Audi - A Kantian Intuitionism PDF

A Kantian Intuitionism
ROBERTAUDI
Kantfamouslysaid that one could not do moralitya worse disservicethan to derive

it from examples, and this pronouncement, taken together with his formulations
and explanationsof the categoricalimperative,has led some critics to regardhim as
too abstract.Ross, by contrast,has been widely viewed as taking individualcases of
duty to have a kind of epistemic priority over principles of duty, and some of his
critics have thus consideredhim insufficientlysystematic,or even dogmaticallylim-
ited to deliverancesof intuition. This paper arises from the conviction that under-
standing of the categorical imperative may be enhanced by reflection on Rossian
principles, and conversely.Kantand other systematicphilosopherswho have done
moral philosophy in the grand style have had too little faith in intuitive singular
moral judgement; Ross and other intuitionists have had too little faith in compre-
hensive moral theory. Drawing in part on an independent account of self-evidence
and its relation to intuition, the paper shows how a Rossianview can be integrated
with a Kantian moral theory in a way that yields the major benefits of both posi-
tions: the moral unification possible through the categoricalimperativeand other
notions prominent in Kantianethics, and the relativeclosenessto moral practiceof
Rossianprinciples.
Reflectivepeople who want moral guidance have often noted that the
help they get from moral theories,particularlyKantianismand utilitar-
ianism, is quite limited. From Kant'scategorical imperative or Mill's
principle of utility, for instance, there is often a long, uncharted dis-
tance to moral decision. Virtue ethics has sometimes appearedto fare
better than rule ethics on this score. But if rule theories seem to their
critics rigoristic or too abstract, virtue theories seem to their critics
unclear in applicationto action, lacking in principlesneeded to justify
moral decisions, or at best derivativefrom rule theories. Intuitionism,
at least in the historicallymost prominent version we find in Ross, is a
candidateto overcome deficienciesboth of virtue ethics and of single-
principle rule theories. But many philosophers regardintuitionism as
dogmatic or consider it inadequate because it lacks a comprehensive
moral theory as a basis for its disparateprinciples. It also shareswith
virtue theories-and arguablywith other rule theories-great diffi-
culty in providing a way to resolve conflicts of duties, those 'knotty
Mind,Vol. 110 . 439 . July2001 ? Oxford University Press 2001
This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

All use subject to JSTOR Terms and Conditions
602 RobertAudi
points' in ethics, as Mill calledthem, that are so salient in practicaleth-

ics.1
Sidgwick was conscious of such difficulties for moral theories. He
saw a need for moral principlesthat lie betweenoverarchingmoral the-
ory and intuitions about cases. Expressingpessimism about agreement
on 'ultimateprinciples'(for example,the principleof utility), he called
for 'middleaxioms'.2He offeredno list, but seemed to mean the sorts of
everydaymoral principles-such as those prohibitinglying, promise-
breaking, and punishing the innocent-that Mill called secondary
rules and Ross called principles of prima facie duty (Ross did not use
'middle axiom' for the latter3).This paper arguesthat a modified Ros-
sian intuitionism can largelyanswerSidgwick'scall for middle axioms
and shows how a Rossianview can be integratedwith a Kantianmoral
theory to yield the majorbenefits of both positions:the moral unifica-
tion possible through the categoricalimperative,and the relativeclose-
ness to moral practice of Rossian principles. There are two sorts of
benefitswe might hope for, either of which might justifymy project:in
ethicaltheory,a betterunderstandingof moral obligationand the justi-
fication of moraljudgements;in moral practice,an improvedabilityto
determinewhat to do. Progresstowardeither of these goals may easily
contributetowardachievingthe other,but it need not do so. The focus
of this paper is mainly theoretical,but my hope is to go some distance
towardboth goals.
1. The possibility of systematizing Rossian principles

One may wonder how, if Rossianprinciplesare axioms, they can medi-
ate betweensomething more basic and somethingless so. Can anything
be more basic than an axiom? Ross claimed self-evidence for his
principles-those expressing prima facie duties of fidelity (promise-
keeping, includinghonesty conceivedas fidelityto one'sword); repara-
tion; justice;gratitude;beneficence;self-improvement;and non-injury.
This status would qualify them as axioms; yet he also seemed to con-
sider them basic in a way that precludesan intermediatestatus appro-
'The referenceis to the closing passagesofch.2 of Mill (1959),esp. pp. 32-33.
2'Wemust remainas faras possible in the "regionof middle axioms".'See Sidgwick(1909),p. 8.
For an indicationof the importanceof middle axioms see SisselaBok'sintroductionto her edition
of this book (1998).
3 In the Foundationsof Ethics(1939),at least, Ross contrasts middle axioms with self-evident
principles.Of 'periodsin which mankind appearsto sink to a lower moral plane' he says,'Whatis
questioned ... is not the fundamentalprinciplesof moralitybut the mediaaxiomata,the rules for
which no a priorievidencecan be claimed...'(p. 21).Cf. pp. 69 and 190.

A KantianIntuitionism 603
priate to mediating between 'deeper' axioms and other principles. In a

famous passage he said:
That an act, qua fulfillinga promise, or qua effecting a just distributionof
good ... is primafacieright,is self-evident... in the sense that when we have
reachedsufficientmental maturityand have given sufficientattentionto the
proposition it is evident without any need of proof, or of evidence beyond
itself. It is evidentjust as a mathematicalaxiom, or the validityof a form of
inference,is evident ... In both cases we are dealing with propositions that
cannot be proved,but that just as certainlyneed no proof.4
A similar unprovability claim appears earlier in Moore and Prichard;5
but (as I have argued elsewhere6) neither the concept of self-evidence
nor anything in intuitionism as such precludes systematizing its moral
principles (in a way described below) in terms of a more general princi-
ple. Ross seems to have failed to see this, as others generally have,7 per-
haps in part because of the influence of familiar paradigms of the self-
evident that are-or appear to be-Aristotelian indemonstrables and
thus incapable of being evidenced by more basic propositions. Surely
we may think of a self-evident proposition as one that is evident in
itself, in the sense implying that adequate understanding of it, as
opposed to inferring it from prior premisses, yields justification for it.
The concept is not that of a proposition incapable of being evidenced
by, or derived from, something more basic or having independent sup-
port. What counts as an axiom, viewed in itself, may be a theorem,
viewed in relation to some other proposition.
4 Ibid., pp. 29-30. Cf. H. A. Prichard's

influential1912paper,'Does MoralPhilosophyRest on a
Mistake?',in his 1949work. The mistakeis 'supposingthe possibility of provingwhat can only be
apprehendeddirectlyby an act of moral thinking' (p. 16).
5Moore had said, of what he called intuitions, 'when I call such propositions "Intuitions",I
mean merelyto assertthat they are incapableof proof'. See his Prefaceto Moore (1993),originally
publishedin 1903,p. 36.
6A detailedcase is made in Audi (1999).
7There are some exceptions. ChristineSwanton, in a rigorousdefence of intuitionism (Swan-
ton 1987),saysthat 'thereis no reasonwhy an intuitionistcould not appealto such a conception in
groundingboth the first-orderprinciplesof the systemand the second-orderprinciplesfor resolv-
ing conflict ... a conception of human flourishing founded on an Aristoteliansystem of human
virtue ... Alternatively,the underlyingmoral conception could be contractualist,involvingan un-
derstandingof the point of moralityas a system which renderspossible co-operation amidst con-
flict of interest'(p. 175).This line is not, however,developed in relationto any account of self-evi-
dence nor shown to be an option for Ross in particular.See also WalterSinnott-Armstrong(1992)
for a formulation of the consistency of intuitionism with a kind of derivability.It is noteworthy
that Sidgwickspoke not only of'middle axioms' but of subordinateones: after granting that we
might have to qualify an apparentlyself-evident formula, he adds that we may wonder 'whether
we have not mistaken for an independent and ultimate axiom one that is really derivative'.See
Sidgwick(1962),p. 341.

604 RobertAudi
Another pertinent distinction is between proving and evidencing,in

the broad sense of providing (objective)reasonto believe. There might
be propositions that can be evidenced even if not proved. Given that
the concept of self-evidence does not rule out proof or evidencing by
somethingelse, it is noteworthythat in at least one place, Ross speaksas
if one prima facie duty might be derivablefrom another.8This remark
seems to acknowledge the possibility that despite many pronounce-
ments to the contraryby intuitionists,the self-evidentcan be evidenced
by something epistemicallyindependentof it.
This possibility suggestsa furtherone: a systematizationof the Ros-
sian duties by appealto a more comprehensiveset of groundsthan Ross
noted. If such a systematizationis realizable,then contraryto what the
dogmatism charge implies, it might provide both reasons for holding
the principlesand a source of correctivesfor false intuitions or merely
apparentmoral principles.By rejectingsome of Ross'smetaethics,then,
we can conceivehis principlesof duty and others like them-call them
Rossianprinciples-as candidatesfor middle axioms: middle because
they can be systematized by an overarching moral theory, axioms
because they are self-evident and can ground propositions plausibly
consideredtheoremsdeduciblefrom them. If, however,Rossianprinci-
ples can be in the middle, what might be at the top? There may be a
numberof good candidates,but the one I want to exploreis a versionof
the categoricalimperative.
Why did Ross apparentlynot considertrying to systematizehis prin-
ciples by appeal to an overarchingtheory such as Kant's?I have sug-
gestedone possible reason:he conceivedthe self-evidentas unprovable.
To be sure, not all systematizationrequiresproof, as opposed to weaker
connections between grounding and grounded propositions (such as
evidencing), but this may not have occurredto him.9 In any case, Ross
might have supposed that there is an ontological reason to reject any
systematizationof his principles.Regardingthe basis of right action, he
asks 'whether there is any general character which makes right acts
right',and he cites egoism and utilitarianism,with special emphasison
8He saysthat '[E]venbeforethe implicit undertakingto tell the truth was established[by a con-
tract] I had a duty not to tell lies, since to tell lies is primafacie to do a positive injuryto a person'
(Ross,1930),p. 55.This seems implicitlyto countenancea derivationof a duty of fidelityfrom one
of non-injury.
9 Evenif systematizationdid requireproof, there is a notion of provingthat does not entail the
use of self-evident premisses,as where we prove a complicatedtheorem from other complicated
ones that are not self-evident.Thereis also the everydayuse of 'proof' in which one may prove the
guilt of a criminal from known facts. In this kind of case we may not even have a valid deduction
and will have contingent premisses.

