Practical Unreason

Mind Association

Practical Unreason Author(s): Philip Pettit and Michael Smith Source: Mind, New Series, Vol. 102, No. 405 (Jan., 1993), pp. 53-79 Published by: Oxford University Press on behalf of the Mind Association Stable URL: http://www.jstor.org/stable/2254172 Accessed: 27/10/2008 09:25 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=oup. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

Oxford University Press and Mind Association are collaborating with JSTOR to digitize, preserve and extend access to Mind.

http://www.jstor.org

Practical Unreason PHILIP PETTIT and MICHAEL SMITH

The philosophical literatureon failures of practicalreason generally takes categories of failure recognised in common-sense morality and in the philosophical tradition-weakness of will, compulsion, wantonnessand the like-and offers a reconstructionof what is involved in such failures. The approachis deferential; it casts philosophy in the role of underlabourerto received wisdom. In this paper we explore a methodologicallybolderapproachto practicalirrationality.We start with a distinctionbetween intentionaland deliberativeperspectiveson the explanation of action and we try to show how it can be used to generatea systematic taxonomy of the differenttypes of failure that we may expect to find in practical reason. The approachwhich we explore is not only methodologically bolder than the standardapproach;it also differs substantively.Some contemporarytheoriestreat phenomenalike weakness of will, compulsion and wantonnessas practicalfailures but not as failures of rationality:say, as failures of autonomy or whatever. Othercurrenttheories-the majority see the phenomenaas failuresof rationality but not as distinctively practicalfailures. They depict them as always involving a theoreticaldeficiency: a sort of ignorance,error,inattentionor illogic. They representthem as failures which are on a par with breakdownsof theoreticalreason; the failuresmay not have exact theoreticalanalogues,exact analoguesin the breakdownof belief, but they are of essentially the same, cognitive kind. Our approachgives us quite a different view of things. The pathologies which we identify in our taxonomyare distinctivelyrationalfailuresand distinctivelypractical failures;they are failures of pure practicalreason. The paperis in five main sections. In section one we introducethe distinction between the intentionaland the deliberativedimensions of decision-makingand in section two we describean ideal of practicalrationalityin which the intentional dimension is in resonancewith the deliberative.This puts us in a position, in the thirdsection, to look at the ways in which that resonancecan breakdown and at the correspondingfailures of practicalreason;the breakdownof resonancemay lead to outrightdissonance, as we describe it, or to a mere consonance between the two dimensions.The last two sections provide a commentaryon the position developed in this discussion. In section four we elaboratea little on the methodological and substantivefeaturesthatmarkit off from more standardapproaches. And in a final, concluding section we characteriseour approachas one under which the heteronomy characteristicof practical unreason contrasts, not with self-rule or autonomy,but with right rule or "orthonomy". Mind, Vol. 102 . 405 . January 1993

(? Oxford University Press 1993

54

Philip Pettit and Michael Smith

1. The intentionaland the deliberative Humanbeings, we assume, are deliberativeagents. As they face a choice, they are capable of registeringconsiderationsrelevant, by their own lights, to what should be done: thus they can register that these are the alternativeoptions and those the associatedpossible outcomes, thatone option has this set of desirable features,anothera differentset, and so on. They are capable,furthermore,of registering that the considerationsoverall supportone or anotherchoice: they can recognise the import of the desirablefeaturesregistered.And they are capable, finally,of being moved by such a patternof reasoning:they arecapableof making this or thatchoice in response to the recognitionthat it is the most strongly supportedalternative. We believe thathumanagents exercise their deliberativecapacityto a limited extent in almost every choice-this exercise, as we shall see, may be successful or unsuccessful-but in any case we shall be concernedonly with choices where there is deliberation.That agents regularly deliberate does not mean that they explicitly weigh the pros and cons relevant to every choice. We think that in approachingaction humanagents registerthe presence and the importof properties that argue for one or anotherchoice; that is why it is reasonableto ask an agent why she thought her action desirable or to ask how she could have been indifferentto featuresthatmade it clearly undesirable.But the deliberativeregisteringof the presence and importof such propertiesmay be a subliminalprocess thatis difficultto reconstructafterwards.Moreover,the process is usually going to be a very incomplete trainof reflection;it is going to directthe agent to some propertiesrelevantfor the choice but almost certainlynot to all. Where our first assumptionis that humanbeings are deliberativeagents, our secondis thatthey arealso intentionalsubjects.The mainelementin this assumption is the assertionthat when humansare moved by deliberativereasoning,not only do theirbeliefs aboutthe desirabilityof the featuresregisteredplay a role in generatingaction, there is also a role to be played by desires. Beliefs alone are not sufficientfor the productionof behaviour(Smith 1987). How can the intentionalconceptionof agentsbe squaredwith the deliberative? How can we find a role for desire in those cases where an agent registersthatone optionhas certaindesirablepropertiesthatareunregisteredin alternatives;where she registers that that option is therefore the most desirable; and where she is moved to action by thatreasoning? The straightforwardanswer to that question is that in such a case, the agent must desire to realise the propertiesdeemed desirable-she must prize or value those properties,at least in the circumstanceson hand-and she must desire their realisationwith sufficientstrengthfor this to lead to a desire to performthe option thatbearsthem:to performthatoptionratherthanany alternative.The idea is that as the agent registers the considerationsrelevant in deliberation,not only does she form appropriatebeliefs in the presence and import of the propertiesregistered;she also forms desires for the realisationof those propertiesandultimately,

Practical Unreason

55

as a net effect of such desires, she conceives a desire to performthe appropriate action. This answer, we say, is straightforward. Our attitude may reflect a third assumptionwe make, apartfrom the assumptionsassociatedwith the deliberative and intentionalconceptions.We assume thatwhen a humanagent comes to form a desire for this or that option among the alternativesthat face her in a decision, she does so as a resultof desiringto realise certainpropertiesthatshe expects the option or its outcome to instantiate.' In particular,she does so as a result of having a strongerdesire to realise those propertiesthanany desire she may have visa-vis the propertiesassociated with other options. If the activationof propertydesires generatesoption-desirein this way, then it is naturalto think that when deliberation issues in action, the agent forms desires for the deliberatively favoured propertiesand these are sufficient to produce the deliberatively supportedaction. Ourpictureof how the deliberativeand intentionalconceptionsgo togetherourpictureof how deliberationand desire fit with one another-raises an obvious question.Do agents always act as they deem to be most desirablein deliberation? Do their desires always answer to their desirability-beliefs?We conjecturethat they do not and thatthis is what createsan opening for a distinctiveform of practical unreason.We hold thatan agent may deliberativelyfavourone option without this impacting suitably on what she desires and what she does. She may choose a differentoption or she may choose the favouredoption but not for the reasons it is deliberativelysupported. Most of our paper amountsto an elaborationof this conjectureand we hope that thatelaboration,with examples, will make the conjectureplausible. But we think that it should be more or less obvious, in any case, that people can form desires that diverge from what they believe desirable.A heroin addictmay think thatthereis nothingat all to be said forjabbingthe needle into her veins; she may resent the "high"thatit gives her and may wish to be rid of the desire for heroin. Yet she may give herself the injection, and do so intentionally, none the less (Frankfurt1988). A woman may know full well that there is nothing at all to be said for drowningher baby in the bathwater,no considerationthatshould be outweighed by other reasons. Yet she may do so, out of a sudden whim, and do so intentionally: that is, do so on the usual belief-desire basis (Watson 1982). In cases like these, the action explicable from the intentionalperspective does not have properties in virtue of which it presents itself as desirable to the agent, althoughit may have propertiesthatengage with the agent's desires.

1 On property-desiressee Jackson 1985 and Pettit 199la: this tries to squarepropertydesires with a decision-theoreticframework.To desire an option is to preferit to the feasible alternatives.To desire a propertyis to be disposed, as between options or, more generally,prospectsthatotherwiseleave one indifferent,to prefera prospectwith the property to any prospectswithout.Thus, if the actualworld lacks the property,it is to preferthatit should have the property:it is to preferthe counterfactualworld in which the propertyis realised, assumingthatits realisationleaves otherthings equal.

