scales with understood properties. mulates an ‘oversimplified general statement While such thought may be more articulate [that] seems to lie behind ordinary explana- and more careful than everyday talk of tions of human action’ (1988: 25): beliefs, desires and actions, it must be recog- nized that something like RRP and RP plays [L] Given any person x, if x wants d and x believes a parallel role there. The central issue for our that a is a means to attain d, under the circum- stances, then x does a. purposes is whether, or to what extent, some- thing like RP has an a priori status there. Taken as expressing a narrow form of ratio- This will need to serve as the main brief nality, the exhibition of which is central to for rationalist approaches to the social intentional explanation, [L] serves to express sciences—particularly with regard to the the sort of thin putatively a priori claim—the putative a priori role of rationality in under- RRP—envisioned by rationalists. standing action. It reflects the considerable Then, reflecting on how beliefs and reasons that one might have for advancing desires are thought to conspire holistically to several claims as a priori in character: produce a choice, Rosenberg is led to suggest a more full-bodied principle representing 1. That actions are behaviors with rational how rationality is putatively involved in the causes—at least the agent must have certain belief states and desire states that make it nar- explanation of action and in the interpreta- rowly rational for the agent to undertake that tion of agents’ beliefs and desires. The sug- action (RRP–a). gestion is that something very like normative 2. That beliefs and desires are states which interact decision represents a (again fairly in the ways that are significantly rational—and commonsensical) refinement on [L]—call it Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 290

290 RATIONALITY, COMPLEXITY, COLLECTIVITY

[Lr]. It is worth noting that these refinements choice or undertaking and (b) added up to a are developed and advanced largely ‘from reason for so acting (jointly portraying the the armchair’. That is, the wrinkles associ- action as good to do). It does not claim that ated with can certainly seem these causal antecedents, these contextually ‘natural’ when systematically reflecting on salient occurrent beliefs and desires, are sorts of cases in which the antecedent of [L] themselves ultimately rational for the agent would be satisfied and yet the agent not to hold—ones that makes rational sense in undertake the indicated action, one is likely light of the agent’s wider set of beliefs and to think: ‘Now, x may want d and x believe desires, or ones that result from rational that a is a means to attain d, under the cir- or deliberation. Second, it does not cumstances. But what if x wants g more than hold that the belief is presented in an overall d and believes that getting g is incompatible favorable light by the total set of the agent’s with doing a.’ Just as [L] expresses a weak beliefs and desires—or even by a very exten- and thin rationality principle thought to a sive set. It does not suppose that the action is priori hold action, so decision theory would ‘all things considered’ rational, given the constitute a highly substantive refinement— agent’s full range of beliefs and desires. [Lr]—and is taken by Rosenberg to express a This said, it becomes plausible that RRP–a much more constraining and substantive, (and Rosenberg’s [L]) might indeed be a putatively a priori claim: in effect, RP. central element of the concept of an action. According to Rosenberg (1988: 30–6), cases of actions are intentional you should come to a striking realization: behaviors undertaken for certain reasons— little testing and refinement of [Lr] is possi- such reasons jointly amount to a representa- ble. The reason is again rooted in the princi- tion of that undertaking as a way of attaining ple of charity in interpretation: [Lr] is certain of the agent’s ends. This does not supposed in arriving at the interpretations mean that every action is intentional—that that would be the necessary preliminaries to (so described) it was intended by the agent. determining whether agents conform to [Lr]. There are familiar ways of describing an Rosenberg concludes that a rationality prin- action that do not turn on the agent’s under- ciple along the lines of [Lr] functions as standings or representation of that undertak- something like a ‘definition’, rather than ing. For example, we sometimes describe an providing an empirically testable or refine- action in terms of consequences that were not able description of cognitive tendencies or intended. Descriptions in terms of institu- as a nomic generalization (1988: 33). tional consequences can be a case in point. It is said that in 2000 a significant number of voters in Florida voted for the Republican CRITICAL , STAGE ONE: candidate for president unintentionally. Yet REGARDING RRP such agents possessed reasons for their undertaking as they understood it. As they Let us begin by looking at the least demand- understood their action or undertaking, they ing element of the position: the Restricted had their reasons, and these made it out to be Rationality Principles. We can focus on desirable in the sense envisioned in RRP–a. RRP–a. We should keep in mind two respects Further, RRP–a should be understood to in which RRP–a advances a very limited apply to less paradigmatic cases of actions or claim. The first has to do with the etiology of undertakings. In impulsive acts which were undertakings or actions; it holds only that, not conditioned by significant deliberation, among the agent’s vast set of beliefs and one might ‘just have felt that it would be nice desires, there are or were some contextually to’ do such-and-such (pinch the child, crack a salient beliefs and desires that both (a) fea- joke, run to the top of the hill …). Yet, there tured in the near causal antecedents of the is a sense in which the agent inarticulately Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 291