the 'ideal utilitarianism' of Moore,'1 as attempting to state one. He

answers in the negative; but he left-and there remains-much
unclarityabout what should count as a 'generalcharacter'underlying
the right.
One might think that provability of one proposition from another
entails certain ontic relations between the content of the former and
that of the latter,for instancethat if Ross'sprinciplesare provablefrom
the principle of utility, then each Rossian duty is constitutively
groundedin utility. But no such entailment holds. Epistemic relations
need not mirror ontic ones (at least in this simple way). For instance,
logically speaking, both the principle of utility and Ross's principles
could be grounded in a third element. We might speculate, moreover,
that Ross's preoccupation with providing a rule theory superior to
Moore's utilitarianismled him to underestimatethe possibility that a
Kantian theory would be a better candidate for comparison with his
view and even an ally.
To see the potential advantagesof integratinga Rossianintuitionism
with a Kantiantheory,it is instructiveto begin with a problemfor Ross.
He was keenly awareof conflicts between duties. He held that no gen-
eral ethicaltheory satisfactorilydealswith these conflictsand that prac-
tical wisdom is our best resource for them." In developing Rossian
intuitionism and integratingit with a Kantianperspective,I will argue
that the resulting comprehensivetheory can take us beyond a Rossian
application of practical wisdom in dealing with conflicts of duty. I
begin by outlining some broad characteristicsimportantfor appraising
the kind of comprehensivenormativetheory in question.
2. Normative completeness, epistemic completeness, and con-

flicts of duties
Let us call an ethical position, such as intuitionism or utilitarianism,
normativelycompleteprovided it accounts for every kind of deed that,
on balance, we (morally) ought (or ought not) to do, say to abstain
from harming others. Roughly,this is to say that it providesprinciples
or standardsin the light of which everyoverall(i.e., final) moral obliga-
tion can be plausiblyexhibited as such.'2Normative completeness is a
10Ross
(1930),p. 16.In ? 7 I brieflyaddresswhat might constitutesa 'generalcharacter'under-
lying the right;the entire paper is in part aimed at clarifyingdifferentkinds of underlying.
n These
points are in ch.2 of Ross (1930);see esp. pp. 41-42.
12
Threepoints should be noted here. First,the 'kinds'of deeds in question must be specifiedin
a generalway;we cannot, for example,list everykind of promise,say maritalor professional.Sec-

606 RobertAudi
kind of adequacycondition for a comprehensivemoral theory; it ena-

bles the theory, at least in principle, to take us from knowledge or a
plausible assumption of an overall (moral) obligation to a plausible
account (and ideally to knowledge) of why the action in question is
obligatory.
Suppose for the sake of argument that Ross considered his list of
prima facie duties normativelycomplete.If it is, then all our final duties
trace to one or more of these prima facie duties, and hence he might
view any finalduty as an overridingcase of one that he could explainas
resting on the relevantkind of ground. But he also saw practicalwis-
dom as essentialfor determiningfinal duty in at least the difficultcases
of conflicting duties. Depending on what practicalwisdom reveals, it
might or might not provide an explanationof why the duty it takes as
final is so. It might lead one to see exploitationunderlyingan ostensibly
permissiblebusinessdecision, and we might therebyexplainthe duty to
avoid it; but in a differentcase it might provide only an intuitive sense
of one option'sbeing morallypreferableto another.
Thereis a relatedkind of completenessthat a moral theory may have.
Suppose we start not with an obligation, but with (non-moral) facts
about our situation:our relationsto others, our resources,our capaci-
ties, and other facts relevantto what we ought to do. Supposeour main
question is not whether, given an obligation, our theory can account
for it, but whether,given the facts about our currentsituation,our the-
ory can tell us what we are (overall)obligatedto do. Calla moral theory
epistemicallycompleteprovided it enables us, at least in principle, to
determine from facts about our situation-facts that it may help us
identify-what we ought (overall)to do. Epistemiccompletenessis an
adequacy condition, or at least desideratum, in a moral theory that
enables it to take us from facts of human life-that we have friends,
that killing and deceit are rampant, that children are starving-to
what we ought to do.
Clearly,both kinds of completenessare important:we should be able
to explainand justifyascriptionsof obligation,and we thus want a nor-
mativelycomplete theory;we should (at least ideally)be able to deter-
mine what we ought to do from the morally neutral point of view of
'the facts', and so want an epistemically complete theory. We want
ond, normativecompletenessdoes not entail the correctnessof the theory in question;our termi-
nology allows that an ethical theory plausiblein the indicatedway is normativelycomplete in the
sense of'covering the relevantterritory'.The better the theory,the more readilyit gives us a basis
for knowledgeas opposed to just plausibility.Third,since my concern here is ethicaltheories, I am
ignoring the point that a normativetheory need not be specificallymoral.

knowledge both of what we should do and of why we should do it. A

theory needs epistemic completeness to provide the comprehensive
moral guidancewe seek as moral agents;it needs normativecomplete-
ness to enable us to explain and justify the moral judgementswe arrive
at on the basis of the facts that indicateour obligations.Moraltheories
can achievethese kinds of completenesswith varyingsuccess.Utilitari-
anism and Kantianismare plausibly viewed as normativelycomplete,
and at least the former may be plausiblyconsideredepistemicallycom-
plete. This apparent twofold completeness of utilitarian theories is a
major reason for their enduring attractiveness:they projectan empiri-
cal, factual route to moral knowledge. To what extent does Rossian
intuitionism approachcompletenessin either sense?
It would take a treatise to clarify in detail each of the wide-ranging
duties on Ross'slist. Intuitionism, moreover,is not limited to that list
(as Ross realizedin deliberatelynot claimingcompletenessfor it13);and
given the potentiallywide scope of Ross'sintuitionism I will not argue
directlyfor its normativecompleteness,but insteadconsiderthe pivotal
question of how the theory provides for resolving conflicts of duties.
Inability to deal adequatelywith these may be the largest obstacle to
regarding it as normatively complete. Two kinds of case are crucial
here.
Suppose a worthy charityasks a donation. For Ross, there is a prima
facie duty of beneficence,grounded in the goodness one may promote
by donating. I propose to say that a normative theory has first-order
normativecompleteness-as opposed to the kind of overallnormative
completeness described earlier-provided it accounts for every
instance in which, as here, we have a prima facie duty and hence (on
plausible assumptions) for all our basic duties: roughly, for each
instance of a (first-order)prima facie duty, it can specify something in
virtue of which it is a duty. Thus, its list of grounds of prima facie duty
will include at least one ground for each such duty; and the better the
theory,the more readilywe can determinesuch a ground for any given
prima facie duty we identify.
One might think that this achievement is trivially accomplished,
since we cannot identify a prima facie duty in the first place without
knowing its ground, as where we know we ought to A in virtue of
knowing we promisedto A. But this is not so. Moralsensitivitycan run
ahead of judgement. We may sense a duty to help someone, without
any good idea of whether the duty is a matter of a tacit promise or of
beneficence or both. We may thus respond to appropriate grounds
'3 He introducesthe list of duties 'without claimingcompletenessof finalityfor it' (p. 20).

608 RobertAudi
before forming a belief that they are present or, especially,making the
correspondingjudgement,here the judgementthat we should help the
person. This case illustratesthe application of a kind of particularism
regardingmoral judgement,as opposed to the subsumptivistview that
sound moral judgement requirestaking an act to be a of certain kind
(for example, a fulfilment of a duty of beneficence). Moreover, even
where moral sensitivity does not run ahead of judgement, judgement
can be producedby its grounds without our recognitionof the process
or even of the groundsas such.'4Thirdly,we may mistakenlyattributea
duty to one groundwhen it restson another.
Does Ross's intuitionism achieve even first-order normative com-
pleteness?Takenat face value, apparentlynot. SupposeI breaka prom-
ise to a friend because of a stronger obligation (owing to sudden
sicknessin my child). I ought to give my friendan explanation,and the
'ought' seems both moral and overriding. If I do not give it, I fail in
some moral respect.But which Rossianduty do I violate?If Ross'sview
has first-order normative completeness, then my not explaining my
failure must violate at least one. Explaining to my friend why I have
failed is not clearlya case of reparation,as my doing the promiseddeed
doubly well later on would be. One may wonder, then, whether Ross's
theory has even first-ordernormativecompleteness.
One defence of the theory is this. We might conceive my apologeti-
cally explainingmy failureas requiredby the duty of non-injury if not
by that of reparationor by both. Ross could say this, and it is plausible.
If the claim is sound, it illustratestwo interestingpoints. First, a Ros-
sian theory can countenance second-orderduties, such as the prima
facie duty to explaina failureto performa (first-order)duty.A second-
order duty of this kind derivesfrom his list, but is not on it. Second, a
theory can be complete in accounting for our duties, prima facie or
final, in more than one way. It might take all of them to be a matterof
meeting a single standard,such as maximizing pleasure;but a theory
might also accountfor duties disjunctively.It might construean obliga-
tory action as called for by at least one of two or more duties, but not
specifyone in particular.15 A complete moral theory need not, then, be
fully distributive:specifying, for each obligatory action, exactly which
duties it fulfils. Its standardsmay only collectively account for all our
duties.
14This point is arguedand illustratedin some detail in Audi (1998).
5
Accounting for a duty disjunctively may be conceived as distributing duties over actions,
whereasan ordinarydisjunctiveduty, say a duty of charity requiringone to give to cause x or to
causey, may be conceivedas distributingactions over duties.