56

Philip Pettit and Michael Smith

Such examples show that the intentionaland deliberativedimensionsof decision-makingmay indeedcome apart.2In particular,they show thatthe conclusion thata certainactionis or appearsmore desirablethanalternativesmay or may not go handin hand with an agent's desiringit. A person may conclude that a particular option is desirable, yet not desire it; and a person may desire a particular option, yet not believe it desirable.The divergence between the intentionaland the deliberative dimensions of decision-making is not surprising, on our preferredaccountof the concept of desirability.We take it thatan action is desirable in certaincircumstancesjust in case, if the agent were fully rational,she would desire that, were she in those circumstances,she performsan action of that kind (Smith 1992, Pettit and Smith forthcoming).3Given this analysis of the concept of desirabilityit is certainly possible for an agent to come to believe a certain action to be desirableand yet not desire to act in that way, and it is equally possible for her to desire to act in a certainway but not believe thatacting in thatway is desirable.And so we have an explanationof why the intentionaland the deliberativeperspectivesmay come apartin the way thatthey do. Moreover,given this analysis, we must also suppose that,otherthings being equal, an agent manifests a form of unreasonin not desiring to act in the way she believes desirable.For, by her own lights, she fails to desire to act in the way she would desire to act if she were fully rational.She is thereforeirrationalby her-own lights. And so we have an explanationof why, in agents who arein this respectrational,the two perspectives marchin step. Ourpurposein this paper,however, is not to defend this particularaccountof desirability,nor to addressotherproblemsrelatedto how the intentionaland the deliberativedimensionsof decision-makingcan come apart.Puttingthose issues aside, we conjecturethatan agent's desires can come apartfrom her deliberative judgmentsand our aim is to show how that hypothesis facilitates the characterisation of practicalunreason.4

2 We have addressedelsewhere some of the problemsgeneratedby the relationshipbetween the intentionaland the deliberativedimensions. See Pettit and Smith 1990, forthcoming; Pettit 1991a; Smith 1992. 3 For a characterisationof this "response-dependent" style of accountingfor concepts see Johnston1989, Pettit 1991b. 4 Among the issues we would like to put aside is the question of whether desire for something is always or ever necessitated just by the belief that that thing is desirable: whetherdesire can be a cognitive state. We write in a way that may favour non-cognitivism, arguingthat a failure of reason, in particulara failure unparalleledin the theoretical forum,can cause a divergencebetween desirability-beliefsand desires. But the cognitivist can give a congenial readingto the claim. The weak cognitivist will have no problem in doing so: she thinks that while a desirability-beliefnecessitates the presence of a correspondingdesire, it does not determinethe degree of strengthof thatdesire, and so she can regardit as a triumphof practicalreason that an agent forms a desire of the appropriate strength.The strongcognitivist, on the face of it, will face a problem.She goes beyond the weak position and holds that the necessitationof desire extends to its degree of strength. She will have to say thatif divergenceoccurs, then the desirability-beliefis not held on the properbasis or intemalisedin the properway or somethingof the kind. Thus she will have

Practical Unreason

57

2. Practical rationality Before looking at how the intentionaland deliberativedimensionscan diverge,it will be useful to examine what happens when they converge. Before looking at the differentmodes of practicalirrationality,it will be useful to examine what practicalrationalityinvolves. We shall give a sketch of what it is for a particular action to be rational.And then we shall add some details about what is required for an agent, as distinctfrom an action, to display rationality. Our discussion of the intentional and deliberative dimensions of decisionmaking alreadygives us a picture of how a rationalchoice will be made, and a rationalaction produced.The agent will register different desirability-relevant propertiesin the options, and in the likely outcomes of the options, before her: differentvalues which the options would instantiateor would be likely to instantiate;she will register,for example, thatreturningthis book would fulfil a promise, not returningit would breakone. The values registeredwill be propertiesin the light of which she tends to desire options, propertieswhich she cherishes or prizes. At a certain point, the propertiesconsidered-together, of course, with any relevant outcome-probabilities-will lead the agent to see one particular option, say returningthe book, as imperativeor prescriptive:as the thing for her to do. And this deliberativejudgment,this more or less explicit self-prescription as to what she ought now to do, will be matchedby a suitabledesire:a desire for the apparentlyprescriptiveoption. As the valued propertiescombine to support the deliberativejudgment,so they will combine to producea desirefor the option thatis deliberativelyfavoured. The dual aspect, deliberativeand intentional,of the propertiesregistered in decision-makingis the key to this pictureof rationalaction. In rationalaction the values which lead an agent to prescribeone option to herself-to see it as desirable, all things considered-are also the values which lead her to choose that option. The values that weigh with the agent in deliberationserve also to arouse a desire for the option which they deliberatively support.Their net impact in arousingdesire-their net desiderativeforce-corresponds to their net deliberative weight. As the agent deliberates,so does she desire. This pictureof rationalaction is drawnbriskly,as the details need not concern us, but there are a numberof points we should notice. a. We referto the perceptionof an option as prescriptiveor as desirableall things considered.This is the perceptionor judgment which the agent forms, having considered all things-or at least having considered all things that strikeher, in the circumstances,as relevant;it is the agent's

to acknowledgethatthe divergenceinvolves a cognitive dimensionand displays a certain parallel with failures of theoreticalreason. But the strong cognitivist will still be able to identify distinctive-in particular,distinctively practical-features in the failure described:the failure will not amountto any familiarkind of ignoranceor error,inattention or illogic. And so she too can endorsethe claim defendedhere.

58

Philip Pettit and Michael Smith

final or operativejudgment of desirability.Notice that the final judgment in this sense is distinctfrom the judgmentthatan option is desirable relative-to-all-considerations: thatit is, as we mightput it, desirableall-things-considered.(Davidson 1980) b. In the example of returningthe book, the propertywhich weighs with the agent is one thatthe choice of thatoption is boundto satisfy: thatof keeping a promise. We shall generally speak, for simplicity, as if the propertiesthat registerwith an agent in producinga choice are properties like this, which are certainof realisationby the appropriateoption. But it should be rememberedthat in most cases the propertiesthatregisterin decision-makingwill be just probabilisticallyconnectedwith the relevant option; they will be propertiesof outcomes which the option has only a certainprobabilityof bringingabout. c. We only mentionthe valuedpropertiesof options, ignoringtheirdisvalued counterparts.This is legitimate,as disvaluedor costly propertiesin any option can be representedas values or benefits of the alternatives. That any option has a given cost means that the alternativesconfer the benefit of avoiding thatcost. d. The valued propertiesof an option will make it seem prescriptiveor desirable,only given the weights which the agent attachesto the values in her reasoning.We allow that the weights attachedto values may be indeterminate,so that the judgment of desirabilitycan often be underdetermined.And we allow thatthe weights attachedto certainvalues may differ between differentagents. But we shall not be commentingexplicitly on those possibilities. So much for our picture of the rationalaction. What of the rationalagent? The rationalagent will certainlyproduce rationalactions. But she must not produce them as a matter of good luck; she must be someone who produces rational actions reliably. So what is going to be required for a person to be a reliable sourceof rationalactions? The net deliberativeweight of a set of values-the net supportit gives to the option prescribed-is determinedby the differentweights associated with each of those values. And the net desiderativeforce of a set of values-its net impact in producingdesire-is determinedby the forces associated with the desires for those values. If an agent is to be reliablyrationalin the choices she makes, if the net desiderativeforce is reliably to correspondto net deliberativeweight, then two conditionsmust be fulfilled. First, all the desiderative forces that determinewhat an agent does must be determinedby the deliberativeweights of values. Theremust be no desires of the kind that move the heroin addict and the distressed parent; there must be no desires that are formed without regardto values. And, second, the desiderative forces contributedby the differentvalues registeredmust correspondsuitablyto theirdeliberativeweights. The weight which a value is ascribedin the balanceof deliberationmust fix the force which the value exercises in the generation of desire;it must determinethe strengthwith which the agent desires to choose an option with that property.If the agent attaches a certain value to helping her friends,for example, then the strengthof the correspondingdesire should not be