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 291

‘chose’ to engage in an action which ‘seemed be termed the reason—a causal vector of good at the time’. Here, ‘the constellation of meaningful states in which some stand out as beliefs and desires that cause’ the action may dominant or controlling of the subsequent be rather thin, but they present it in a good course of action. (if feeble) light, and this conforms to RRP–a. The central idea in the more robustly ratio- I do find it plausible that this much is con- nalist approach to the human sciences, and ceptually guaranteed by the very concept of to the place of rationality in those sciences, an action. Were we to give up on the notion turns on one simple idea that is said to be that this holds for a wide range of those guaranteed by the coordinate concepts of events that we have been thinking about as belief, desire and action/undertaking: that undertakings, we would thereby have com- beliefs and desires as considerations interact pelling reason to give up on the idea that there according to holistic rational principles, and are any undertakings and any actions at all. thereby compose themselves, all things con- RRP does not provide much guidance in sidered, into rational choices—choices that settling the real substantive questions that are, from the point of view of such consider- concern social , or even those that ations, a rational resolution. As reflected concern folk in everyday contexts. It does in Davidson’s and Rosenberg’s writings, not, for example, give much direction for the projected a priori rationality in action determining what are ‘the agent’s reasons’ is of the sort represented by normative for a given stretch of behavior that might decision-theory (and , and epistemology). well be an action. For many episodes that It is thought to be conceptually necessary one plausibly treats as some undertaking or that, in the preponderance of cases, actions action, there may be various constellations of undertaken are the rational product of the beliefs and desires that the agent might hold, strengths of the agent’s various desires and of and that would put the undertaking (under the agent’s beliefs concerning the propensity some interpretations) in a favorable light, of various course of action to produce or and it is a significant question which of these frustrate those desires. The rational course of sets the agent had, which were the agent’s action for an agent is that with the highest reason for so acting, and what then is the expected value among those courses of action intentional character of this undertaking. If open to the agent—where the expected value there is to be strong a priori guidance or of an action is understood as the sum of the constraint on the explanation of actions, possible (positive and negative) outcomes of there would need to be markedly stronger a that course of action (as conceived against priori principles. the agent’s background beliefs), each weighted by the agent’s understanding of the of that outcome given that course EVALUATION, STAGE TWO: of action. REGARDING RP Such is the full-blooded conception of rationality in action that the proponent of Empirical Resources and the RP commonly envisions. This understanding itself supposes an understanding of a corre- Revisability of Rationality sponding holistic rationality in belief and Expectations desire. In keeping with the principle of char- We need to understand how the range of an ity, agents are understood as possessing a individual’s beliefs and desires interact as rich set of standing beliefs and desires. These considerations and compose themselves so may evolve over time, under prompting by as to yield a choice. We need to understand and reflection. But, these stand- how various reasons or considerations ing states are understood to be ‘reasonably’ resolve themselves into something that might constant, and changes in them are thought to Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 292

292 RATIONALITY, COMPLEXITY, COLLECTIVITY

be of a largely rational character. Their inter- subject to empirical test, i.e., [Lr] can be actions or interrelations are said to be such empirically shown inadequate and refined. as to evince a preponderance of rationality. The best reason for thinking so is that there is Rationalists seem less articulate on the pre- reason to think that [Lr] has been tested cise character of the rationality that is sup- and found inadequate—prompting empirical posed to be guaranteed here—and they refinements. An apparently instructive exam- resort to general and hedged formulations. ple can be found in the well-known work of Davidson writes of a ‘large degree of consis- Tversky and Kahneman (1974). Tversky tency’ (1980b: 221), and of significant (1975) contrives experimental situations in conformity with ‘stipulated structures’ of a which people’s responses give us empirical normative character (1980c: 6–7), of ‘impos- reasons for revising our understanding of ing our logic’ in interpretation. ‘It is uncer- human cognitive tendencies—evidence indi- tain to what extent these principles can be cating that [Lr] must be abandoned or, what made definite—it is a problem of rationaliz- amounts to the same thing, significantly ing and codifying our epistemology,’ says revised. Consider a set of situations and Davidson (1980c: 7). It should now be clear results that Tversky discusses. The situations that RP amounts to a rather significant ratio- are of a common sort found in studies of nality claim regarding both cognition and decision-making under : choices action—a claim taken to hold a priori of all between gambles. (Using the standard creatures with beliefs and desires, creatures notion, (X,P,Y) will represent a gamble where who undertake actions. Compared to the first one will receive X with a probability of P, or small rationalist step (RRP), this second step Y with a probability of 1 – P.) Tversky pre- (RP) seems quite a stretch! In keeping with sented subjects with a choice between the conceptual status claimed for it, it is said gambles A and B: to so constrain both the attribution and expla- nation of actions and cognitive states that it is A= ($1000, 1/2, 0), B= ($400) neither at risk of significant empirical chal- lenge nor susceptible to significant empirical Presented with this choice, almost all sub- refinement. Call this the strong rationalist jects prefer the ‘sure thing’, B. They do this position regarding rationality in the human despite the fact that A has a greater actuarial sciences. value: $500. Confronted with such sweeping claims Such results are not themselves news derived from abstract philosophical reflec- within standard decision theory, and present tion, one does well to approach them with no immediate threat to [Lr]. After all, it is caution. If strong rationalism is correct here, common to distinguish between the amount then something along the lines of full norma- of goods or money to be had and its ‘’. tive decision theory descriptively applied— The latter is conceived as a subjective, non- [Lr]—must be correct. If strong rationalism is linear, function of the former. The common correct, then [Lr] should not be subject to postulation of decreasing marginal utility— empirical test or refinement. [Lr] could not a concave positive utility curve—is clearly be subjected to empirical refinement or test enough to accommodate the results because [Lr] would play a conceptually obtained in connection with choice situa- grounded constraining role in the attribution tions of just this first sort. One need only of beliefs and desires; attributions involving claim that, commonly, u($400)>1/2u($1000). significant violations of [Lr] would count as This response is just what the strong ratio- problematic (indeed as conceptually incoher- nalist would anticipate: [Lr] is not impugned ent) interpretations. by the above results because we interpret However, there is reason to believe that our subjects on the basis of its charitable [Lr]—in effect, decision theory deployed as a insistence on standard normative decision descriptive account of human cognition6—is theory. Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 293