A theory exhibitingfirst-ordernormativecompleteness,even if fully

distributive, need not exhibit overall normative completeness, i.e.,
completeness in accounting for final duty. Accounting for that is no
easy matter,and Ross did not claim to do it, if indeed he thought it pos-
sible. Might a Rossian theory do this? Might it account for, say, the
finality (roughly, overridingness)of my duty to aid my child (thereby
breaking my promise)? It is one thing to account for my prima facie
duty; it is much more difficultto account for my final duty. Ross denies
that rivaltheories can providea generalaccount of overriding,'6and he
himself suggestsonly a rough procedurefor this. His intuitionism does
not provide a theoreticallyplausible ground for the finality of a duty,
even though this is something we need to determine in some way
whenever we must resolve a conflict of duties or, especially,when we
must explain or justifysuch a resolution.
Let us say that a normativetheory that (like any plausiblemoral the-
ory) countenances conflicts of prima facie duty has second-ordernor-
mative completenessif and only if it accounts for the finality of any
(first-order) duty that prevails in such a conflict (and for the equal
stringency of two conflicting duties that are equally stringent).'7How
might we extend Rossian intuitionism to achieve both first- and sec-
ond-order completenessand thereby(for everydaylife) overallnorma-
tive completeness?l8
3. A Kantian approachto conflicts of duty

A broadlyKantiantheory can help in this by providingfor some exten-
sion and unification of Rossian intuitionism without weakening it
either epistemicallyor normatively.I stressthat I am speakingof a Kan-
tian theory,not specificallyKant's.The integrationof a broadlyKantian
theory with a Rossian intuitionism may indeed produce a significant
gain for the Kantian position as well. These points can be seen more
16I refer here to Ross (1930), ch 2. He implies, however,that 'the rival theory' (p. 23) is in no
better position, which confirms that his preoccupation here is with Moore. He seems not to be
consideringKantianismas providingan answer.
1 Since I am
only sketchinga normative theory, I largelyignore the point that one may have
conflictingsetsof duties, say two pulling one way and two pulling another.Moreover,I takea set of
duties to A that conflictswith a set of duties to B (whereA and B are incompatible)to be final only
if the deontic weight of the first set is greaterthan that of the second. If they are equally weighty,
presumablyone is morallyfree to A and to B (though the choice may be difficultor even in some
way tragic). Cf. Brink(1994).
s1Since the notion of completenessin question can applyat still higher orders,what I call over-
all completenessis not the most comprehensivekind possible;what is said about it herewill, how-
ever,suggesthow one might characterizeprogressivelyhigher-ordercompleteness.

610 RobertAudi
clearly if we first consider the kind of deliberation appropriate to decid-

ing which of two conflicting duties (if either) is overriding. Consider
the common kind of case in which I realize that my sick child could
have a serious setback and that only I can keep watch; I see that if I
break a promise to meet my friend, the friend will only have to make a
needless trip to the restaurant and, not finding me, phone to see what is
wrong. Now suppose that, following Kant's categorical imperative-
though not his own interpretation of it'9-I consider both universal-
ity and intrinsic end formulations of it.20
Kant's universality formulation is 'Act as if the maxim of your action
were to become through your will a universal law of nature'.2 I take the
explicit use of this to require asking whether one can (rationally) uni-
versalize one's maxim, say 'If the only way to keep my sick child safe is
to break a promise to a friend at the cost of inconvenience, but in a way
the friend would not (at least on careful reflection) resent, I will break
it'.22The intrinsic end formulation is 'Actin such a way that you always
treat humanity, whether in your own person or in the person of any other,
never simply as a means, but always at the same time as an end.'23Kant
apparently considered these formulations equivalent (though not nec-
essarily identical in content24). I leave aside the question of their equiv-
alence; in any case, they function differently. The universality
191 referto Kant'swidely known view, suggestedin the Groundwork and other work of his, that
a perfectduty,such as the duty to keep a promise,alwaysoutweighsan imperfectduty, such as the
duty to help someone in distress.Since perfect duties can conflict, even if Kantwere right about
the formercase,he would presumablyneed to appealto the categoricalimperative,in the way sug-
gested in the text, to deal with those conflicts.
2"There is disagreementover whether these are equivalentin content. I provisionallyassume
that if only becausethe intrinsicend formulationprovidesthe main materialsneeded for interpre-
tation of the universalityone, the equivalenceclaim is plausible. The falsity of this claim would
not, however,substantiallyalter my project.For a relatedstatement of how Kant'sethical system
may be viewed, includinga discussionof its apparentdeductivecharacter,see Donagan (1993).
21Kant (1961),p. 89.
22I insert 'rationally'to capture Kant'sintention and because it is in any case not plausible to
think the requirementconcernseitherpsychologicalor strictlogicalpossibility-as is well known,
in the GroundworkKant grantsthat there is no inconsistencyin universalizingthe maxim corre-
sponding to failureto do good deeds and to develop one's talents. In the application at hand we
could be more cautious and say 'could not reasonablyresent',but this may not be necessaryfor a
sound maxim in such cases.The notion of what is reasonablein such a case is by implicationclari-
fied in manypartsof this paper.A similarnotion is illuminatinglydiscussedby Scanlonin Scanlon
(1998).
23
Kant(1961),p. 96. Some translatorsuse 'merely'ratherthan 'simply',and I follow their prac-
tice as preferablein capturingthe relevantnotion as expressedin English.
241 am here
assuming that even logically equivalent propositions need not be identical, and
where the equivalenceis in addition synthetic, which is presumablywhat Kant intended, this as-
sumption seems plainlywarranted.

formulation is highly appropriateto testingacts for permissibility,but

not particularly helpful in discoveringwhat to do where we have no
promisingoptions to consider,since it is at best difficultto arriveat rel-
evant maxims to universalizeif one does not have a definite or at least
limited range of acts for which to formulate them. The intrinsic end
formulation is appropriateto both tasks:it articulatesa constraint on
maxims and therebyprovides a test for them, and it sets a twofold aim
for action and thereby indicates, if not specific options, at least the
directions in which to seek guidance of our conduct. Let us consider
how these formulationsmay help with conflictsof duties.
Would universalizingmy maxim permittingbreakingthe promise to
my friend undermine the practice of promising?Surelynot. We regu-
larly accept promises fully awarethat things like illness prevent their
fulfilment in cases where non-fulfilment is not serious. Would my
breaking my promise in such a case offend a reasonable promisee? I
believe not. Certainly I would not be offended or resentful if the cir-
cumstanceswere reversed.A problemthat remainsis how to formulate
maxims. This difficultycannot be discussedin detail here.25The intui-
tionist theory I am developing, however,does not depend on maxims
as a basis for singular moral judgement-particularly where there is
no conflict of duties-and in that respecthas an advantageover Kant's
own account. Indeed, given the way in which the theory focuses on
understandingthe categoricalimperativein relation to selection, for-
mulation, unification, and justification of Rossian principles and of
rules subordinateto them, we may often bypassthe problemof how to
formulate maxims-at least of the specific kind Kanthad in mind in
giving the four famous examplesin the Groundwork.
In any case, I suggestthat whetherwe are framingmaxims or testing
principles alreadyformulated,we might take Kant'sintrinsic end for-
mulation of the categoricalimperativeto be essential (at least for any
Kantiantheory) in determining what can be rationallyuniversalized.
Supposemy aim is to treathumanitynevermerelyas a meansbut at the
same time as an end. If I risked my child'shealth to keep a promise of
the kind in question, I would apparentlyfail to treatthe child as an end
in the relevantsense. That treatment implies not only that I value the
child's well-being for its own sake, but also that I am acting in a way
that, from the point of view of the universalityformulation, a rational
person in the child'ssituation could accept.Yetin breakingthe promise
I would not be using my friend merelyas a means. I would not (at least
25A brief account of how maxims should be formulated, with special emphasis on intention, is
given in ch.3 of Audi (1989).

612 RobertAudi
in any obvious way) use the friend at all, as I would by lying to get the
friend's car for a trip to the pub. Using others in the Kantian sense
seems to require their being one's means to some end (or at least an
attempted means that one in some sense uses). Perhaps more impor-
tant, since my explanation of breaking the promise would be accepted
by any reasonable person in the situation, the friend is not treated
objectionably. The intuitive idea is roughly that we may not use some-
one exploitively. Kant perhaps thought, or in any case we may plausibly
claim, that his view does not depend on any prior moral notion of
exploitation.26 The notion of exploitation we need has, as one anchor,
the idea of using something merely as an instrument: it matters only in
getting the job done; it may be damaged in the process and trashed
thereafter.
Ross might think that we have no theory of conflict resolution here,
only rules of thumb to facilitate using practical wisdom. We certainly
need practical wisdom to apply the categorical imperative, particularly
since duties to avoid treating people merely as means can conflict with
duties to treat them as ends. (Such conflicts are most likely for different
sets of people, and I assume that at least when other things are equal,
the avoidance of treating one set of them merely as a means takes prior-
ity over treating another set as an end.) Still, I cannot see that we do not
gain some help from the imperative beyond what we derive from just
gathering the facts in conflict cases and trying to make a wise decision
apart from reliance on such a principle. Indeed, Ross would agree that
if what we do is morally obligatory, it should in principle be describable
in a way that is generalizable. For he regards moral properties as conse-
quential upon natural ones, such as those involving the results of an
action for pleasure and pain, approval and resentment. If it is natural
facts, ultimately, that ground and justify our true moral judgements, it
is plausible to hold that- in principle-one could describe these
facts in a way that yields, for each sound moral judgement, a non-trivial
general description of its grounds. If they justify our judgement, then
(on plausible assumptions) we can become aware of them through suit-
able reflection and, given sufficient conceptual clarity, formulate a
description of them that expresses our justification.27 We may then for-
mulate a general principle. It may or may not be Rossian, in the sense
26 leave
open the prospects for developing a prior notion from a Kantianperspective.I also
leave a greatdeal unsaid regardingthe notions of treatingpersonsas ends (or merelyas means).
27The relevantdependenceof moral on naturalpropertiesis widely discussed in recentlitera-
ture and, in chs.4 and 5 of Audi (1997),is explicatedin a way that supportsmy points here. The rel-
evant passagesin Ross (1930)are mainly in chs.2 and 4. See esp. pp. 33,105,and 121-3.

implying that it is at least a candidate for self-evidence. Whether Ros-

sian or not, it should be in some sense subsumable under the categori-
cal imperative, and it may thereby clarify both the overall normative
content of that principle and the broadly Rossian framework with
which we are integrating the imperative.
If we further assume that a justified moral decision to act requires a
sense of the identity and bearing of the relevant facts, then we cannot
reasonably deny that the categorical imperative provides a test even for
judgements reached without its help. Ross might reasonably insist that
we can be guided by facts without being able (at least apart from
Socratic prodding) to articulate how they bear. Granted. This is indeed
a place where moral intuitions may guide moral judgement-and the
application of the categorical imperative. But his view implies no rea-
son to deny that the effort to articulate the bearing of facts is appropri-
ate and often successful; and he would surely grant that in a similar way
the categorical imperative may at least be intelligibly invoked where
duties conflict. The appeal to it will not always be conclusive; but given
the vagueness of moral terms, one would expect some recalcitrant bor-
derline cases, and the Kantian intuitionism we get by integrating a Ros-
sian framework with the categorical imperative may do as well on this
score as any plausible alternative theory. It might still turn out that
bringing this two-tiered theory to bear can always yield at least a mini-
mally satisfactory answer as to which of two (sets of) conflicting duties
is final.28 The theory would then have second-order normative com-
pleteness.
4. The beneficence problem

We can see further reason for integrating a Rossian intuitionism with a
Kantian view if we consider one of the most serious challenges any Ros-
sian theory faces: conflicts between the duty of beneficence and other
duties. Indeed, this duty poses serious problems for any plausible ethi-
cal view that takes the value of the welfare of persons to imply weighty
obligations to better their lives. It is not clear that any plausible theory
provides a solution to it that resolves all of associated theoretical wor-
ries or cuts through all the 'knotty points' Mill noted, those hard cases
where even the wisest may be uncertain. Why doesn't my duty of benef-
icence virtually always outweigh my ordinary fiduciary duties to, say,
my family, and nearly all my duties of self-improvement?
28It
may turn out that two conflicting duties are equally compelling, in which case we might
take our final duty to be disjunctiveand may select an option of a non-moral basis.