Practical Unreason

59

so low thata considerationto which she gives lesser deliberativeweight can sway her in a differentdirection;and it should not be so high that a considerationto which she gives greaterweight is unableto deflect her inclination. Our image of the rationalagent, then, is one of a person in balance:a person in whom desiderativeforces are matchedto deliberativeweights. The language of weights and forces is metaphoricalbut it is not empty.5Thata value has a certain deliberativeweight means, more prosaically, that the agent is disposed to give it a certain importance vis-'a-vis other values. That a value has a certain desiderativeforce means thatthe recognitionof the presence of the propertyvalued generatesa desire with a certainstrength:with a certaincapacity to win out over the desires occasioned by other evaluations or occasioned exogenously. Wherethe measureof deliberativeweight is given by the agent'sreasoningpractices, the measureof desiderativeforce is given by her dispositionsto action. This completes what we need to say aboutpracticalrationality.It remainsonly to commenton an objection.Ouraccountmay be resistedon the groundthateven if an agent desires as she deliberates,and even if she does this reliably,the action which she produces on a given occasion may be irrationalin other ways. An action is irrational,it appears,if the deliberativeprescriptionis not actuallysupportedby the valued propertieswhich the agent registers,even if the desire she forms matches thatjudgment;in this case the agent displays inferentialfailure. Again, an action is apparentlyirrationalif the valued propertiesactively registered are not all of those which the agent takes as relevantin, say, earlierreflection, or if they are not weighed as in earlier reflection: the agent displays a selective or biassed attentionto the values on offer,being unfaithfulto her reflective perceptions. An action is apparentlyirrational,furthermore,if the valued propertiesregisteredby the agent do not actually belong to the options or outcomes which she surveys or if she wrongly surveys those options or outcomes, being mistakenabouttheirfeasibility or likelihood;here the agent is in error,we may put it, about mattersof value. And finally, an action is irrational,some will say, if the values registeredare not suitableor objective or whatever;in this case the agent can be said to be in ignoranceaboutmattersof value. The objectionraisedis fairenough. But we need not be particularlyconcerned, for it simply serves to remindus thatpracticalrationalitycan be more or less narrowly conceived. An agent's decision-makingmay certainlybe marredin any of the ways illustrated.And, to that extent, the agent may well be said to exhibit "practical irrationality"in her choice of action. But the form of irrationality exhibitedby an agent on such occasions is not especially practical,for the failure is a purely theoreticalone: it is a failurein the way she forms herjudgmentas to what is desirable all things considered. Our interest is in practicalirrationality, more narrowlyunderstood:if you like, in pure practicalunreason.We are concerned with the failures of practicalreason that can be exhibitedby agents quite independentlyof whether their deliberationsare flawed in theoreticalrespects. 5 The language can be misleading in other ways and needs to be used with care. See Pettit 1987.

60

Philip Pettit and Michael Smith

And so we need not worrythatactions and agentsmay not fail in this way but still count, in a wider picture,as practicallyirrational.We returnto this topic later.

3. Practical irrationality Given our picture of rationalaction and rationalagency, we can now approach the question of how agents may fail in the exercise of practical reason. We approachthe issue in a geometricalspirit.Firstwe devise a geometryin which to representrationalaction, then we indicatethe differentways in which this geometry may be disturbedand, finally,we identify each departurefrom the geometry with a more or less familiarpatternof practicalunreason. First, the geometry of rationalaction. Imagine two closed figures or spaces, each enclosing a range of points:

V

A

Callthe figureon the left the "values"spaceandthe figureon therightthe"actions" space. Let differentpoints in the values space representdifferentpackages of values thatmightbe recognisedby an agent. Since we areabstractingaway fromthe correctness of an agent's values, some points will representwhat some may regard as non-values. Thus one point might represent a package comprising just the value of prudence,anotherjust the value of beneficence,anotherjust the value of friendship,and anotherjust the value of fairness,while yet otherpoints represent packages:say, packagesof the values of friendshipandprudence,or of the values of beneficence and prudence,or of the values of fairness and prudence.And so on. In such packages, notice, the values are unweighted. Let different points in the actions space represent different options that an agent might choose. Think of the options as described in a way that does not reflect the values which support them; think of them as presented in a mode which makes them suitable items for the agent to control.Thus the descriptions under which a set of options present themselves remain constant, as the agent considersthe differentvalues thatthey promise, certainlyor probabilistically,to realise. In our now well worn example, the options are to returnthe book or not to returnit, whateverthe values thatthe agent comes to see on either side.

Practical Unreason

61

In orderto map rationalaction on a diagramconstructedout of these spaces, we need to introducetwo furtherrepresentationaldevices: a broken line and a solid line. The interpretationof these devices is of the utmost importance. Brokenline. The brokenline will always connect a point in the values space to a point in the actionsspace. If the values point is w and the actionspoint is b, then the interpretation of the line is this: given the alternative options, and given the relevant option-outcomeprobabilities,the value-set, w, and no set larger or smaller,leads the agent to see option b as prescriptive;6the agent weights those and other values in such a way that w supportsb. The agent may have registeredmany values not includedin w but the w -values are those in the light of which she judges that b is the best thing to choose: they are the values that serve in the circumstances to make b seem superiorto the other options. The non-w values which the agent may have registeredwill include the outweighed values presentonly in alternatives, and they will also include those values registeredin the favoured option which did not count with the agent: the values which did not serve in the determinationof the agent'sjudgmentas to what she should do. Solid line. The solid line will always end at a point in the actions space and may begin at a point in the values space or at a point in between. If it connects a values point, say x, with an actions point, say c, then it means: given the alternativeoptions, and given the relevantoption-outcomeprobabilities, the value-set, x, and no set larger or smaller,leads the agent to desire and choose c. If it ends at thatactions point, but does not reach back to the values space, then it means:withoutregard to any valuedproperties-any properties representedin the V space-the agent desires and chooses c. In order for certainx-values---or indeed for non-valued properties-to lead to desire and choice, the agent must have registered their presencebut, as in the other case, she will have registeredmany otherproperties too. Otherthings being given, the generativepropertiesarethose thatquickenthe agent's desire for the option chosen: those thatplay a causal role in giving rise to a preferencefor thatoption.

6 Two assumptionsto note, both made for reasons of simplicity.We assume, first,that there is always a single option which the agent sees as prescriptive.Of course there will also be cases where the agent sees two or more options as having equal claims and, as David Lewis has remindedus, there will be cases where the agent's weighting of the values is insufficientlydeterminateto fix one of a numberof options as the most desirable. We would have to stretchour geometricalresourcesin orderto representsuch cases, perhaps allowing a numberof brokenlines to originateat a given point in the values space. We assume, second, thatthereis no overdeterminationin the relationrepresentedby either line. We would also have to stretchour geometricalresourcesto representoverdetermination.

62

Philip Pettit and Michael Smith

With this frameworkin hand, the ideal of rationalaction we describedearlier can be representedas follows:

V

A

Fig 1: Reason vindicated. In this case thereis a vindicationof reasons.The agent is led by certainvalues, w, to see a certainoption, b, as prescriptive.And those same values, those same reasons, lead the agent to desire and choose b. Given the alternative,the fact that returningthe book will fulfil my promise,leads me to see thatoption as prescriptive. And thatvery fact leads me also to desire andchoose to returnthe book. Reason is vindicatedin my action. Even so, I may fail to be a practicallyrationalagent;I may producethe rational actionby good luck. We shouldrememberthatfor an agent to be rationalin making a certain choice, she must not only act rationally;she must be reliably disposed to produce such a rationalaction. She must not only act in a way that fits the diagram;she must be reliably disposed to act in that way. We returnto this point presently. Whenreasonis vindicated,whetheror not the agent is rational,a certainaction is judged to be right in the light of certainvalues and then that action is desired and chosen under the influence of those values. The right action is desired, and desiredfor the right deliberativereasons;deliberationand desire, as we may say, resonatein harmony. There are five ways in which this resonance may break down. The wrong action may be desired in three differentways: for the right deliberativereasons, for the wrong deliberativereasons, or for no deliberativereasons at all. And the rightactionmay be desiredin two differentways: for the wrong deliberativereasons or for no deliberativereasons at all. In the firstthree cases, the resonanceof reasonvindicatedgives way to dissonance,we shall say, whereasin the othertwo cases it gives way to consonance.7 7 Sometimes it may be to an agent's credit,by generallyacceptedcriteria,thatshe displays dissonance or consonance ratherthan resonance:this, because the deliberativepattern with which she breaks is not particularlycreditableand the achievement of narrow resonancelooks unattractivefrom a broaderperspective. Dissonance may be creditable, because it may be to an agent's creditthatshe is moved in desire by a propertyshe ignores or plays down in deliberation.And consonancemay be creditablefor the same sort of reason. For example, it may be to an agent's creditthat she is moved in desire by the considerationof just the honesty of an action when she takes accountat the level of deliberation