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 293

However, this is an overly simple descrip- mathematical relations between 1000 and tion of our interpretive practice. At some 400—or that whatever could be purchased point, and Tversky’s work takes us to such a with $400 can typically be purchased in a point, the [Lr]-informed identification of matched pair with $200 remaining from values held by subjects comes to clash with $1000? The reason that such differences in other constraints—and [Lr] can give way. understanding are not plausible is that in This begins to be in evidence in connection addition to some expectations for certain with a second choice situation, one produced forms of rationality, we also have expecta- by multiplying the of gains by tions regarding roughly when people learn 1/5. That is, subjects are presented the choice rudimentary matters of importance within between C and D: their society. We expect such elementary math and monetary units to be learned much C= ($1000, 1/10, 0), D= ($400, 1/5, 0) earlier than college. Such relatively mun- dane, but nevertheless empirical, expecta- If the explanation of the choices found in tions effectively block positing significantly the first situation were really the concave different understandings of the relevant shape of the subjects’ curves, then aspects of the situations Tversky presents to we could expect a preference of D over C. his subjects. However, that is not what is observed. Within It is more plausible that Tversky’s subjects the confines of standard decision theory, the understood the probabilities stipulated in overall pattern of choices is ‘incompatible ways differing from Tversky’s (and ours). with any utility function’ (Tversky, 1975: And it certainly is true that they may not 166). Tversky’s results suggest that there is a have developed any sophisticated under- ‘positive effect ... [in which] the standing of probability. But, Tversky’s utility of a positive outcome appears greater results do not require sophisticated under- when it is certain than when it is embedded standings. It seems quite likely that his sub- in a gamble’ (1975: 166). He also provides jects could have applied talk of probabilities evidence for a negative certainty effect. Such to matters such as coin tosses, urns with col- interactions of utility and probability violates ored balls, and whether their car would start aspects of standard normative decision next time tried. That would be enough to make theory, where it is supposed that there are Tversky’s results telling. (If it was lacking in utility functions (unique up to a certain trans- most people’s thoughts, normative decision formation) associated with particular goods theory is likely in trouble anyway.) Again, we and that such utility functions interact simply find some relatively mundane and empirical with subjective probabilities according to the expectations constraining interpretation— rule: u(x)p(x). and these could be given further empirical One tempting response would be to development. insist that Tversky’s subjects just did not Thus, in addition to some [L]-like expec- understand the situations in the way he sup- tations serving as constraints on interpreta- poses. This would be to invoke the strong tion, we find various more or less mundane, rationalist position regarding charity and [Lr]. more or less empirically developed, expecta- However, and this is crucial, one can raise and tions also constraining interpretation—with address this issue in a principled fashion— the result that suggested refinements of [L] one that itself seems empirically informed. can be put under significant empirical test in Consider, just what was it about the situa- ways reflected in Tversky and Kahneman’s tions that Tversky’s subjects plausibly under- work (to name just one prominent example). stood differently? They were American Call these empirical constraints—empirical college students. Is it plausible that they expectations (or EE). These EE are a diverse did not understand talk of ‘dollars’? Of that lot. Some are fairly general in character—for they did not understand the rudimentary example, they may have to do with the power Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 294