614 Robert Audi
Ross heightened the beneficence problem by describing the duty of

beneficenceas resting'on the mere fact that thereareotherbeings in the
world whose condition we can make better in respect of virtue, or of
intelligence,or of pleasure'.29 Still, he also stressed'the highly personal
characterof duty' and maintained (without explainingthe point) that,
other things equal, the duty of fidelity outweighs that of beneficence.30
A Rossianintuitionism can thus takeit that even a largecontributionto
the welfare of humanity does not necessarily outweigh all duties of
fidelityor all duties of self-improvement,and that there is no quantita-
tive criterion-such as maximizationof welfare-that we can appeal
to in deciding each case. There are at least two respects in which the
Kantianapproachoutlined here can help both to clarifythis point and
to rationalizeit. The first concerns the stringencyof the duty of benefi-
cence in relationto conflictingdutiesto avoidtreatinganyonemerelyas
a means-i.e., to meet the first demand of the intrinsic end formula-
tion. The second concerns what it is to treat others as ends, a further
demand of the intrinsic end formulation,and how understandingthat
demand may clarifythe duty of beneficence.Letus takethese in turn.
Supposing that all normal adults do have weighty duties of benefi-
cence, must we all become full-time philanthropists, at least until
human life on earth is far better?Must we, as certain consequentialist
views apparentlyimply,devote our lives chiefly,or at least overridingly,
to maximizing the good of persons (I leave open whether only actual
personscount and whetherother sentientbeings are to count)?Thereis
a sense in which, in submittingto this duty-as opposed to embracing
it out of a voluntary benevolent ideal-one would be using oneself
merelyas a means, or would at leastbe liableto doing this. Forinstance,
your personal commitments and talents would not matter at all inde-
pendently of their contribution to the general good in question (the
general good of persons includes one's own, but only as a tiny and
sometimes perhaps negligible part). To be sure, your good matters as
much as anyoneelse's,but this can be not at all if, say,you are incapable
of pleasure (or whatever intrinsic value is in question). This is not a
kind of treatment we can expect rational persons to agree to. Thus, if
two options enabledyou to contributeequallyto overallgoodness and
were otherwise alike except that, in pursuing one you could also
29Ross (1930),p. 21. Cf. Sidgwick(1907):'The Utilitariandoctrine ... is that each man ought to
considerthe happinessof any other as theoreticallyof equal importancewith his own, and only of
less importance practically,in so far as he is better able to realise the latter',p. 252. Clearly,this
leavesproblemsabout how to decide what we arebetterable to do and what degreeof prioritythat
determinationhas.
30
Ross (1930), pp. 21 and 17-18.

develop your talents and in pursuing the other you could not, then
althoughyou might permissiblypreferthe formerfor personalreasons,
developingyour talentswould be irrelevantto what, morally,you ought
to do. This lends colour to the idea that you would be using yourself
merely as a means: your interests matter only as possible contributors
to the overall good of persons conceived collectively.Similarly,even if
one did not try to maximize this overall good, but always preferred
large contributionsto it over one's personal commitments and talents,
one would, in a world like this, be liable at almost any point to be using
oneself merelyas a means.
A plausibleapplicationof the categoricalimperative,then, would at
least block an indiscriminatelarge-scalepreponderanceof the duty of
beneficence over competing duties. Giving it such automatic priority
ill-befits our dignity as persons. This is particularly evident where
someone harmsanotherperson for the sake of the overallgood of per-
sons, as in harvestingorgans from a salvageableaccidentvictim (with-
out consent). A defenderof autonomy might arguethat one has a right
to treat oneself so; but the right would be cancellableat will in a way
that implies that the associatedduty of beneficenceis self-imposed (say
by promises or vows one has made), a point also confirmedby the fact
that no one else would have a right to demand such beneficenceof one.
From here one might argue that-to put it in Rossianterms-there
is a prima facie duty not to use oneself even mainly as a means. But I
shall not pursue that line now: it is enough if the strong beneficence
thesis, which implies that one matters only in relation to the overall
good of persons, is justifiablyrejectableby a Kantianintuitionism.
One might wonderwhetherthis kind of use of the categoricalimper-
ative cuts the other way.Might omission of beneficentdeeds, at least in
a world like this, also in some way use others merelyas a means?Con-
sider deciding not to do any charitabledeeds towardthe poor, with the
resultingbenefit of retainingone's resources.In many cases this seems
morally reprehensible. But suppose it is. Still, one need not be using
them. (It is different,of course, if they are one's employees).Moreover,
suppose that-in the right spirit-one does some charitabledeeds.
This is one way of treatingas endsthose for whom one does such deeds.
To be sure, my suggestedstrategyfor applyingthe categoricalimpera-
tive does not provide a clearly successful procedure for deciding each
case (as opposed to implying-what is of great importance in con-
firming or disconfirmingmoral judgements-that once a case is cor-
rectly decided a universalizableprinciple is extractablefrom it31).But
intuitionism is not committed to the availability,in advanceof reflec-

616 RobertAudi
tion on actual conflicts of duty, of precise rules for deciding such confl-
icts.
A third respect in which Kant's intrinsic end formulation can help in
understanding the scope and stringency of the duty of beneficence
comes to light if we distinguish two interpretations of treating persons
as ends. On an impersonal reading, to treat people as ends is simply to
promote their good for its own sake, something possible when one has
no relationship to them, as where one contributes to a charity one sim-
ply knows, through assurances from a friend, is philanthropic. On a
personal reading, it applies only to people to whom one has some per-
sonal relationship. On this reading, you cannot fail to treat people as
ends if you have no personal relationship to them, since there is no way
you 'treat them'. The latter reading fits Kant's main illustrations better.
In applying the imperative to beneficence, for instance, he says that the
agent 'sees others who have to struggle with great hardships (and whom
he could easily help)'.32
Kant's theory as he develops it apparently does, however, counte-
nance at least indirect relationships, say where one has a definite
description adequate to provide a sense of who is in question, for
instance poor children in one's own city. Suppose these relationships
are such that we can fail to treat these children as ends. Still, doing or
giving something to help may suffice to prevent failure to treat them as
ends. It may express a recognition of their humanity and of one's obli-
gation to address it. Doing this- like many other ways of treating
someone as an end- is compatible with one's failing to do the best
thing one can. But on a Kantian theory, as for any Rossian intuitionism,
there need be no final duty always to do the best thing one can in pro-
moting the welfare of others.33This is supported by the intuitive point
that one cannot be said to fail to treat such children as ends simply
because one does not do as much as one would if one were maximizing
the overall good of persons.
A similar point holds if the intrinsic end formulation is applied to
suffering children we see on television. They certainly can give us a
sense of personal involvement. But here an additional point emerges.
31One worry is that in practicewe can extractonly such vague principlesas 'In cases like this,
preferspending money on educatingone's childrenover savingchildrenabroad',whereat best we
can be specificby listing so many circumstancesthat the principleis unlikelyto reapply.
32
See Kant (1961),?423, emphases added. The parenthesizedpoint may indicate that Kant is
thinking of the duty as restrictedin scope dependingon one's capacitiesand other factors.Cf. his
point that 'Tohelp others whenwe can is a duty'(?398),emphasisadded.
33This
point appliesto any plausiblemoral theory that countenancessupererogation,as most
versionsof utilitarianism,for example,arguablydo not.

The personal-relationship reading of the duty to treat others as ends

narrows, and may clarify, the field in which the duty of beneficence
conflicts with other duties. This reading goes only partway toward solv-
ing the beneficence problem, but by distinguishing the two readings we
can at least see that the issue is in part the extent to which moral obliga-
tion is personal and the degree to which it may arise from the mere
opportunity to do good. Kantian intuitionism takes it to be predomi-
nantly personal and, on principle, does not quantitatively define the
degree of predominance nor require a maximizing standard of benefi-
cence.
At least two points about the application of the categorical impera-
tive have now emerged. First, if we prefer duties of general beneficence,
in the ways that are intuitively objectionable, over (among others)
duties of fidelity or self-improvement, we are not only likely to use our-
selves merely as a means, but will also fail to treat certain others, such as
a promisee or our friends, as ends, since the pull of beneficence will
often override their justified claims on us. Second, where fulfilling a
duty of beneficence would result in either or (especially) both of these
deficiencies under the intrinsic end formulation, one is justified in giv-
ing a very high degree of preference for duties of self-improvement or,
especially, of fidelity, over the duty of beneficence (and much the same
would hold for other duties that might conflict with beneficence).
These points do not solve the beneficence problem, but they at least
enhance our theoretical resources for dealing with it.
5. Kantian intuitionism as providing a plausible grounding of

Rossian duties
There is at least one further and more general consideration suggesting
that a Kantian intuitionism is a good extension of a Rossian version. It
concerns the moral principles we would choose to live by in the first
place if the categorical imperative is our basic ethical principle.
Rossian principles of duty (though perhaps not exactly Ross's list)
may be argued to be just the general moral principles one would
derive-even if not strictly deduce-from a careful application of the
categorical imperative to everyday life. For instance, if one is to avoid
treating people merely as means, one must recognize (prima facie)
duties of non-injury (including avoidance of murder, brutality, and
deceit), of reparation, and of fidelity and honesty; and if one is to treat
people positively as ends, one must recognize duties of beneficence,
gratitude, self-improvement, and justice (meaning, as Ross intended in