Practical Unreason

63

The five ways in which the resonanceof practicalrationalitymay breakdown are nicely representedin the five available ways of disturbingthe geometry of reason vindicated.Keep the dotted line that connects w and b in place, since all this means is that deliberationis present:there is one option that is seen by the agent, in the light of certain values, as the thing to do. There are five ways in which the solid line may then be varied, consistently with the interpretation given. Eitherthe solid line leads to a differentactions point from b (dissonance) or to b itself (consonance).If it leads to a differentpoint, then thereare threepossibilities:it begins from w (the right deliberativereasons), it begins from another point in the values space (the wrong deliberative reasons), or it begins from somewherein between the spaces (no deliberativereasons).If it leads to b itself, then there are two possibilities: it begins from a point other than w in the values space (the wrong deliberativereasons) or it begins from somewhere in between (no deliberativereasons). We now go on to characterisethese five forms of practicalunreason.It turns out thatthey are illustratedby common-or-gardenfailures. (i) Reason misfires

Agents do not always do what they take themselves to be justified, all things considered,in doing; they act in a deliberativelydissonantway. In one such case we can say that reason misfires. Certain values lead the agent to see a particular option as prescriptivebut, thoughthose same values lead her to desire and choose something,they lead her to desire and choose a differentoption from that which she sees as desirableall things considered.Reason misfiresbecause, while acting on a certainset of values, she acts in a way thatis not supported,in her own deliberativeview of things, by those values. This case can be representedas follows.

vI

Fia 2: Reason misfires. The value-set,w, leads the agent to see b as prescriptive,given the alternativeson offer. But that very same set of values leads her to desire and choose a different option, c.

of both its honesty and prudence:these values overdetermineher perceptionof the option as the right choice. Thanksto Denys Turnerfor a relatedcomment.

64

Philip Pettit and Michael Smith

Considerthe following example, by way of illustration.Suppose I value conveying informationclearly when I speak, but also value conveying that same informationhumorously.These values lead me to judge that, given the alternatives available,the best way for me to presenta lecture is by means of a certain mix of anecdotes and formal definitions.The definitionswon't do much for the humourof the occasion but they are necessary for clarity and I give clarity considerableweight, in particularmore weight thanhumour,in my deliberations. But now suppose that as I speak I find myself loathe to go to the definitions;I find myself sacrificingclarity to humourin a greaterdegree than I judge desirable. In this case, thoughI act on the basis of the very values thatdeterminemy all things consideredjudgement-clarity and humourin the conveying of information-I do not performthe action I take to be desirableall things considered.I do not convey informationwith the right mix of anecdote and definition. Reason misfires. In the misfiringof reason the relative importanceof the values which lead the agent to see one option as prescriptiveis not reflectedin the relative strengthof her desires for those properties.Given the deliberativeweighting of the values, one option seems prescriptive. Given the desiderative forces associated with those properties,a differentoption is desired and chosen. The agent's desire for clarityis too weak, or her desirefor humourtoo strong,or a combinationof these things obtains. In any case there is a failure of the balance requiredfor rational choice. Some have thought that an arbitrarychoice is requiredin deciding between such descriptions of the relative strength of an agent's desires (Watson 1987). Whatdifferenceis there,they ask, between the case in which the desires reflecting considerations of clarity are too weak and the case in which the desires reflectingconsiderationsof humouraretoo strong?Aren't these two ways of saying the same thing? Such scepticism is misplaced. We call a desire too strongor too weak dependingon whetherits strength,relative to the strengthof the agent's otherdesires, tracksthe deliberativeweight of the correspondingvalue: its weight relative to the weight of the values to which the agent's other desires correspond.And this fact about a desire would in turn emerge in decision-makingcontexts, actual or counterfactual.Thus, the desire reflecting a value that is habituallydefeated in action by the desires associated with other,less weighty values, is too weak; whereasthe desirereflectinga value thathabituallydefeats other,weightiervalues in action is too strong. But thoughreasonmisfiresbecause of a failureof balancebetween the weights and the degrees of strengthassociated with certain values, this imbalance does not mean that reason will always misfire. Consider someone whose degrees of desire as between values like clarity and humourare slightly out of alignment with the relative,deliberativeweights thatshe assigns to such properties.We can imagine thatthis person could make the right choice; she could choose the same option thatwould be chosen by the rationalagent, in whom deliberativeweights and desiderativeforces are perfectly aligned. We returnhere to an observation

Practical Unreason

65

made in connection with the case of reason vindicated.The agent who acts in a fashionthatvindicatesreasonmay not be rationalherself. She may not be reliably disposedto act in thatway; she may fail to be rationalby a degree which does not show up in this particularcontext of choice. The diagramwhich representsreason vindicated marks what is in common between the rational agent and the person envisaged here: their choosing the same option. But it also enables us to bringout the differencebetween them. For, given thatthe personenvisaged does not exemplify the requiredbalanceof deliberative weights and desiderativeforces, there are bound to be some situationsin which her reason will misfire in the mannerillustrated.There are bound to be at least some counterfactualdecisions wherethe options are such thatthe imbalance between the weights and forces leads her to desire and choose a differentoption from thatwhich she sees as prescriptive.In those situationsthe choice she makes will display the geometry of reason misfiring. (ii) Reason internally undermined

Thereis another,perhapsmore familiar,case in which an agent acts in a deliberatively dissonantway, failing to do what she takes herself to be all things consideredjustified in doing. In this case, the agent acts, and acts on the basis of one or more of her values, but does not act on the basis of the values which lead her to see a particularoption as desirableall thingsconsidered.Reason does not misfire, as the values the agent acts upon are not the very values which direct her deliberativeconclusion.Reason's verdictis underminedby values ignoredin the framing of that conclusion. It is internallyundermined,underminedfrom within, in the sense that at any rate it is values, and not any more exogenous forces, which cause the problem. We can representthis case as follows, in our second diagram.Given the alternatives, the set of values representedby w leads the agent to see the action, b, as prescriptive;but it is the set of values, x, which leads the agent to act and it leads her to choose action c, not b.

V

A

Fig 3: Reason internallyundermined Consideran exampleby way of illustration.SupposeI am in companyandpeople begin to make jokes at the expense of my absent friend. Though the jokes are funny,they are also moderatelyhurtful,sufficiently so that a good friend would not go along with them, though not sufficiently so that a complete stranger

66

Philip Pettit and Michael Smith

wouldn't rightly find them funny. In this situationconsiderationsof loyalty support my withdrawingfrom the conversation altogether,letting others carry on with the jokes if they so desire, whereas considerationsof humoursupportmy going along with the joke. However, all things considered,we will suppose, loyalty presentsgetting up and leaving as the option to perform,the one desirableall things considered. Imaginenow thatthoughloyalty leads me to see leaving as prescriptive,though the weight attachedto thatpropertyis greaterthanthe weight attachedto the fun of going along with the jokes, the strengthof my desire to be loyal is not correspondinglygreaterthan the strengthof my desire to enjoy and contributeto the humour.In this case I will stay and go along with the fun, despite my recognising that this is not the desirableoption all things considered.Reason will be undermined, and underminedby considerationsof the kind from which it takes its own lead. Reason will be undermined,as we say, from within.The explanationof reason's being underminedfrom within is that, though relative to the agent's other values, a certainvalue or value-set has a given weight, the desirefor thatvaluable propertydoes not have a correspondingdegree of strength,a correspondingforce, relativeto the desiresfor the othervaluedproperties.The possibilitiesdivide, then, as before.The desiremay be too weak relativeto those otherdesires,or thoseother desires may be too strong,or the case may involve both deviations. (iii) Reason externally undermined In the cases just described,in acting contraryto what she takes herself to be justified in doing all things considered,the agent still acts upon a value she has. But sometimes agents act intentionally and knowingly contraryto what they take themselves to be justified in doing all things considered,because they act on the basis of desires that do not reflect their values at all. Sometimes agents act in a deliberatively dissonant way, without acting in the light of any propertiesthat they value. The registeringof certainpropertiesmay serve to arouse desire and choice but the propertiesregistereddo not figureas values in theirdeliberations; they have no weight whatsoever. We can representthis case in our thirddiagram.