294 RATIONALITY, COMPLEXITY, COLLECTIVITY

of human cognitive abilities, whether it is accommodates Tversky and Kahneman’s reasonable to expect that someone would , but those observations are ‘put certain things together’ and appreciate highly plausible; it seems, in fact, that they certain implications, with whether it is likely are most plausible, given the sorts of empiri- that certain learning or would be cal constraints in question. recalled from memory, with certain com- Those of a strong rationalist bent may monality in human , ‘needs’ and have conceived of a rejoinder. They will note the like. They may have to do with various that the various above-mentioned concrete domains of human cultural phenomena: empirical constraints themselves turn on , group identity, political phenomena, antecedent interpretations of human the flow of information within various generally, and of those in more narrow popu- groups, economic phenomena and the like. lations such as those from which Tversky They may have to do with particular cultures and Kahneman’s subjects are drawn. After or groups, as in what things are learned when all, how do we know that most folk of col- within a certain culture. The point is that lege age were long ago exposed to informa- such diverse EE provide a significant con- tion regarding certain topics? How do we straint on our understanding of people, and know that humans are capable of learning can make it empirically plausible that one what little math is needed to recognize the has encountered a case where some proposed points of significance? These empirical con- development on [L], such as [Lr], is violated. straints seem to be ploddingly obvious With systematic enough violation, one can generalizations arrived at on the basis with have empirical basis for abandoning some everyday experience with those very popula- proposed development on [L] in favor of tions, or presumably similar populations. As others.7 such, they depend on antecedent interpreta- The essential issue in evaluating strong tion. The rationalist would insist that such rationalism is whether [Lr] plays such a deci- interpretation must have been constrained sive and dominant role in informing what and informed by something like [Lr] all beliefs and desires are attributable to agents along. It then seems that, in retaining these that [Lr] is itself rendered immune to empiri- interpretations—and in making the revisions cal pressure and revision. We have just con- that constitute PT, or something on this sidered a kind of empirical inquiry in which order—one must be, or should be, seeking to it seems that significant basis is provided for diverge from [Lr] in the most minimal fash- revising [L] in ways that amount to abandon- ion. This is to say that [Lr] cannot be empiri- ing [Lr]. The suggestion has been that there cally revised much, and that [Lr] itself serves are multiple empirical constraints on inter- as a irrevisable constraint from which devia- pretation—EE—that can provide leverage tions under interpretation must be mini- for abandoning or deeply revising [Lr]. mized. The strong rationalist point might be Tversky and Kahneman (1974) then advance put in terms of the performance/competence an alternative to [Lr]—an alternative descrip- distinction: one may need to attribute tive account of human choice behavior— moments of , it is conceded, but which they term prospect theory. It counts as these must always be isolated enough to an empirical refinement for several reasons. count as mere performance errors against a First, the motivation for abandoning [Lr] in background of rational competence. favor of some alternative is empirical. There are reasons for doubting that the Second, the particular alternative is judged rationalist has things quite right here. The promising and worthy of further empirical essential issue has now to do with the char- investigation because it accommodates acter of the ultimate constraints on interpre- the observations obtained in Tversky and tation. Is it really the case, a priori, that any Kahneman’s work. Third, PT not only empirical refinement of [Lr] would need to Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 295

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 295

rely on interpretations that largely confirm The rationalist understanding of the principle [Lr]? The rationalist rejoinder requires that of charity contrasts with Quine’s understanding some rather powerful normative model of of that principle, according to which the char- rationality (something like standard decision itable constraint is derived and plastic. Quine theory together with some parallel account of (who is writing somewhat narrowly of transla- epistemic rationality) serves as an invariant tion) says that we must translate others so as constraint on interpretation—that there is to preserve ‘the obvious’—where what is some such constraint on interpretation, obvious is a matter of empirical . which is invariant, does not evolve, being Before discussing attributions of rationality set a priori. Again, in philosophy, this view and irrationality, consider the implications of is prominently associated with Donald Quine’s admonition for attributions of true Davidson’s writings. and false beliefs. Some truths are (in context) The Principle of Charity vs. The Principle relatively obvious—for example, given a of Explicability (Sub-heading level 2) context of good illumination where one’s The principle of charity in interpretation is subject is at arms length, it should be obvious roughly that one must so interpret as to find that one is faced with a rabbit (and not a griz- those interpreted to be preponderantly ratio- zly bear). (Of course, there might be less fre- nal and believers of mostly truths. (This for- quent contexts involving good light and mulation reflects Davidson’s influential proximity where it would be at least equally development of the principle.) Let us focus obvious that one is faced with a grizzly bear, on the idea that we must find rationality not a rabbit.) So, if one’s scheme for translat- under interpretation. Here it is crucial to ing some people has them regularly misiden- distinguish between two understandings of tifying instances of these two kinds— this supposed constraint on adequate or insisting (obviously mistakenly) that they acceptable interpretation. One sees the first have killed a grizzly bear, and warning of the as absolutely fundamental, and the other as rabbit protecting some winter kill—one derivative and plastic in certain respects. The would have reason to rethink one’s transla- strong rationalist idea is not merely that we tions. Such matters are perceptually obvious must find our subjects to be reasoning in cer- (and this is in large degree an empirical tain ways, and that many of those ways hap- matter): people tend to get right such every- pen to be rational ways to think, so that we day matters about middle-sized physical and need to find such rationality under interpreta- biological objects in plain sight, and with tion. (That much is congenial to one who respect to which they have a significant prac- sees the constraints on interpretation as a tical interest in developing a competence. On matter of deploying expectations for human the other hand, what is not perceptually obvi- reasoning that are commonsensical but ulti- ous need not be treated as true under transla- mately empirical in character.) Rather, the tion. The ill-glimpsed form in the brush strong rationalist idea is that that the need to might be misidentified, and this does not attribute rationality is fundamental, that ratio- indicate a problem with translation. nal ways of thinking (and acting) must— Similarly, translation that has us attributing because they are rational ways of thinking glaring errors in reasoning of sorts that ‘one and acting—be supposed in interpretation. It would find obvious’ should not be accepted, is the idea that it is by virtue of being unless there are mitigating circumstances. rational that certain ways of thinking must (Factors that might count include the pres- be found under acceptable or adequate inter- ence of drugs, alcohol, sleep deprivation, pretation. Not all understandings of the very strong personal interest in a conclusion principle of charity turn out to suppose this, other than that rationally indicated, and but strong rationalist understandings (such as some kinds of defective training.) Empirical Davidson’s) do. results regarding human foibles seem highly Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 296