618 RobertAudi
this context,34rectification of injustice one discovers, as opposed to

avoidanceof injustice,which is at least typicallya kind of injury).
The kind of referencewe find in Sidgwickand others to axioms and
to theorems ('middle axioms') invites expectation of strict deduction
and is to that extent misleadingabout Kantianintuitionism. Not every
derivation is a strict deduction, as opposed to, say, the provision of a
rationale. A deduction might require regimentation on the Kantian
side, say taking the duty not to treat people merely as means as abso-
lute, ratherthan suppose that the only absoluteduty underthe categor-
ical imperativein this formulationis-more plausibly,if less clearly-
to do deeds optimal with respectto jointly treatingpeople as ends and
avoiding treating them merely as means. Suppose, however, that the
prohibitional clause expresses absolute obligation. Then, with the
added (arguablynon-contingent) premissthat breakingpromisestends
to treat people merely as means, it follows (given the assumption that
treating people merely as means is absolutely wrong) that breaking
promises tends to be wrong, which, for the relevant, non-statistical
kind of tendency,is roughlyequivalentto its being prima facie wrong.
Similarpoints hold for the other Rossianduties. The derivabilityclaim
here is not that everyviolation of a Rossian (final) duty either treats
someone merely as a means or fails to treat someone as an end; it is
roughly that violations of these duties tend to do that, and that those
that do not haveclose affinitiesto those that do. Perhaps,with sufficient
qualification of the crucial Kantian and Rossian notions, the former
claim can be established,but even then, from the categoricalimperative
takenwithout artificialregimentation,we may be able to achieveonly a
weakerderivation,a rationaleratherthan a strict deduction, of Rossian
duties.
This derivationalstrategygives the categoricalimperative a double
role. First, it yields, at least in providing an inferential pathway to,
Ross's prima facie duties-and possibly some further, independent
ones. Second, as illustratedin the case of the sick child, it provides an
account (a non-quantitativeaccount, to be sure) of how to weight the
factors associated with those duties in cases of moral conflict. It does
this not by assigning an absolute weight to those factorsbut by telling
us what sorts of variables to consider and helping us to determine
which of the alternativeactions we then identify is morally acceptable.
Granted,one could perhapsjustifythe same first-orderprinciplesusing
34 I referto Ross'ssetting out of the prima facie duties in (1930),p. 21; on p. 35 he expressesthe
duty of justice differentlyand positively,as that of 'producinga distribution of goods in propor-
tion to merit'.

the principle of utility, as is suggestedby Sidgwick'sappealto his utili-

tarian principle as a basis for the principles endorsed by 'dogmatic'
intuitionists. But one could achieve such a justificationonly by invok-
ing auxiliary assumptions that are both contingent and quantitative,
whereasno parallelassumptionsare requiredin the derivationof Ros-
sian principlesfrom the categoricalimperative.35
Does the suggestedKantianintuitionism imply that (paceRoss) there
is some 'generalcharacterthat makesright acts right'?It does not. This
thesis about the ontology of rightness is left open. The position does
not deny that there is a plurality of moral reasons for action, say for
avoiding injury, for rectifying injustices, and for keeping promises.
These reasons may even be considered basic in three closely related
ways. (1) Any of them can be known to be morally relevant to action
without being derived from more fundamental considerations; for
instance,that one has injureda person (say,in rushingto make a train)
can be non-inferentiallyknown to be a moral reason to make repara-
tion. (2) Their normativeforce,conceivedas sufficientto justifyaction,
does not consist wholly in their indicating some otherfactor, such as
enhancementof pleasure;in the injurycase, for example,neithera duty
to relievepain nor any independentnon-moral considerationis needed
to ground the prima facie duty of reparation.(3) Our justified level of
confidence in the principles of prima facie duty can be higher, espe-
cially pretheoretically,than our justifiedlevel of confidencein the cate-
gorical imperative.
If Rossianprimafacie duties arebasic in this threefoldsense, they are
in a certain way irreducible, yet it does not follow from this Rossian
point that therecanbe no significantgeneralcharacteristicpossessedby
all duties or even some property that 'makes right acts right'. There
might still be some general characteristic,say befitting the dignity of
persons, such that, first, the moral reasonscan be known to be morally
relevantin the light of their connection with it and, second, they derive
from it normative force sufficient to justify the kinds of action for
which they are reasons.36The mode of this derivationof Rossianduties
is in part explained,by Kantianintuitionism, on the hypothesisthat the
categoricalimperativeis the centralprinciple-though perhapsnot the
35It is
helpful to comparethe case for a justificationof Rossianduties from rule-consequential-
ism made by Hooker in Hooker (1996), pp. 531-32.This paper is critically discussed by Philip
Stratton-Lakein (1997).Hooker providesfurthersupport for his view in Hooker (2000).
36
In the light of these points we can see that the Kantianintuitionist (or even the theoretically
enlightenedRossianintuitionist) could replyto ChristineKorsgaard'sworrythat, to the question

620 RobertAudi
only principle-expressing in very general terms what sort of action

befits the dignity of persons.
So viewed, the Kantianimperative also helps us to see a difference
between positive and negativeduties as well as associateddifferencesin
stringency often attributedto perfect as opposed to imperfect duties:
roughly, perfect duties (often intuitively negative) may be viewed as
those whose violation treatssomeone merelyas a means (roughly,in an
exploitive way) or shows a similar disregardfor them, and imperfect
duties (often intuitivelypositive) may be viewed as those whose viola-
tion does not necessarilydo this but does fail to treat someone as an
end (for example, without appropriateconcern for the person'sgood
for its own sake).37
In closing this section, I want to indicateone furtherreasonto think
that Kantiannotions can help to develop an intuitionism more system-
atic than Ross's.The notion of respectis centralfor understandingKan-
tian intuitionism,whose widest injunctionis that we must show respect
for persons.This notion is partiallyexplicableby appealto the concepts
of the dignity and autonomy of persons, which may be understood
both through Kant'swork and on their own terms. But the notion of
respect can also be clarified to some extent from a virtue-theoretic
point of view. Respectfor persons,as a trait of characterand not just an
attitude, is a moral virtue. If respect is partly constituted by less com-
prehensivevirtues such as justice,it is not just an equivalentof the vir-
tue of justice. Even if it is equivalent to some integrated group of
virtues, our understandingof them can contributeto our understand-
ing of respectand therebyto our understandingof the Rossianprinci-
ples of duty it helps to systematize.
This is not to suggest that the relevantvirtues can be entirelyunder-
stood apart from independent normative notions, such as some con-
cept of the intrinsicallygood and the moral notion of fair treatment.38
whetherone must face death ratherthan do a certaindeed, 'The realist'sanswerto this question is
simply "Yes"... This is of course especiallytroublesomewhen the rightnessof the action is sup-
posed to be self-evidentand knowablethroughintuition,so thatthereis nothing more to sayabout
it.'SeeKorsgaard(1996),p. 38 (emphasisaddedto suggestthat she herepresupposesthe self-evident
is ungroundable).
37 do not mean to put much weight on the clarityof the distinctionsbetweenpositiveand neg-
ative duties and between perfect and imperfectones. The (perfect) duty to keep promises, for ex-
ample, can be taken to be the duty not to break them, though it is more natural to express its
content positively.If it is positive, then a positive duty would be perfect.The Kantiandistinctions
partiallydevelopedin this papermay be a betterbasis for classifyingduties than such elusiveterms
as 'positive'and 'negative'takenindependentlyof them.
38That aretaicnotions are not understandableentirelyapart from other normative notions is
arguedin detail in Audi (1994).

But Ross could readily agree that our moral intuitions can clarify are-
taic notions, including those, such as that of respect, to which practical
wisdom may fruitfully appeal in understanding both the requirements
of our prima facie duties and the resolution of conflicts among them.
As for Kant, even supposing he had theoretical reason to resist an are-
taic conception of respect as a comprehensive moral notion, counte-
nancing such a conception in some significant role is nonetheless
consistent with affirming a central role for the categorical imperative in
normative ethics.39
There is, then, a plausible understanding of the categorical impera-
tive on which it provides a measure of clarification, explanation, and
(additional) justification for Rossian principles. None of these points is
highly restrictive regarding the ontology of ethics. The epistemological,
conceptual, and normative points essential for Kantian intuitionism
can probably be accommodated to a plausible version of constructivism
or even to noncognitivism. In any case, surely any ontology of ethics
adequate to the categorical imperative should be adequate to the Ros-
sian principles, and conversely; and I cannot see that any of the plausi-
ble candidates precludes the moderate commitments in moral
epistemology needed for Kantian intuitionism.40
6. Kantian intuitionism as a plausible development of Kantian

ethics
A Kantian intuitionist approach has a further merit. It helps in under-
standing central elements in Kantian ethics where it is at best quite
abstract, for instance concerning the notions of universalizability and
treating persons as ends. In at least three significant places Kantian
intuitionism can help to clarify and extend Kantian ethics conceived as
a system of moral principles unified by, and in an epistemic sense
groundable in, the categorical imperative. First, one can think of Ros-
sian principles as part of that system and indeed as formulating stand-
ards for treating people with the kind of respect appropriate to their
39It is
noteworthy that in the Critiqueof PracticalReason(1949) Kant ties respect not only to
'consciousnessof the directconstraintof the will throughlaw' (?117),which I taketo be conscious-
ness of a majoraspectof autonomy,but also, in TheDoctrineof Virtue(1964),to dignity,conceived
as groundedin the worth of persons:'man regardedas a person... is exaltedabove any price ... He
possesses, in other words, a dignity(an absolute inner worth) by which he exacts respectfor him-
self from all other rationalbeings:he can ... value himself on a footing of equalitywith them' See
?433-34, P. 99.
40 I omit considerationof Kant'stwo-worldsview. Interpretingthis is a majortask,and so faras
I can tell this paperis compatiblewith at least some plausibleinterpretationsof it.