*

V

*l

i-~~ftc

i

A

Fig 4: Reason externallyundermined The value-set representedby w leads the agent to see option b as prescriptive, given the alternativesavailable.But the desire on which the agent acts is not gen-

Practical Unreason

67

eratedby those values: it is generatedwithout regardto value, so that the solid line does not begin in the values space. And thatdesire leads the agent to produce action c, not action b. We might illustratethis case by reference to the heroin addict, or indeed the distressedparent,that we mentionedearlier.The heroin addict describedthinks, not just that taking heroin is undesirableall things considered,but that there is nothing at all to be said on the side of taking heroin.And equally the distressed parentthinksthatdrowningher baby is undesirablein every possible respect.Yet the addicttakesheroin,the parentdrownsher baby,andin doing these thingsthey each act intentionally. We describethis sort of case as one where reasonis underminedfrom without, reason is externally undermined.Reason is underminedratherthan supported, because the agent does not choose the option which she sees as prescriptive;it exhibits deliberative dissonance. Reason is underminedratherthan misfiring, because the agent does not act on the values which lead her to see that option as prescriptive.Reason is externallyratherthaninternallyundermined,because the desire which producesher action is formed withoutregardto values of any kind. Reason is usurpedby a complete outsider,a desire thatreflects none of the considerationswhich weigh with the agent. (iv) Reason internallyunderpinned In most circumstancesin which we act therearemany reasonsfor doing what we judge we have most reason to do, all things considered.For many reasons converge on a single course of action in particularcircumstances.This fact is, we believe, now widely accepted. It serves to explain why, for example, in adjudicating the debate between consequentialism, deontology and commonsense moraltheory,we have to consider fantasticcases, not ordinarycases, in orderto see how these theories differ from each otherin theirpracticalupshot. However, even when different reasons all converge on a single course of action, thereis still a questionas to which reasonslead the agent to see the action as prescriptive.And, given thatquestion, thereis also the question as to whether the considerationswhich lead her to see the action as prescriptiveare the reasons which lead her to desire andchoose the action.Thus we can see roomfor a further failure of practicalreason, albeit a relatively benign one: reason is underpinned ratherthanundermined,for the agent does what reasonrequires,even if she does it for the "wrong"reasons.Reason is internallyunderpinned,because it is reasons or values, and not any more foreign influences, which lead the agent to perform that action. The case is representedby our fourthdiagram.The value-set w leads the agent to see option b as prescriptive,given the alternativesavailable. And the agent does indeedcome to desire andchoose b. But the agentis led to desireandchoose option b, not by the value-set, w, but ratherby a differentset, x.