296 RATIONALITY, COMPLEXITY, COLLECTIVITY

significant when determining what percep- in everyday and common experience (of tual matters should be relatively obvious in common perceptual capabilities and limits, context. and of common capabilities and Empirical results also seem significant for foibles). It is a matter subject to systematic determining what forms of reasoning are cog- study (as in the empirical work on human nitively obvious and which are not. People inferential strategies and errors). On Quine’s seem rather better at working with conjunc- understanding, the charitable constraint that tion and negation than with conditionals. we preserve the obvious provides substantive They seem better at working with condition- constraints on interpretation only when con- als when these involve concrete matters with joined with such empirical information—so which they have significant experience. They that the substantive constraint here does not can believe the damnedest things when gods constitute an a priori constraint on interpreta- or governments are involved. They may tion to the effect that we must find rational- believe contradictory things when that contra- ity. Rather, if anything is a priori demanded diction is ‘well hidden’—so that it might here, it is that we should seek to find others require subtle proofs or particularly agile reasoning in ways characteristic of the class to appreciate. To insist that we ‘pre- of cognitive systems to which they belong serve the obvious’ in our interpretation is to (or characteristic of such critters in relevantly insist that there are certain ways of reasoning similar circumstances). The principle of that should be found in those we interpret:— charity is thus understood as an empirically the ones that characterize reasoning in the informed constraint on interpretation, one relevant set of critters (say humans). Some of that results from the application of our evolv- these ways happen to be rational. This is a ing empirical understanding of the relevant fact about human beings about which we are cognitive systems. ‘The translator will getting a progressively better grasp as we depend early and late on psychological con- investigate human inferential tendencies. jectures as to what the native is likely to Since there is arguably significant human believe’ (Quine, 1987: 7). Since human rationality, the advice to interpret so as to beings are given to some significant forms of ‘save the obvious’ would have us interpret so rationality (as well as to some significant as to find significant rationality (the obvious forms of irrationality), the derivative demand rationality). Still, to put it mildly, humans that our interpretation be informed and con- turn out to be subject to non-negligible irra- ditioned by these expectations for some tionality. The rational principles that would forms of rationality (and some irrationality) serve as a corrective to such tendencies are, can be termed a principle of charity. We are emphatically, not generally obvious to folk in to seek to find rationality of the common context. So, if we must preserve the obvious, sorts—and failure to do so results in an then findings of such irrationality (cases of account which is likely mistaken (for when forms of irrationality to which folk are given, the errors attributed to folk are highly cases where the contrasting form of rational- unlikely, mistaken interpretation is relatively ity is not obvious), are no strike against an likely). interpretation. To be fair to the rationalists, one must What is crucial on Quine’s understanding notice that they would typically acknowl- of the principle of charity is that while there edge a role for empirical information about are real substantive constraints on interpreta- human cognitive capacities and incapacities. tion, the substantive constraints here do not Such information is acknowledged to be constitute a kind of a priori constraint. What important in determining what interpretation is obvious perceptually or cognitively is an is the best interpretation of an agent or empirical matter: it is a matter of psycholog- people. Thus, Davidson (1984c, 1984d) ical tendencies. It is a matter regarding which insists that all attributions of error and irra- we have significant empirical access rooted tionality count against an interpretation in Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 297

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 297

some measure (this is the a priori part), but articulates a powerful a priori constraint on that some count more strongly than others interpretation, an invariant a priori to (this is at least partly an empirical matter). which all interpretation must approximate. So, in determining what is the best interpre- What makes for the best approximation may tation of an agent or community of agents, be empirically informed, as expectations we seek to attribute no irrationality; but, as for human capabilities and limitations may some attribution of irrationality will be inform what errors make for significant unavoidable, we should settle for attributions divergence and which do not, but the model of irrationality that violate our learned expec- of reasoning to which the interpretations tations the least—i.e., the empirical informa- must ultimately approximate is invariant. tion contributes to the negative weighting of On the Quinean understanding, that model of attributions of error. (Of course, one would reasoning to which our interpretations must also want to allow that expectations will con- find our subjects approximating is neither a tinue to evolve over time.) priori nor invariant. It is rather our evolving This seems reasonable, but it also seems to empirical understanding of human reasoning. make for a more attenuated form of rational- As our understanding of human reasoning ism. It originally seemed as if the strong tendencies evolves under work in cognitive rationalist could insist that there are certain psychology (for example), the resulting levels and forms of rationality that would expectations for both rationality and irra- need to be found under interpretation. If this tionality form a composite model of human much could be taken to be a priori, and if the cognition, and this model is that to which levels and forms of rationality corresponded interpretation should conform for now. An at least to [Lr], then [Lr] becomes unassail- element of this model—say, the expectation able (at least at the level of competence)— for certain forms of valid deductive reason- and a RP is vindicated. But once one allows ing, or the expectation for certain (fallacious) that empirical expectations can modulate the overuse of some judgment —is a putatively a priori demand for finding ratio- piece of the model because it is a piece of our nality under interpretation, it becomes less present best understanding of humans, not clear what is a priori guaranteed. As noted because it is given a priori as rational.8 earlier, the rationalist tends to adopt some- To illustrate the between the what hedged formulations at this point. Thus, strong rationalist approach and the empiricist Davidson writes of it being a priori that approach, we can contrive a cartoon history beliefs, desires and actions are preponder- of our interpretive practice. So suppose some antly rational. Here it seems that RP/[Lr] point in that practice that should surely be serves to characterize an a priori ideal to congenial to the rationalist. Suppose that which all adequate interpretation must there were a time in which interpreters had approximate, from which no acceptable no empirical expectations regarding human interpretation can diverge too much. It is then reasoning—only a normative model that acknowledged that what counts as ‘too includes things like normative decision much’ is at least partially an empirical matter. theory, statistical reasoning, basic logic and (If it were wholly an empirical matter, then the like. (I that there ever was such a again it seems that the a priori element here point, but let us not pause over this point.) becomes vacuous.) So, as long as the ‘too According to the strong rationalist, this is not much’ is not much, interpretations will need too impoverished a position from which to to conforms largely to [Lr], and background begin, for they insist that interpreters yet have interpretations will not provide the basis for the a priori ideal to which all interpretations any but minor revisions of [Lr]. must approximate anyway. Admittedly, inter- The contrast boils down to this: On the preters would have no nuanced way of Davidsonian understanding (the strong ratio- weighting divergence from the ideal—no nalist understanding) the principle of charity empirical weighting of errors. But, we may Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 298