622 RobertAudi
dignity. The principles are constraints on the use of the imperative:

interpretationsof it that do not yield the principles,and applicationsof
it that are inconsistentwith them, are prima facie defeatedby that fact.
Second,one can conceivethe basis of the appropriaterespectas the dig-
nity of persons, understood concretely as, in good part, a status pro-
tected by adherence to Rossian duties and as undermined more by
violation of a perfect duty like that of non-injury than by violation of
an imperfectone like that of beneficence.Third,if we take dignity to be
a fundamentalelement in a Kantianview, we can conceiveit to underlie
rights of autonomy and can take the proper sphere of personal auton-
omy to be that in which the Rossian duties, particularlythe negative
ones, areobserved.41 In outline, the idea is that each of us properlyexer-
cises autonomy up to that borderwhose crossing entails failurein our
duties to others. This border is not alwaysclearlymarkedand is often
undefended.
One might think that if the categoricalimperativecan serveas a basis
from which Rossianprinciples can be seen to follow (or to be implied
in some weaker sense), then they can hardly be said to clarify it, as
opposed to being part of what it implicitly says. But first, what is
implicitly said may be unnoticed even by those capable of seeing the
implication and, where the implication is by way of unobvious or
numerous intermediate steps, even denied by some who affirm the
principlein question. Second,far from being partof what a principlein
any intuitive sense says, some of its implications may never be discov-
ered:that there is a reliable,or even deductivelyvalid, inferentialpath
from one proposition to another does not entail that anyonewho con-
siders or even reflectson the first proposition will traversethat path to
the second. Theorems need not be discoveredby reflection on axioms
from which they follow, and sometimes axioms are not discovered
unless someone seeks grounds or explanationsor a unifying rationale
for theorems that follow from them. We can reason downwardto axi-
oms, as well as upwardfrom them.
These points about the possible relations between the categorical
imperative and Rossian principles are easily missed in contexts in
which, as here, we are consideringboththe categoricalimperativeand
normativeconsequencesof it that are clearlybefore us, such as the duty
of reparation.Nor would I deny that in fact any thoughtfulmoral agent
41 The second and third

point are supportedby a number of considerationsdeveloped in ch.n
of Audi (1997).I should add here that takingdignity to be part of what underliesautonomy rights
is compatible with considering the propertyof autonomy-roughly, a kind of capacity of self-
government-to be a basis of dignity.

reflectingon the categoricalimperativein the light of wide experience

of human life is likelyto think of many Rossianprinciplesas implicit in
it. But that may be so even if-and perhaps in part because-in the
natural order of discovery, Rossian principles precede such general
standardsas the categoricalimperative.Whateverthe case for the cate-
gorical imperative from general principles of practical reason, it is
doubtful that on the basis of those alone, a priori reasoningwould suf-
fice to articulatethe imperativewith sufficientspecificityto enable one
to derive all of the Rossian principles.42In any event, the order of dis-
coveryis neutralwith respectto the both logical and the epistemicrela-
tions. The first discovered may be derivable from what we discover
through it; yet our justification for the latter may be less strong than
our justificationfor the former.
Fromthe point of view of a Rossianintuitionism, it is naturalto dis-
tinguish two kinds of relation that Rossian principles can bear to the
categoricalimperativeor to any other candidateto ground them: one is
a specificationrelation,the other a derivationrelation-not mere logical
derivability,but an illuminatingkind of derivabilityhaving some justi-
ficatoryor explanatorypower. If Ross agreedthat his principles could
be inferred from a version of Kantiantheory, then, he might go on to
say that his principlesspecifywhat the categoricalimperativecomes to,
ratherthan being derivablefrom it in a way that gives it any justifica-
tory or explanatorypowertowardthem. Farfrom its telling us why they
hold, it is they that tell us what it says.A steadfastRossianmight claim
that Kantianism is clarified by intuitionism, but deny that the latter
gains much from integration with the former. A steadfast Kantian
might make the converseclaim.
Thereis a truth close to the specificationview, but it does not under-
mine the basic point that the categorical imperative frameworkpro-
vides both support and unification to Rossian duties. In a sense, any
non-trivial derivationof consequencesfrom a proposition specifies,at
least in part, what it comes to. Still, the categorical imperative has
meaning independentlyof the Rossianduties, howevermuch they clar-
ify it. For instance, from our understandingof instrumentalrelations,
we have a sense of what it is to treat someone merely as a means. We
regularlyuse tools and far too often similarly use other people. Here,
what happens to the tool is of no concern-unless we may need it for
42Kantsays, for example, 'Supposethere were something whoseexistencehas in
itselfan abso-
lute value, something which as an end in itselfcould be a ground of determinatelaws;then ... in it
alone, would therebe the ground of a possible categoricalimperative'.See Kant1961?428. Thereis
of course controversy about just how, in or outside Kant'swork, the categorical imperative is
grounded. Fora sustainedattempt inspiredby Kant,see Gewirth(1975).

624 RobertAudi
another job or happen to like it for its own sake. Granted, the notion of
a person as an end is technical, since ends in the ordinary sense are real-
izable; but from Kant's-and Ross's-writings we can acquire a sense
of what it is to treat someone as an end. For those we love, we do things
with no further end than some aspect of their good.
It may seem that since Rossian principles express prima facie duties,
and since our duties under the categorical imperative are final, the
former principles cannot be groundable in the latter. Indeed, some of
what Kant himself said may support this charge, since in places he
apparently took perfect duties, such as promissory ones, to be abso-
lute.43I have already suggested that if the imperative itself-apart from
any inferences from it such as might yield Rossian principles as
'theorems'-expresses any absolute duties, they are highly abstract.
Suppose there is an absolute duty always to act harmoniously with
some rationally universalizable maxim, and an absolute duty never to
treat people merely as means. To fulfil or even be guided by these duties
one must interpret such directives; and for reasons evident above, we
should not expect the plausible interpretations to yield only principles
of final duty. They may indeed yield no such principles except where
very general or quite open-ended language is used, as where it is said
that the only absolute duty is to do deeds optimal with respect to jointly
treating people as ends and avoiding treating them merely as means.
We saw that the categorical imperative is not plausibly thought to pro-
hibit all promise-breaking; and it is not clear that it would warrant, any
more than Ross would, an exceptionless rule requiring any specific act-
type without something like a ceteris paribus clause. Again, seeing the
categorical imperative in the light of its interpretation as guided by the
project of systematizing Rossian duties can significantly help in under-
standing Kantian ethics.
The categorical imperative, on the other hand, can also serve both to
connect the Rossian duties with one another and (as already suggested)
to provide a kind of rationale for them. Notice, for instance, how culpa-
ble failures to fulfil duties of gratitude, beneficence, and self-improve-
ment seem (to some degree) to be or to imply cases of failure to treat
one or more persons as ends, and (with a few exceptions that need not
be discussed here) culpable failures to fulfil duties of fidelity, repara-
tion, justice, and non-injury seem (to some degree) to be or to imply
43McNaughton(2000) might havebeen thinking of such passagesin saying,'I am more skepti-

cal than Audi about the possibilityof any other theory providingindependentsupport for a list of
duties of Ross'skind. Kantianism,for example,appearsto hold that some principlesare exception-
less, and not primafacie' (p. 283).

casesof treatingsomeone merelyas a means. These points interconnect

the Rossianduties as the standardsenjoined on us if we areto live up to
the imperative'srequiredtreatmentof persons;and the points provide
a rationale for the duties by exhibiting their fulfilment as obeying its
double-barrelledinjunction.44
It should now be clearthat just as our applicationof Rossianprinci-
ples and our resolution of conflicts among them may be affected by
how we interpretthe categoricalimperative,our understandingof the
categoricalimperativemay be affectedby what we learn from using it to
systematize such normative principles as those expressing Rossian
duties. Both sets of principles embody vague elements that provide, as
it were, open spacesto be filled by elements from the other set. Rossian
principles add determinacy to the notion of treating as an end; the
notion of treatingmerely as a means can make it clearthat what looks
like beneficencein a business decision is reallyexploitivebecauseit uses
someone merely as a means. The moral principlesin question are also
open-ended: indefinitely many kinds of acts can instantiate them, for
example by being cases of treating someone as an end or of helping
someone improve in knowledge.Here, too, understandingof the cate-
gorical imperative may be enhanced by reflection on Rossian princi-
ples, and conversely. Suppose an associate uses 'mitigate against' to
mean 'militateagainst'(as is not uncommon in America).To promote
knowledge,as a Rossianduty of beneficence,should I politely note the
substitution, or is this patronizingtowardan adult who is not my stu-
dent? It might be felt to be; but doesn't treatingothers as ends require
helping them avoid embarrassingslips?And wouldn'ta rationalperson
want to learn here?It turns out that how I can point out the slip is cru-
cial for whetherI should. This bears on both what counts as treating
someone as an end and on when the duty to enhance knowledge is
final. There may still be no clearly right choice; but reflection on the
Kantian notion can bring insight to the determination of Rossian
duties, and the habitualadherenceto those can set one in the right gen-
eral direction to understand and realize the standards abstractly
expressedin the categoricalimperative.
These points are largelyconceptual,concerningmoral concepts and
their application.Thereis a relatedepistemologicalpoint: our justifica-
tion for accepting the categoricalimperative can be enhanced by our
44I leave open whether adequatenon-moral concepts of treatingpeople as means and of treat-
ing them as ends may be devised from a Kantianperspective,so that applicationof the categorical
imperative does not require independent moral standards. If this is not so, then the Kantian
frameworkin question needs supplementationby, for instance,an intuitionist perspectivesuch as
the one presentedhere.

626 RobertAudi
justification for accepting the principles of duty it systematizes; and our

justification for accepting them may be enhanced by our awareness of
their support from 'above'-from the imperative-as well as by our
awareness of their being intuitively confirmed from 'below'-in appli-
cation to concrete moral cases about which we have clear convictions.
The systematization in question has many dimensions. One has been
stressed: the derivation of the target (systematized) propositions from
the grounding proposition(s). This may be by strict deduction or by
plausibility argument, as where the categorical imperative provides a
rationale for, but does not entail, a prohibition of conversational inter-
ruptions. Another dimension of the systematization is interconnecting
moral principles under a concept that figures in the grounding proposi-
tion, such as the concept of the dignity of persons. Dignity is central in
the status of persons as ends in themselves and broad enough to consti-
tute a main basis of Rossian duties.
Still another dimension of Kantian intuitionism is this. The categori-
cal imperative partially explains why Ross's principles hold. Its doing
this is not just a matter of an inferential connection. In part, the expla-
nation consists in presenting dignity as a basis on which people are
owed the duties. (I take Kantian ethics, broadly conceived, to include an
ideal of human dignity that goes beyond commitment to the categori-
cal imperative.) And there may be further elements, such as a ground-
ing principle's playing a role in resolving borderline cases. For instance,
where it is not clear whether an assurance amounts to a promise, it is
pertinent to ask whether if, without an overriding duty, one did not do
the deed in question, one would be using the person merely as a means
or at least treating the person disrespectfully.
One way in which this bi-directional justification can occur is this.
The categorical imperative can figure as at least part of our best expla-
nation of why Rossian principles of duty hold; hence, so far as we have
independent justification for them, we may gain justification for it (and
if they are self-evident, there is independent, non-inferential justifica-
tion for them). Similarly, so far as we have independent justification for
accepting the categorical imperative, we gain justification for the Ros-
sian principles as derivable from it (they can receive additional justifi-
cation even if they are self-evident). Moreover, insofar as the moral
importance of respect for persons-or a principle expressing this
value, such as that we must treat persons with respect-can explain
both the categorical imperative and the Rossian duties, both are unified
and receive some justification from their connection to this common
ground.