68

Philip Pettit and Michael Smith

~~~ V

;

A

Fig 5: Reason internallyunderpinned Let's consideran example. Suppose that,in certaincircumstances,all things consideredI have most reason to returna book I have borrowedto the person who gave it to me. Among the values thatarerelevantare honesty andprudence,each of which requireme to returnthe book. Given the alternatives,the honesty considerations,andthose alone, lead me to see returningthe book as prescriptive.My all thingsconsideredjudgementis, as we might say, "determined"by honesty,not by prudence;the prudenceof returningthe book plays no role in my seeing that option as most desirable. This fact about my judgement can be capturedcounterfactuallyas follows. Imagine that I had believed that it was not prudentfor me to returnthe book to the personwho gave it to me, but still honest. Then I would still have found most reasonto returnthe book; I would still have seen the returningof the book as prescriptive.And if I had believed thathonesty had requiredme to give the book to someone else, though prudence still requiredme to returnit to the person who gave it to me, then I would have found most reason to give the book to someone else; I would have seen that action as prescriptive. But now imnagine furtherthatthoughI see returningthe book as prescriptivein light of the honesty considerations,I do not producethe choice in light of those considerations.RatherI produce it in light of the prudenceconsiderations,or in light of the prudenceconsiderationscombinedwith the honesty considerations.I do whatreasonprescribesbut I do not do it for the reasonsin virtueof which reason prescribesit. Reason is underpinnedby considerationsit does not invoke; it is underpinnedfrom within. This fact about the basis of my desire and choice can also be capturedcounterfactually.Suppose that it had not been honest to returnthe book to the person who gave it to me, though it remained prudent to do so; I might have found out she had stolen it. In that case I might still have given the book back: I would have done so if moved by prudencealone, or if moved by a combination of prudenceand honesty in which prudenceplays the more powerful role. Suppose on the other hand that, though honest, it had not been prudentfor me to return the book. In that case I might not have given it back: I would not

Practical Unreason

69

have done so in the case of being moved exclusively, or in majorpart, by prudence. The case we have in mind here is like the internalunderminingof reason, so far as the agent is led by one set of values in her deliberation,and by anotherin the formationof desire and choice. The difference between the two is that the desiderativelyeffective value-set leads in this case to the same choice, whereasit leads in the otherto a differentone. The action is deliberativelyconsonantrather than dissonantbut it does not display the resonance of desire with deliberation which we associate with reason vindicated.The agent does the right thing, intuitively, but for the wrong reason.8 The failure involved in the internalunderpinningof reason is obviously very differentfrom either of the firstthreefailures.It is benign ratherthanmalign, for the consonanceit secures is behaviourallyindistinguishablefrom the case where reason is vindicated.9The fact that reason can be underpinnedinternallyin this way means that there are devices imaginablewhereby I can try to ensure that I behave as reason requires,or otherscan try to ensurethis for me. Consideronce again the case where I go along with thejoke againstmy friend. Whateverthe source of the failure-whether it be a case of reason misfiringor reasonundermined-it might well be that,thoughI am disposed to go along with the joke, I wouldn't be disposed to go along with the joke if the company knew thatthe person at whose expense the jokes were being told was a friendof mine. For I would then have, as I now do not, reasons of reputationto quit. They would thinkbadly of me if I were to go along with the joke. Reasons of reputationare, perhaps,not the most admirablereasonsfor refrainingfrom going along with the joke. Certainlymy friend wouldn't be too pleased to find out that that is why I refrained. But these reasons might be enough to get me to do what reason requires.And so, concerned as I am with whether or not I do the right thing, I might findmyself with sufficientreasonto say "He's a friendof mine you know", so changingmy circumstances,and therebychangingthe reasonsavailableto me for refrainingfrom going along with the joke.1I 8 There are differentkinds of deliberativeconsonance-consonance as distinct from resonance-which our approachallows us to distinguish.One sort involves the replacement of the honestyconsiderations,to take the examplejust given. In this case the honesty considerationswould not have producedthe behaviouron their own. They are incapable of getting me to be honest aboutreturningthe book: either they are too weak for the job, to returnto a familiardichotomy,or the considerationsthatarguefor keeping the book are too strong.Thus it is only because the prudenceconsiderationsintervenethat I am saved from an underminingof practicalreason.Anothersort involves the buttressingor supplementationof the honesty considerations,ratherthan their replacement.Here I returnthe book in the light of a combinationof the honesty and prudenceconsiderations.The honesty considerationswould not have been sufficient on their own to get me to returnthe book-indeed the same may be trueof the prudenceconsiderationson theirown-but the combinationof both sorts of reasons is sufficientand indeed effective. 9 Benign? Perhapsnot by all lights. By some, as MarkSainsburyhas pointedout to us, "The last temptationis the greatest treason/Todo the right deed for the wrong reason" (T.S.Eliot,Murderin the Cathedral). 10 On this topic we are especially grateful to JeanetteKennett for helpful conversations. For similarthoughtssee Kennettforthcoming.

70

Philip Pettit and Michael Smith

But not only can I try myself to ensure that I do the right thing in such cases. Others,particularlythose others who have a hand in shaping the institutionsof our society, may try to ensure that I, like everyone else, do what most of us recognise as the right thing. The enterpriseof institutionaldesign, an enterprisethat is as old as democracyitself, is concernedprecisely with ensuringthat if people are not spontaneouslyvirtuous in this or that regard,if they do not do the right thing for the right reasons, then at least they will conform to virtue's demands; they will have reasons enough of other kinds to behave as the public good requires. Such reasons may be provided, under appropriateinstitutionalpressures, by fear of the law, fear for one's financialfortunes,fear for one's reputation, or whatever(Brennanand Pettit forthcoming). Even republicantheoristsof democracywho have arguedfor the need for public virtue,and who have seemed to stressthe need for virtueif institutionaldesign is to be successful, have often had in mind just behaviouralconformity to the demandsof virtue.Thus Tocquevillewrites of Montesquieuon virtue:"Wemust not take Montesquieu'sidea in a narrowsense... When this triumphof man over temptationresults from the weakness of the temptationor the considerationof personal interest,it does not constitute virtue in the eyes of the moralist,but it does enterinto Montesquieu'sconception,for he was speakingof the effect much more than the cause'.1 (v) Reason externally underpinned

There is a second sort of case in which an agent does the right thing but not for the right reasons. This also representsa variety of deliberativeconsonance that falls shortof resonance.In this case the agent sees a choice of action, b, as prescriptivein light of a value-set, w. But while the agent does producebehaviourb, she is not led to do so by the reasons provided by w. She produces b without regardto any values whatsoever.The case is representedin our last diagram.

0

0

V

- --

--s

A

Fig 6: Reason externallyunderpinned. Considerthe following example. As I am walking down a footpathI see a ladder leaning up againstthe wall. I reflecton my reasonsfor walkingunderit as against walking aroundit and decide that, all things considered,I have more reason to 1l Quoted from the preparatorynotes to Volume 2 of Democracy in America in Aron 1968, p. 201.

Practical Unreason

71

walk around.ThoughI can't see anyoneandcan't see any equipment,I know that laddersare usually put up against walls like this because people are working on the roof. If there are people working on the roof then there is some chance that they have equipmentnearthe ladder,and so thereis a chance that,if I walk under the ladder,I will be hit by a falling object. Since it is only mildly inconvenientto walk around,it is best to walk around.And so I walk around.Now it seems perfectly possible that,in such circumstances,though I walk aroundthe ladder,and so do what I have most reasonto do all things considered,I may not do so for any reasonthatI have. I may walk aroundthe ladderfor no deliberativereason. We have probablyall been introducedto the superstitionthat bad things happen to people who walk underladders.ImaginethatI walk aroundthe ladder,not because of believing the superstition-that would give me a reason, albeit a bad reason, to do so-but because my introductionto the superstitionin childhood has left me compulsively and reasonlessly inclined to do so. If someone asks me why I don't walk underthe ladderI have to answer:"I don't know, I just don't want to. I really don't want to". When I refrain from walking under a ladder because I really don't want to, in this way, then I refrainfrom doing so for no reason. Or so it seems to us. Now we see how it can be that someone who does what she is most justified in doing may yet do what she does for no reason.For she may do what she is most justifiedin doing on the basis of a desire-thatin no way reflectsany of her values. Reason is underpinned,as in the previous sort of case, but now it is underpinned from without;it is underpinnedby a more or less brutedesire.The actionis deliberativelyconsonantbut it is producedwithoutregardto any values or reasons. Withthis example described,othersshouldreadilycome to mind. One obvious example is a variationon a case introducedby Donald Davidson (1980). I am in bed but rememberthatI forgot to brushmy teeth. Suppose thatgiven the importance of dental health to me, I see getting up and washing my teeth as the thing to do; I see the choice as prescriptive.Whatis perfectlypossible, by analogy with the examplejust described,is thatwhile I do get up and wash my teeth, r do not do so underthe influence of the value of dentalhealth.I may do so out of a compulsive feeling of guilt or discomfortat lying in bed with my teeth unwashed,a feeling laid down in the drill and trainingof childhood.In this example, as in the other, my reason is underpinnedfrom without, underpinnedby a force which owes nothingto the influence of values.12 In discussing the internalunderpinningof reason we said that there are various devices whereby such underpinningis ensured and behavioural virtue is 12 We distinguishedsome varieties of consonance in a footnote to our discussion of internalunderpinning.In one kind the right reasons are replacedby the otherreasons and in anotherthey are supplementedby those reasons.Clearlythereare correspondingpossibilities here. The desire that I have without regardto value may replace the effect of the rightreasons,the case being one where those reasonsdo not possess the force to move me to action. Or the desire may supplementthe independentlyinadequateforce which those reasonshave. In orderto representthis last case, we would need to extend the resourcesof our geometry,allowing a solid line terminatingin an option to be forked, with one point of origin in the values space, the othernot.