298 RATIONALITY, COMPLEXITY, COLLECTIVITY

suppose that they might then count apparent confidence of researchers that their interpre- divergences equally against an interpretation— tations are reasonable and that the diver- and decide on an interpretive scheme for a gences from some rational ideal that they people by choosing that scheme that mini- seem to find are indeed real and systematic. mizes divergence. But at this point there is a As a result, investigators will come to have wrinkle to consider. On the one hand, inter- empirical theories or expectations having to preters could go on ever modifying their do with human rationality and irrationality. interpretations so as to ‘explain away errors’, To this point, the cartoon history has been adding ever more epicycles to their interpre- developed in a way that is highly favorable to tive schemes, or they could have some sense the strong rationalist. For purposes of illus- of ‘reasonable’ or ‘plausible’ complexity. At tration I have supposed that the sort of nor- some time, they may sense that an error in mative model that the strong rationalist inference, an unacknowledged inconsistency envisioned as anchoring interpretation does or some other piece of irrationality is ‘more indeed constrain interpretation, at least at a likely’ then yet another sense of the relevant mythical beginning in which no empirical terms in the subjects’ lexicon, yet another expectation regarding human reasoning is epicycle. Perhaps, drawing on with brought to the table. Now we can let the dis- their own reasoning, they may sense that the agreement between rationalists and empiri- avoidance of attributions of error is making cists emerge. for an unrealistically baroque interpretive Suppose that we now undertake to under- scheme. It would be natural to think of such stand some new agent or people. According to judgments as empirically informed, as draw- the rationalist, the model that is presupposed— ing on courses of experience. But this is no from which attributed deviations are to be problem for the strong rationalist who is counted against the interpretations that we will ready to acknowledge that empirical infor- entertain—continues to be the same, invariant, mation can help ascertain when an interpre- normative model (logic, statistical methods, tation closely enough approximates to the a the rest of normative epistemology, and deci- priori ideal. All that has been supposed here sion theory). What has changed over time is is that at some point interpreters do not feel a the empirical background understanding priori obliged to continue to complicate their which may influence how we weight the seri- interpretations to avoid yet another attribu- ousness of attributed divergence from this tion of inconsistency. model (but any divergence counts in some So interpreters now find themselves measure against the interpretation). So, when attributing some irrationality to their sub- we find agents to be reasoning in defective jects: the subjects are found to be marginally ways that we have come to expect, this counts diverging from the ideal that was initially against our interpretation, at least a little. supposed. Significantly, they will find a pat- According to the strong rationalist, there is tern in the divergence; they will find that always some ‘tax’ on any attributions of there seem to be systematic ways in which irrationality—so that an interpretation that pro- folk diverge from the normative model. ceeds smoothly and corresponds to our empiri- Interpreters will also come to appreciate cally informed expectations for certain forms of much about when folk in a given social con- irrationality will have yet thereby incurred an text. learn certain things, for example, and ‘error tax’ on its acceptability. Further attribu- what patterns of motivation are prevalent tions of irrationality—even if they conflict with within a society or within a profession within no empirical expectations, and even if these that society (recall the EE that seemed rele- expectations conform to EE-like expecta- vant when thinking about Tversky and tions—may then be difficult to sustain. The a Kahneman’s work). As suggested earlier, priori normative model continues to anchor such empirical expectations serve to con- and constrain our interpretations—and error strain an interpretation, and can add to the taxes on attributions of irrationality preclude Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 299