Adequately interpreted, then, the categorical imperative and the set

of Rossian principles may each help in clarifying the content of the
other, in applying the other to concrete moral decisions, and, in differ-
ent ways, in providing evidence for the other. We may, then, reasonably
seek to bring our interpretations of the categorical imperative broadly
understood and of Rossian propositions into reflective equilibrium.
Strictly conceived, reflective equilibrium is a relation among proposi-
tional attitudes; but it is achieved largely by adjusting them-especially
by refinement, deletion, and addition-in a way that yields a set of
them with contents that are mutually coherent, complementary in
explanatory power, and reciprocally clarifying. If the method seems to
foist a coherentist reading on intuitionism or on Kantian theories, it
should be stressed that on both approaches-and certainly on Kantian
intuitionism-reflective equilibrium is not considered confirmatory
unless the cognitions placed in it have some intuitive plausibility inde-
pendently of it.45
This procedure for achieving reflective equilibrium may seem impos-
sible in a territory where successful formulations may often capture
self-evident principles. It is not. First, in the matter of content, the
vagueness and open-endedness of the Kantian and Rossian principles
makes room for the procedure. Second, regarding justification, the
defeasibility of our justification for both kinds of principles makes
room for rational substitution of revised formulations for those discon-
firmed by disequilibrium. Third, regarding explanation, the complexity
and partial conceptual independence of the two sets of principles
makes possible explanatory connections in both directions. Once we
free ourselves from a narrow theory of self-evidence and from the con-
finement of intuition to self-evident propositions, our overarching
moral principle and the set of specific prima facie duty principles it
generates can each admit of clarification, justification, and explanation
from the other. To be sure, we may not have any comprehensive moral
principle deeper than the categorical imperative, but this does not pre-
clude every kind of grounding of such a principle in normative stand-
ards. We may at least consider it a reasonable principle to hold if respect
for persons is the overarching moral standard that, above all, it is to
express, and if the dignity of persons is the primary value that the insti-
tution of morality is to serve.
5 Complex epistemologicalissues arise here. A major source of the common-though by no
means universalor uncontested-view that the confirmatoryuse of reflectiveequilibriumimplies
a coherentistepistemology is Rawls (1971).The use of it I suggest comports betterwith moderate
foundationalism.Reasonsfor this are suggestedin the next paragraphbut are explainedin detail
in chs.i and 2 of Audi (2001).

628 RobertAudi
It may now seem that I have in effect suggested not just a Kantian
intuitionism but an intuitionist Kantianism. There is some truth in
that. I do not see how a Kantiantheory can be plausiblewithout both a
high degree of epistemic dependence on intuition and, in practical
application, a dependence on secondary principles, as Mill termed
them: roughly,categoricalimperativeswith a small ''.46But, althoughI
also do not see that anythingimportantin Kantianethics is falsifiedby
the suggestedaccount of how Kantianconsiderationscan yield a more
comprehensiveintuitionism,47the position I am proposingis intuition-
ist. It says,in bare outline, that we have a pluralityof moral obligations
expressiblein Rossianprinciplesof prima facie duty, and that, although
these are non-inferentiallyand intuitivelyjustified,they are systematiz-
able by, and stand in a mutually clarifyingrelation to, the categorical
imperative.
7. Between the middle axioms and moral decision: grounds of

duty
If a Kantian intuitionism is viable, and if principles of the kind Ross
proposed can serve as middle axioms, we have made a theoretical
advance. Kant famously said that one could not do morality a worse
disservicethan to derive it from examples,48and this pronouncement,
taken togetherwith his formulationsand explanationsof the categori-
cal imperative,has led some critics to regardhim as too abstract.Ross,
by contrast, has been viewed as a particularist, in taking individual
cases of duty to have a kind of epistemic priority over principles of
duty, and some of his critics have thus considered him insufficiently
systematic,or even dogmaticallylimited to deliverancesof intuition.49
With these interpretationsof Kantianand intuitionist ethics in mind,
one might call the former approach-at least as often apparent in
Kant's writings-top-down and the latter bottom-up. Kantian intui-
46 It is
noteworthythat in Foundationsof EthicsRoss spoke of 'Kant'sform of Intuitionism, in
which it is held [contraryto Ross] that the rightnessor wrongnessof an individualact can be in-
ferredwith certaintyfrom its fallingor not fallingunder a rule capableof being universalized'See
Ross 1939,p. 189.
47I distinguishbetweenthe theory Kantpresentedand his pronouncementsin interpretingit. I
cannot, for example, see that anything fundamental in the categorical imperative framework
makesall suicides immoral,as Kantis commonly read as holding in Kant(1961).
48See Kant(1961),?408.
49 For discussion of Ross'sparticularismand an indication of some common criticism of him
see ch.2 of Audi (1997), Dancy (1993), and Audi (1998b), which assesses some points in Dancy's
particularism.

tionism, by contrast, provides for the searchfor reflectiveequilibrium

to go in both directions and to yield adjustments or clarifications at
either end or anywherein between.
A Kantianintuitionism, then, can endorse a third strategy:up and
downfrom the middle,or indeed from any point between the 'top' and
the 'bottom'.This strategyis unlike Sidgwick'sand Moore'sin not sub-
ordinating Rossian principles to an overarchingmaster principle: the
systematizationachievableby the categoricalimperative may depend
on, or lead to, reinterpreting(or even modifying) our interpretationof
it, as well as to reinterpreting (or modifying) Rossian principles.50
Indeed, Kantianintuitionism is compatiblewith the plausibleview that
our justified confidence level is higher for Rossian principles than for
the categoricalimperativeand, in some cases,higherfor singularmoral
judgements than for any Rossianprinciplesthat subsume them. Com-
mitment to the integrationcalled for by Kantianintuitionism is neutral
with respect to epistemic priority. Purists in the Kantianand Rossian
traditions may divide here, the former giving epistemic priority to the
'top' the latter to the 'bottom'.I cannot see that either level must have
hegemony. Kantianintuitionism also allows that justified confidence
can not only changewith new or lost evidence,but reversebetweenlev-
els, such as the level of singular moral judgement and that of Rossian
principle.Thereis no one placewe must startin ethicalreflection;there
are many directions we may go once we begin; and no one level has
unqualifiedpriorityover any other. Much remainsto be said, however,
to fill out the Kantianintuitionist view. The most importantremaining
question is perhapshow this frameworkcan approachepistemic com-
pleteness nearly enough to give us fruitful general results in practical
(or 'applied')ethics.
To see how Kantianintuitionism applies in practicalethics, we must
view moral obligation not only from the top down, as is naturalwhen
the emphasisis on applyingprinciples,but also from the bottom up. It
is useful to begin with a question Ross did not adequatelyaddress.To
what extent can we factuallyspecifythe groundsof duty?Can we say in
(non-moral) factual terms what it is to promise,or for someone to be
injured,or to needour help?Such questions are crucialfor the problem
of the epistemic completeness of Kantian intuitionism, and they are
importantfor a number of reasons.
50Sidgwickheld, for example,that 'Utilitarianism... must be acceptedas overrulingIntuition-

ism and Egoism'(1907),p. 420) and 'must show to the Intuitionistthat the principlesof Truth,Jus-
tice, etc. have only a dependentand subordinatevalidity ...' (p. 421), footnote omitted.

630 RobertAudi
First,if any set of principlesis to be an independentlygood guide in

moral matters, we need a way to apply them that does not require an
appeal to an independent basis of moral judgement, say an independ-
ent standardof justiceby which we identifya Rossianduty of justice. If,
moreover,we want a theory that approachesepistemic completeness,
we need a way to proceedfrom relevantfacts to overallobligation. Sec-
ond, in addition to requiring the guidance of facts in applying moral
principles,we need at least a rough account of the sorts of facts that call
for moral decision or moral action. This point is easy to ignore because
the literatureof ethics is dominatedby problemsawaitingjudgement;it
pays too little attention to conditions under which moral decision or
action is called for in the first place. Not everything we do is morally
significant.When must we make a moral decision, and when may we
simply pursue non-moral ends?Third,we must teach moral principles
initially by pointing to such facts as one person's physically hurting
another.Childrenneed experienceof these objectivegroundsfor judge-
ment before they can develop moral concepts. How may these morally
significantfactsbe fruitfullyconceivedfrom the point of view of ethical
theory? Fourth, in practice, at least, we must sometimes ultimately
explain or justify a moral judgement by appeal to facts, as opposed to
citing some other moral judgement.This point is a constrainton nor-
mativelycomplete theories;it is doubtful that our judgementsof over-
all obligationcan be adequatelyexplainedor justifiedwithout appealto
facts at certain points, particularlywhen they result from resolving a
conflictof duties. The point does not imply reducibilityof moraljudge-
ments to factualones, but the formercan and often should be shown to
be groundedin the latter.51 What sorts of factshavethe basic (or at least
special) moral significancethat this explanationand epistemicground-
ing requires?
Ross'slist of grounds of duty is far from being entirely factual;but
this apparently did not seem to him to pose a serious problem. He
seemed quite unawareof it in setting out his list of duties:the duties of
reparation'rest on a previous wrongfulact';the duties of justice con-
cern distributions 'not in accord with the merit of the persons con-
cerned'; and the duties of both beneficence and self-improvement
requiremaking someone 'better in respect of virtue.'52Can we under-
stand the relevantnormativeterms from insidethe Rossianframework?
51If one wonderswhy moral judgementsshould not alwaysbe groundedin factualones, I sug-
gest that sometimes the former are non-inferentialresponses to facts, somewhat as perceptual
judgementsare. I discussthis possibilityin some detail in Audi (1998a).
2 These terms are found in Ross (1930),p. 21 (emphasesadded).