72

Philip Pettit and Michael Smith

produced.A similar point obtains in this case. Much of the drill and training whereby we try to get our children to do what is right, by the values we instill in them, is likely to have the effect of inducing more or less compulsive, and perhapsguilt-driven, desires to behave in appropriateways. A trivial example might be getting them to clean their teeth every evening, with the sort of result illustrated above. But we can easily imagine examples of a more substantive import. As institutional measures may serve to prop up reason internally, so many of the means of moral education may serve to prop it up from without.

4. Comparisonsand contrasts We have distinguishedbetween the deliberativeand the intentionalperspectives on the explanationof action. We have arguedthat though, in the rationalagent, these perspectives march steadfastlyin step, in the irrationalagent, they all too often come apart.And we have provideda geometryof these failuresof practical reason,a geometrywhich directsus to the differentways in which actionscan be irrational.Reason may misfire, reason may be internally or externally undermined, or reason may be internally or externally underpinned.A resonance of desirewith deliberation,to invoke a differentmetaphor,may be replacedby a dissonance or a mere consonance. At thebeginningof thispaperwe saidthatourapproachis distinguishedfromthe establishedtraditionof discussingpracticalirrationalityby two features.First,it is not deferentialto common sense or the philosophicaltradition;it derives the differentsortsof practicalunreasonfromnovel premises,ratherthanstartingwith the received categories.And second, the failuresit identifiesare at once distinctively rationaland distinctively practical failures. We can now elaborateon these two matters.We will discuss themin reverseorder,beginningwith the second feature. Some treatmentsof practicalunreasonfail to preserve anythingof unreason, anythingof irrationality,in the phenomenadiscussed. They discuss phenomena like compulsionor wantonnessor weaknessof will but depictthem only as departuresfrom autonomy,for example: only as failures of self-commandor self-rule (Frankfurt1988; Bigelow, Dodds and Pargetter1988, 1990; Bigelow and Pargetterforthcoming).The idea here derivesfromKant,for whom reasonrequiresselfrule, but the interpretationenvisaged loses the connection with reason. Self-rule is takento meanjust the triumphof higher-orderdesires:the triumphof reflexive desires as to what desires to have. Ourapproach,by contrast,identifies a distinctive irrationalityinvolved in the differentways in which deliberationanddesirecan come apart.If an agentjudges that a certainoption is to be done, if she sincerely sees that option as best, then any failureto take thatjudgmentfully to heartis a failureof reason.It represents a failurethatis continuouswith the failureinvolved in believing certainproposi-

Practical Unreason

73

tions and seeing that a furtherpropositionfollows from them without being led to make any consequentadjustmentin one's beliefs. But the approach we have taken not only enables us to recognise distinctively rationalfailures at the origin of action; it also allow us to cast those failures as distinctively practical.Here there is a second point of contrastwith the contemporaryliterature. In our earlier discussion we mentioned a variety of theoretical ills that might beset deliberative judgment: ignorance or error in regardto the valuable propertiesregisteredin the differentoptions; selective or biassed attentionto the valuable propertiespresent, with the propertiesactually registered differing in identity or weight from those that are reflectively acknowledged;and illogic or inferentialfailure in the derivationof the deliberative conclusion: the conclusion derived is not actually supported,even in the agent's own lights, by the evaluative premises. Most approaches to practical irrationalityassimilate failures of practicalreason to one or other of these categories.13 Take, for example, the many differentways of understandingwhat is known in common sense as weakness of will. The age-old Socratic approach, under which virtue is knowledge, assimilates weakness of will to ignorance or error about relevant matters of value (McDowell 1979). And approaches that have commandedmore interest in recent times assimilate it to other theoreticalfailures. One assimilates it to the pathology of selective or biassed attention:the agent in reflection sees the options in one way, the agent in action sees them in another,so that the deliberativejudgment acted on is not the deliberativejudgment reflectively endorsed (Jackson 1984, Schick 1991). And another,popularised by Donald Davidson in particular,assimilatesit to inferentialfailure:by the agent's own lights, the evidence supports the deliberative judgment that one option is desirable-in Davidson's way of thinking, the agent judges that that option is desirable-all-things-considered,is desirable-relative-to-all-considerations-but the agent, in an inferentiallapse, forms and acts on the judgmentthat a differentoption is desirable(Davidson 1980).14 We agree that practical reason is plagued by ignorance, error,selective and biassed attention,and inferentialfailure. Not only that. We also think that these sorts of failure are of greatimportanceand thatthe literaturewhich characterises them makes an enormous contributionto our understandingof practicalunrea13

An exception is Michael Stocker,whose approachwe find congenial. See Stocker

1979. 14 Susan Hurley (1989, Chs. 7 and 8) introducesan interestingvariation.Under this approach,as underours, the inferentialinput to deliberativejudgment is a registeringof valued properties,such thatany one propertycan presentan option as desirablepro tanto, and can continue to present it as desirablepro tanto, even after the option is seen as not desirable simpliciter.Here there is a contrastwith Davidson (1980), for whom a prima facie considerationin supportof an option-the counterpartof the pro tanto supportceases to provide any supportif, all things considered, the option does not appearto be desirable. Given the apparatusof pro tanto reasons, one which we essentially endorse, Hurley arguesthat weakness of will is characterisedby acting on a pro tantojudgmentof desirabilityratherthana judgmentof desirabilitysimpliciter.

74

Philip Pettit and Michael Smith

son; what reservationswe have bear on mattersof detail.1IBut we believe that, however serious, the maladies characterisedin this literaturedo not exhaust the ways in which practicalreason may breakdown and, more particularly,thatthey neglect the breakdownsof pure practicalreason: the breakdownswhich are not distinctivelytheoreticalin character.All the maladies discussed affect the final, deliberativejudgmentof desirability:the judgmentof desirability,as we thinkof it, all things considered.But the breakdownsof purepracticalreason, the breakdowns which are not particularlytheoretical,bearon the connectionbetween the finaljudgmentof desirabilityand the agent's desire, not on the statusof the judgment of desirabilityitself. And it is such breakdownsthat are identifiedand taxonomised in the approachwe have takenhere. We all have a powerful,pretheoreticalintuitionthathumanagents can be fully cognisant of, and fully sensitive to, the reasons which supporttheir performing one action, and yet go on and perform another.The distinctive feature of our approachis that it supportsthis intuition.We recognise all the failures to which the ordinaryapproachesdrawattentionbut we give countenanceto otherfailures of reason as well: failures of pure practicalreason, failures which occur without any lack of cognisance or sensitivity on the partof the agents. So much for the substantive contrast between the approachtaken here and more standardapproachesto practicalunreason.The other featurewhich marks off our approachis methodological in characterratherthan substantive.It consists in the fact that we are not deferentialto the categoriesof common sense, or of the philosophical tradition,in delineatingthe possibilities of practicalunreason; we derive a taxonomyof failuresfrom novel ratherthanreceived premises. The literatureon practicalunreasonemphasises a variety of departuresfrom practicalreason or, understoodin a narrow sense, virtue. An agent can depart from virtue by displaying mere continence, for example, or by being weak of will, or compulsive or capricious. The main divide is between departuresfrom virtuethatresultin a rightactionfrom the deliberativepoint of view, as with continence, and departuresthatresult in a wrong action, as with weakness of will, or compulsion or caprice. Can our schema substantiate these distinctions? We believe thatit can. Weakness of will, compulsion and caprice are all instantiatedboth in cases where the agent does the wrong thing for the wrong deliberativereason and in cases where she does the wrong thing for no reason at all. So what is the difference between them?With all threephenomena,there is a mismatchbetween the degrees of desire presentin the agent and the values that she recognises in delib15 One matteris worthy of particularnotice here. Even under inferentialfailure, the rationalresponse will be to desire and choose the option that is seen as prescriptive.The rationalresponsewill have to match the mistake made in the deliberationwith a mistake, if you want to call it that, in the generationof desire and choice. But the rationalagent's dispositionsare unlikely to be able to producethe desire and choice requiredby an inferential mistake in a reliable way. And so the agent may find herself, happily,incapableof living up to herjudgment.This sortof inabilityshouldbe educative,as remarkedby Alison Maclntyre(1990) and JeanetteKennett(forthcoming).

Practical Unreason

75