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 299

interpretive findings that are too much in as an evolving attractor. Because the divergence from that model. Such is the strong background empirical expectations—EE— rationalist picture. are not tethered to an invariant model along In contrast, the empiricist need count little the lines of [Lr], there is no a priori guarantee more than RRP as conceptually grounded. As that these background interpretations, and expectations for ways of reasoning, includ- the interpretations/ that they sup- ing forms of irrationality and forms of ratio- port, cannot give rise to deep challenges to nality, emerge in the course of empirical [Lr], in fact they themselves could already work drawing on acceptable interpretive reflect deep revisions of [Lr]. Thus, on the schemes, these constitute an evolving model empiricist understanding, there can be ade- of human cognition—one that then serves for quate interpretations that attribute deep vio- the empiricist as the model from which diver- lations of [Lr], and that thus can occasion gence is counted against an interpretation. revision in [Lr] treated as an account of On this view, there is no ‘tax’ on attributions human cognition. conforming to these expectations, this empir- How can one decide between the two ical model, even where these diverge from understandings of the principle of charity on the normative model from which (for pur- offer—the strong rationalist understanding poses of illustration we are supposing that) and the empiricist understanding (itself com- earlier interpretation took its departure. patible with a weak a priori element such as Focus now on the issues left hanging at the RRP)? To settle the matter would require close of our discussion of Kahneman and an extended reflection on the considerations Tversky’s challenge to RP/ [Lr]. The sugges- adduced in a range of interpretive inquiries tion was that a range of empirically informed (such as those found in history and cultural expectations of a diverse sort—EE— serve to anthropology) and on a range of investiga- provide support for an interpretation that has tions that suppose and sometimes reconsider our subjects systematically violating [Lr]. It interpretations (such as careful work in cog- would seem that such results can accumulate nitive psychology). For my own part, I am so as to support an understanding of human convinced that the empiricist model provides cognition that is deeply at odds with [Lr]. The the most adequate and best-motivated under- projected rationalist response was that such standing of the relevant inquiries, but devel- expectations were themselves dependent on oping the support for this conclusion is interpretations and thus hostage to [Lr], so beyond the scope of this article. that deep challenges to [Lr] were foreclosed. In effect, while [Lr] might be given some ‘tweaking’, it remains an invariant a priori REFERENCES AND SELECT attractor to which all interpretive results BIBLIOGRAPHY remain tethered. The strength of the tether may vary somewhat with empirical results, Bealer, G. (1987) ‘The philosophical limits but these themselves remain conditioned by of scientific ’, Philosophical interpretations tethered to [Lr]. Perspectives 1: 289–365. The empiricists think differently of the Burge, T. (1979) ‘ and the mental’, fundamental constraints on interpretation. Midwest Studies in Philosophy, 4: 73–121. On their view there is no such a priori, invari- —— (1992) ‘ and language: ant and substantive, model serving as an 1950–1990’, Philosophical Review, 101: 3–51. attractor for interpretation (nothing beyond Chalmers, D. (1996) . something like RRP). As empirical under- Oxford: . standings of human reasoning evolve, so does —— (2002a) ‘Sense and intension’, in the ‘attractor’ to which interpretations must J. Tomberlin (ed.), Philosophical Perspectives approximate—for those understandings con- 16: Language and Mind. Oxford: Blackwell, stitute the model that informs interpretation pp. 135–82. Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 300