eration.The differencebetween the phenomena,so we conjecture,relates to the causationand characterof this mismatch. A difference of causationmarks off caprice from the other two pathologies. With weakness of will and compulsion, the mismatchis somethingvisited upon the agent from without:it is a legacy of her nature,her past or whatever.With caprice, that is not so: the mismatchis somethingfor which she, as she is at the moment,is blameworthy;it involves a more or less wilful departurefrom reason. As a differenceof causationmarksoff caprice,so a differencein the characterof the mismatchmarksthe divide between weakness and compulsion.Roughly, we thinkthatit is appropriateto ascribeweakness of will when the mismatchis one that the agent is capable of handling:recognising where her desires are leading, she is capable of inhibitingtheir effect, say by reflectingon the long-term,more or less egoistic costs of following them. We thinkthatit is appropriateto ascribe compulsionratherthanweakness of will to the extent thatthis contemporarysort of self-control is not possible: to the extent that the agent is enslaved by the desires that move her away from the pathprescribedin deliberation. So much for actions that are compulsive or weak or capricious. Finally, we trunto continence, in particularthe continent agent. This type of agent is traditionally taken to be someone who struggles to do the right thing, and generally succeeds, making distinctive efforts of self-masteryor self-management:efforts which the virtuous agent does not need to make. Does our schema allow us to make sense of this picture?We believe it does. Continentactions must come out, on our approach,as right actions done for the wrong reasons. It is naturalto assume, then, thatthe continentagent is someone who producessuch continentactions andproducesthem non-accidentallyor reliably.And thatassumptionexplains why the continentagent fits the traditional image. If an agent is reliablyto producethe right action,but not for the rightreasons, then she must rely on providingherself with special incentives, or more or less blind habits, to get her moving in the right direction.She must equip herself with resourceswhich will wring from her a compliance that is not in her nature. She must make up for a lack of spontaneousvirtueby becoming a successful tactician in the artof self-management.In a word, she must conformto the received image of continence."6 We hope thatthese remarksareenough to show thatalthoughour taxonomyof purepracticalfailuresof reason is generatedby a non-trivialdistinctionbetween deliberativeand intentionaldimensions, although it does not startout from the received wisdom on practicalunreason,it does serve to make sense of received categories. The taxonomydirectsus to differentfailuresfrom those that are generally emphasised in the currentliterature,and connects'equally well with the long, partlycommon-sensicaltraditionof thinkingaboutpracticalirrationalities.

16

Thankshere to MarkJohnstonfor a helpful comment.

76

Philip Pettit and Michael Smith

5. Conclusion We mentionedin passing that Kant introducedthe idea that non-heteronomyis requiredfor practicalrationality.This idea gave rise to the characterisationof all formsof practicalunreasonas varietiesof governmentfrom without,government by somethingotherthanthe self. One way of summingup the approachadopted here is to show how it gives new, non-Kantianlife to this political imagery. Our concern has been with narrowpracticalrationality,as we have stressed. So what should such rationalityinvolve, in ternis of the metaphorof non-heteronomy? What should narrownon-heteronomybe takento require?The received line would say, autonomy:the rule of the self-the rule of the autos-rather than an alien rule. But on our approachthe naturalresponseis to say thatin the narrow sphereof practicalrationality,non-heteronomyis not self-rule or autonomy;it is right rule or "orthonomy"(Pettitand Smith 1990, p. 588). What is wrong with heteronomy,on our approach,is not that it involves the rule of the "heteros"in the sense of the exogenous; what is wrong with it is that it involves the rule of the "heteros"in the sense of the inappropriate.We see the non-heteronomous agent, the agent who is practically rational in the narrow sense, as someone in whom desire is appropriatelygoverned,notjust as someone in whom the governmentof desire is exercised by her. Thus we take a very differentview of non-heteronomyfrom post-Kantianexistentialistslike Sartrewho requireany operative desires to be affirmedin an act of radical choice (Sartre 1957, Part 4, Ch. 1). And equally we see things very differentlyfrom someone like HarryFrankfurt,who requiresany operativedesires, or at least any operative ground-leveldesires, to be desires that are endorseda level up: desires that the agent desires to act on (Frankfurt1988). Ourimage of non-heteronomyis driven by a more traditionalmetaphorof good governmentthan the democraticmetaphor which seems to inspire such visions. The good governmentof desire is a regime under which desire is faithful to the rule of deliberation;being endogenously inspiredand maintainedis not enough, even if it is necessary. The notion of orthonomy,however it contrastswith post-Kantianideals, connects up with the traditionwhich emphasisesthe requirementof executive virtues in a rationalagent. The non-executive or substantivevirtues requirean agent to be a lover of the good; the executive virtuesrequireher to be a good lover. Examples of virtues that are predominantly,if not exclusively, executive include temperance,courage, fortitude,and an impartialityacross times and persons:if you like, justice. As we see such virtues, they are requirementsor aspects of orthonomy.To be orthonomousrequiresa temperanceaboutthe things thatcan let loose uncontrollabledesires;a couragewhich does not let the desirefor one's own welfare excessively warp one's choices; a fortitude which enables one to bear up underadversity,maintaininga desiderativeconnection with the things one values; and an impartiality which keeps the claims of one's future self, and the claims of otherpersons, as powerfulin the generationof desire as more immediate counterparts.

Practical Unreason

77

Such executive virtues are derived in the Aristoteliantraditionfrom a preference for the middle way. It is not surprisingthatour notion of orthonomyshould be seen as a generalisedversion of the executive virtuesfor, intuitively,the ideal of orthonomyrepresentsa version of the Aristoteliandoctrineof the mean. Or at least it does to the extent that the doctrinebears on the narrowmatterof how an agent desires ratherthan the broaderquestion of what she desires.17The importantthingis not to assumecontrolof one's desires, as in the existentialistor quasiexistentialistvision. The importantthing is to be someone in whom desires are neithertoo strongnor too weak. It is to be someone in whom desiresaregenerated by values, and generatedwith forces equivalentto the weights that those values are accordedin deliberation.The forces must not fall short of the weights, nor must they rise in excess of them. The forces and the weights must be in balance. Ourconceptionof narrowpracticalrationalitythusgives us at least one reason, if we areto stick with the Kantianimagery,for takingnon-heteronomyas orthonomy ratherthan autonomy.But thereis also a more general considerationwhich favours this rendering.The ideal of right governmentmay be understoodnarrowly or broadly,dependingon how far we are preparedto specify the goals of the governors.As we have characterisedorthonomy,it describes only a narrow ideal of practicalrationality:an ideal of pure practicalreason. But that narrow ideal fits naturallyinto a broaderone: an ideal under which desire answers to deliberation,as the narrowideal requires,and deliberationitself escapes theoretical defects like illogic, inattention,errorand ignorance;an ideal underwhich the agent is substantivelyas well as executively virtuous. The fact that orthonomy can be understoodnarrowlyor broadlymeans that as an ideal of pure practical reason,it is continuouswith a fuller and moreroundedpictureof practicalrationality. As an ideal of pure practicalreason, it proclaimsits incompletenesson its face; it does not suggest, as the ideal of autonomyhas sometimesdone, thatit represents the be-all and the end-all of morality.And that, surely, is to its credit.'8 AustralianNational University, Canberra,ACT2601, Australia. Monash University, Clayton, Victoria3168, Australia.

17

PHILIPPETTIT

MICHAELSMITH

See Urmson 1980, especially the summaryon p.163. But see also Hursthouse1980-

81. 18 We are gratefulto Geoffrey Brennan,RichardHolton, Lloyd Humberstone,Jeanette Kennett,Rae Langton,Peter Menzies and MarkSainsburyfor helpful comments.We are also very gratefulfor the many helpful commentsthatwe received at presentationsof the paper,especially at its initial presentationin a seminaron "Valuein Action",held at Monash Universityin August 1991.

78

Philip Pettit and Michael Smith

REFERENCES Aron, Raymond 1968: Main Currents in Sociological Thought, Vol 1. Harmondsworth:PenguinBooks. Bacharach,Michael andHurley,Susan,eds., 1991: Essays on the Foundationsof Decision Theory.Oxford:OxfordUniversityPress. Bigelow, John, Dodds, Susan and Pargetter,Robert 1988: "Against the Will". Pacific Philosophical Quarterly,69, pp. 307-24. 1990: "Temptationand the Will". American Philosophical Quarterly,27, pp. 39-49. Bigelow, John and Pargetter,Robert forthcoming:"Autonomy and Integrity". MonashUniversity,mimeo. Brennan,Geoffrey and Philip Pettit forthcoming:"HandsInvisible and Intangible". Synthese. Charles,David and Lennon, Kathleen, eds., 1992: Reduction,Explanationand Realism. Oxford:OxfordUniversityPress. Dancy, Jonathan,ed., forthcoming:Reading Parfit. Oxford:Basil Blackwell. Davidson, Donald 1980: "How is Weakness of Will Possible" in his Essays on Actions and Events. Oxford:OxfordUniversityPress. Flanagan,0. andRorty,A.O., eds, 1990. Identity,Characterand Morality.Cambridge, Mass.: MIT Press. Frankfurt,Harry1988: "Freedomof the will and the concept of a person",in his The importanceof WhatWe Care About.Cambridge:CambridgeUniversity Press, 1988. Hurley, Susan 1989: NaturalReasons. New York: OxfordUniversityPress. Hursthouse,Rosalind, 1980-81: "A False Doctrineof the Mean".Proceedings of the AristotelianSociety, 81. Jackson,Frankl984: "Weaknessof Will". Mind, 93, pp.1-18. 1985: "InternalConflicts in Desires and Morals".AmericanPhilosophical Quarterly,22, pp.105-14. Johnston, Mark 1989: "Dispositional Theories of Value". Proceedings of the AristotelianSociety, Supp.Vol. 63, pp. 139-74. Kennett,Jeanetteforthcoming:"Mixed Motives". AustralasianJournal of Philosophy. Maclntyre,Alison 1990: "Is AkraticAction Always Irrational?",in Flanagan.0. and Rorty,A.O., eds., 1990. McDowell, John 1979: "Virtueand Reason".Monist, 62, pp. 331-350. Pettit,Philip 1987: "UtilitarianismwithoutUniversalisability".Mind, 96, pp. 7482. 1991a:"Decision Theory and Folk Psychology", in Bacharachand Hurley, eds, 1991, pp. 147-75. 1991b:"Realismand Response-dependence".Mind, 100, pp. 587-626. Pettit,Philip and Michael Smith 1990: "BackgroundingDesire". ThePhilosophical Review, 99, pp. 565-92. forthcoming:"Parfit'sP", in Dancy forthcoming.

Practical Unreason

79

Rorty,A.O., ed., 1980:Essays on Aristotle'sEthics. Berkeley:Universityof CaliforniaPress. Sartre,Jean Paul 1958: Being and Nothingness (tr. H.Barnes). London: Methuen. Schick, Frederic 1991: UnderstandingAction. Cambridge:CambridgeUniversity Press. Smith,Michael 1987: "TheHumeanTheoryof Motivation".Mind,96, pp. 36-6 1. 1992: "Valuing: Desiring or Believing?", in Charles and Lennon, eds., 1992. Stocker, Michael 1979: "Desiring the Bad: An Essay in Moral Psychology". Journal of Philosophy, 76, pp. 738-753. Urmson,J.0. 1980: "Aristotle'sDoctrineof the Mean",in A.O. Rorty,ed., 1980. Watson, Gary 1977: "Scepticism about Weakness of Will". Philosophical Review, 86, pp. 316-39. - 1982: "FreeAgency", in his Free Will. Oxford:OxfordUniversityPress.