300 RATIONALITY, COMPLEXITY, COLLECTIVITY

——— 2002b. ‘The Components of Content’, in ——— (1975b) ‘The analytic and the syn- D. Chalmers (ed.), Philosophy of Mind: thetic’, in Mind, Language and . Classical and Contemporary Readings. Oxford: Cambridge: Cambridge University Press, Oxford University Press, pp. 608–33. pp. 33–69. Davidson, D. (1980a) ‘Actions, reasons, and Peacocke, C. (1992) A Study of Concepts. causes’, in Essays on Actions and Events. Cambridge, MA: MIT Press. Oxford: Clarendon Press, pp. 1–19. Risjord, M. (2000) Woodcutters and ——— (1980b) ‘Mental events’, in Essays on Witchcraft. Albany: State University of New Actions and Events. Oxford: Clarendon Press, York Press. pp. 207–25. Quine, W. (1953) ‘Two of ’, ——— (1980c) ‘Towards a unified theory of in From a Logical Point of View. Cambridge, meaning and action’, Grazer Philosophical MA: Harvard University Press. Studies, 2: 1–12. ——— (1960) Word and Object. Cambridge, ——— (1982) ‘ of irrationality’, in R. MA: MIT Press. Wollheim and J. Hopkins (eds), Philosophical ——— (1970) ‘Philosophical in lan- Essays on Freud. Cambridge: Cambridge guage theory’, 1: 2–19. University Press, pp. 289–305. ——— (1981) ‘On the very Idea of a third ——— (1984a) ‘Radical interpretation’, in ’, Theories and Things. Cambridge, Inquiries into Truth and Interpretation. MA: Harvard University Press, pp. 38–42. Oxford: Clarendon Press, pp. 125–40. ——— (1987) ‘Indeterminacy of translation ——— (1984b) ‘Belief and the basis of again’, Journal of Philosophy 84: 5–10. meaning’, in Inquiries into Truth and Rosenberg, A. (1985) ‘Davidson’s unintended Interpretation. Oxford: Clarendon Press, attack on psychology’, in E. LaPore and pp. 141–54. B. McLaughlin (eds), Actions and Events: ——— (1984c) ‘Thought and talk’, in Inquiries Perspectives of the Philosophy of Donald into Truth and Interpretation. Oxford: Davidson. Worchester, MA: Blackwell, Clarendon Press, pp. 155–70. pp. 399–407. ——— (1984d) ‘On the very idea of a concep- ——— (1988) Philosophy of Social . tual scheme’, in Inquiries into Truth and Boulder, CO: Westview Press. Interpretation. Oxford: Clarendon Press, Stich, S. (1990) The Fragmentation of Reason. pp. 185–98. Cambridge, MA: MIT Press. Henderson, D. (1987) ‘The principle of charity Tversky, A. (1975) ‘A critique of expected utility and the problem of irrationality’, Synthese, theory: Descriptive and normative considera- 73: 225–52. tions’, Erkenntnis, 9: 163–73. ——— (1990) ‘An empirical basis for charity in Tversky, A. and Kahneman, D. (1974) translation’, Erkenntnis 32: 83–103. ‘Judgments under uncertainty: ——— (1993) Interpretation and Explanation and ’, Science 185: 1124–31. in the Human Sciences. Binghamton: State Whorf, B. (1956) ‘The punctal and segmenta- University of New York Press. tive aspects of verbs in Hopi’, in J.B. Carroll ——— (1994) ‘Conceptual schemes after (ed.), Language, Thought and Reality: Davidson’, in Gerhard Preyer, Frank Siebelt Selected Writings of Benjamin Lee Whorf. and Alexander Ulfig (eds), Language, Mind, Cambridge, MA: The MIT Press, pp. 51–6. and Epistemology: On Donald Davidson’s Philosophy. Dordrecht: Kluwer Academic Publishers, pp. 171–97. NOTES Jackson, P. (1998) From to : A Defense of Conceptual Analysis. Oxford: 1 Putnam (1975a) provides a useful discussion of Clarendon Press. this , which he then criticizes. Kripke, S. (1972) Naming and Necessity. 2 There is, of course, an obvious possibility which Cambridge, MA: Harvard University Press. would take the air out of the neo-rationalist program: Putnam, H. (1975a) ‘The meaning of meaning,’ the element of the semantics of concepts that is acces- in Mind, Language, and Reality: Philosophical sible to one who is conceptually competent, merely by Papers, vol. 2. Cambridge: Cambridge University virtue of that person being conceptually competent, Press, pp. 215–71. might turn out to be such a wimpy component of the Outwaite-3587-16.qxd 6/12/2007 5:32 PM Page 301

RATIONALITY AND RATIONALIST APPROACHES IN THE SOCIAL SCIENCES 301

conceptual semantics that traditional empiricist defla- would be a case in point. He thinks that there are sig- tionary responses seem appropriate. nificant, conceptually mandated, a priori constraints 3 Although the transition just now suggested on interpretation—ones that are ‘almost definitional’ reflects a grounds for regarding whether of action and related concepts—and then he insists there is any line to be drawn between conceptual that this renders such concepts unworkable for any truths (analytic claims) and truths that are central to respectable science. The conceptual constraints cer- empirically supported theories of some matter (which tainly do not seem trivial, as they amount to the idea would be synthetic). Famously, Quine (1953, 1960) that actions, beliefs, and desires interact so as to argued that central elements of our empirical largely conform to the rather elaborate dictates of theories or understandings may seem relatively safe decision theory. Since he thinks that the social from revision, but that this matter of degree should sciences are so constrained, he thinks that these are not be confused with the supposed status of being not respectable sciences, and would have us change ‘true by meaning’ or being ‘purely conceptually the subject of inquiry. The same verdict is applied grounded’—or any status that would make for a to any intentional psychology. In effect, Rosenberg prioricity. accommodates the rationalist brief presented here by 4 And to cut down on alternative understandings insisting that such concepts and constraints have no of choices/undertakings, we are to progressively con- place in any respectable empirical science. We do not strain our understanding of these in terms of wider study unicorns—for good empirical reasons. Neither sets of standing beliefs and desires attributable to the should we study actions and reasons. agent or agents. 6 Of course, the rationalist would insist that [Lr] 5 Those familiar with Rosenberg’s work will doubt- serves as an a priori truth regarding all agents—all less find is strange to read of him as advancing a who act for reasons—not just humans. But, given ‘rationalist’ position. This is a function of the early that humans are supposedly such agents, it would choice to treat the claim that there are significant a need to serve as an a priori truth of human cognition: priori principles as a mark of rationalism, even when if humans have beliefs and desires, if they act, then these principles are understood as conceptually [Lr] must (on the rationalist account) hold true of grounded. As explained, many contemporary self- human cognition. labeled ‘rationalists’ fall into this camp—for example, 7 For further development of these themes, see Peacocke, Chalmers and Bealer. Using the designa- Henderson (1991). tion ‘rationalist’ is so broad a fashion that one can be 8 For a more sustained development of this an empiricist and yet still be a rationalist. Rosenberg contrast, see Henderson (1993), chapters 2–3.