Here is the beginning of a solution. If the relevant normative terms

(for example, 'virtue') are understandable by appeal to the factual
grounds Ross cites in setting out the list of duties (or that might be
cited in clarifying a different well selected group of Rossian duties),
then each Rossianduty has a roughly factualbase, in the sense of a set
of factuallyspecifiablegrounds sufficientto yield it-to renderprima
facie obligatory some action of the kind it calls for. But this factual
specifiabilityof the grounds of Rossianduties may be in practiceunat-
tainable (even though it does not requireeither analyzabilityor even a
full indication of the grounds of the duty in question, as opposed to an
adequatebasis for ascribingit). I suspectthat Ross, steepedas he was in
Aristotle, took it that without begging any questions, we may assume
that both virtue concepts and such 'thick' moral concepts as those of
injury,lying, and malice,may be presupposedby mature,moral agents.
This may be a safe assumptionin a civilizedsociety.I am not suggesting
that Ross begged questions, but we should try to advanceintuitionism
in ways that he and later intuitionists did not. In this, as in other ways,
Kantianintuitionism providesadditionalresources.
I have alreadyindicatedhow a Kantianintuitionism helps us to deal
with the factual specification problem at the theoretical level. But in
practice the problem requires us to tie our understanding of benefi-
cence and other Rossianduties to factualmattersas closely as we can. I
want to addressan aspect of this problem that Ross does not. Suppose
he is right in taking virtue to be intrinsicallygood and thus an appro-
priate characteristic to honour and promote. Can a Rossian-or
Kantian-intuitionism, without begging moral questions, take pro-
motion of virtue as a ground for duty, notably for the apparentduties
of self-improvementand beneficence,each of which requirespromot-
ing virtue as essential to the good of the person(s) in question? How
does such a theory identifyvirtue without countenancinga moral cate-
gory independent of its basic concepts?A theory that cannot do this is
epistemicallyincomplete.Then, even if normativelycomplete,it cannot
do all one would like it to do in takingus from factsto duties.
Here is an approachopen to any plausibleintuitionism. Supposethat
moral virtues are roughlythe sorts of traitsof characterone would have
if, guided by practicalwisdom, one internalizedthe Rossianprinciples
of duty understood apart from their appeal to moral virtue and in the
light of the categoricalimperative.Roughly,the idea is that the duty to
make people better in respect of virtue might be interpretedas requir-
ing us to help them better internalizethe Rossian principles, taken in
the context of the injunctionto treatpeople as ends and nevermerelyas

632 RobertAudi
means, and with this proviso: any principle that essentially appeals to
virtue is understood with that appeal replaced by appropriate reference
to all the other principles so far as they are intelligible without invoking
that concept. By this strategy, one could identify at least many of the
important moral virtues, even if it does not provide a route to an analy-
sis of virtue concepts, and even if, as suggested above, some aspects of
virtue concepts are not capturable in terms of rules. If we habitually
observe-in the way appropriate to practical wisdom-Ross's princi-
ples of fidelity, justice, non-injury, reparation, gratitude, and benefi-
cence, each principle being understood in relation to all the others and
to the ideal of the material and intellectual improvement of persons as
ends as and as beings never to be treated merely as means, we are likely
to have a good number of important virtues. If, in addition, we can
bring to bear an account of the categorical imperative as systematizing
Rossian duties, then any factual considerations drawn from that
account may also be used to give a factual specification of the grounds
of those duties.
If we view virtues in this light, then some of the points about how
Kantian intuitionism can accommodate the moral importance of vir-
tues become clearer. First, a virtue can be constituted by internalization
of one or more basic moral principles together with an appropriate sec-
ond-order understanding-such as Kantian intuitionism would
provide of how to deal with conflicting moral considerations. This is
not a formula for the genesis of a virtue, nor need the second-order
understanding be articulate. Take the duty of beneficence. Above all,
one may have a suitably deep, long-standing desire to promote the
good of others, an understanding of how and when to do it, a tendency
to do it on that basis, and a sense of what duties override this goal. Sec-
ond, none of these points requires that in the context of acting benefi-
cently one must rehearse a moral principle. Internalized principles can
guide us without our calling them to mind. This even applies to using
the categorical imperative framework: I can act under the ideas of uni-
versalizability and of means and ends without reciting, or even being
readily able to formulate, the imperative that guides me. Thus, a Kan-
tian intuitionism does not require agents to be self-conscious rule-fol-
lowers. Internalized rules may yield moral conduct much as virtues do,
and both can allow spontaneous moral conduct. Neither here nor else-
where does the theory imply an implausible moral psychology.
There is much more one could say about the shape and implications
of Kantian intuitionism. But if we take this theory as a basis for further
reflection on moral obligation, then at least this much should be clear.

The categoricalimperative,conceivedin the light of respectfor persons

as a guiding moral standard, and with its intrinsic end formulation
taken as primary,can systematizeRossianprinciplesof duty,which can
be brought to bear more directlythan that imperativeitself in formu-
lating, understanding,and applying still more specific principles. The
entire frameworkmay be espousedwith an eye towardinterpretingele-
ments at any level in it in the light of their interactionswith elements at
any other level. The frameworkhas a good claim to normative com-
pleteness,and it may be arguedto approachepistemic completenessas
nearlyas one should demand in a moral theory.The use of it that I sug-
gest we make in understandingduties of beneficenceand self-improve-
ment can be clarified considerably by employing its own resources.
These include conceivingrespectfor personsas a moralvirtue. The the-
ory does not depend on virtue ethics, but it does demandvirtue in eth-
ics. Virtue is important in ethical theory and indispensable in moral
practice.
Kantian intuitionism does not claim to supply a monolithic or
reductiveaccount of moral obligation, nor a 'generalcharacter'in vir-
tue of which all morallyright acts are right;but the theory may perhaps
be seen as construing the property-if we may call it that- of befit-
ting the dignity of persons as belongingto morallyobligatoryactions in
general.This idea needs furtherclarification,but can receivesignificant
elucidation from the Rossian duties conceived as I have suggested.
Some of the grounds of the property,such as the capacityfor rational
thought and for autonomous conduct, are directlyconnected with the
grounds of the Rossian duties. Our understandingof those duties, in
turn, can be clarifiedby connecting them with the categoricalimpera-
tive. They can also be clarified from below: by framing more specific
practicalprinciples that addresscommon problems in the professions
or ordinaryaffairs.Kantand other philosopherswho have done moral
philosophyin the grandstyle have had too little faith in intuitivesingu-
lar moral judgement; Ross and other intuitionists have had too little
faith in comprehensivemoral theory. Kantianintuitionism is designed
to accommodateboth approaches.
There is still no bypassingpracticalwisdom, either in theoreticalor
in practical ethics. Ross was right about that, and rule ethics in the
grand top-down style epitomized in Kant does not easily do justice do
it. But practicalwisdom is independentlyessentialfor any approachin
ethics. This one has, among other merits I hope, the advantageof coun-
tenancing many differentinteractinglevels of reflectionfrom which to
understandmoral concepts and guide conduct in everydaylife.53

634 RobertAudi
Departmentof Philosophy ROBERT AUDI

University of Nebraska
Lincoln
Nebraska
68588-0321
USA
raudil@unl.edu
References
Audi, Robert 1989: Practical Reasoning. London: Routledge.
- 1994: 'Acting from Virtue'. Mind, 104, pp. 449-71.
- 1997:Moral Knowledge and Ethical Character.Oxford: Oxford Uni-
versity Press.
--1998a: 'The Axiology of Moral Experience'. Journal of Ethics, 2,
PP. 355-75.
1-998b: 'Moderate Intuitionism and the Epistemology of Moral
Judgment'. Ethical Theory and Moral Practice, i, pp. 14-34.
-1999: 'Self-Evidence'. Philosophical Perspectives,13, pp. 205-28.
-2001: The Architectureof Reason. Oxford: Oxford University Press.
Bok, Sissela, (ed.) 1998: Practical Ethics. Henry Sidgwick. Oxford:
Oxford University Press.
Brink, David O. 1994: 'Moral Conflict and its Structure'. Philosophical
Review, 103, pp. 215-47.
Dancy, Jonathan 1993:Moral Reasons. Oxford: Basil Blackwell.
Donagan, Alan 1993: 'Common Morality and Kant's Enlightenment
Project'. In Outka and Reeder (eds) 1993.
Gewirth, Alan 1975: Reason and Morality. Chicago: University of Chi-
cago Press.
Hooker, Brad 1996: 'Ross-style Pluralism versus Rule-consequentialism',
Mind, 105, pp. 531-52.
- 2000: Ideal Code, Real World. Oxford: Oxford University Press.
53Earlierversions of this paper were given at the AmericanPhilosophicalAssociation'sPacific

Division meetings in 1999,where GeoffreySayre-McCordpresenteda helpful commentary,at the
1999annual meeting of the Associationfor Practicaland ProfessionalEthics,and to the JowettSo-
ciety in Oxford,as well as at DartmouthCollege,San JoseStateUniversity,SantaClaraUniversity,
the Universityof Missouri,Columbia,the Universityof Texas,Austin,and WestPoint, and partsof
it derive from a section of ch.12of Audi 1997.For helpful comments I am gratefulto many mem-
bers of these audiencesand to HilaryBok, Roger Crisp,RobertFogelin, BernardGert, Amy Gut-
mann, Robert Kane, ChristopherKulp, Michael Meyer,JamesMurphy,Louis Pojman, William
Prior, ElizabethRadcliffe,Russ Shafer-Landau,David Sosa, BradWilburn, and, especially,John
Deigh, DerekParfit,WalterSinnott-Armstrong,and readersfor the Journal.

Kant, Immanuel 1949: The Critique of Practical Reason and Other Writ-
ings in Moral Philosophy. Trans. Lewis White Beck. Chicago:
University of Chicago Press.
- 1961: Groundwork of the Metaphysics of Morals. Trans. H. J. Paton.
London: Hutchinson and Co.
--1964: The Doctrine of Virtue. Trans. Mary Gregor. New York:
Harper and Row.
Korsgaard, Christine 1996: Sources of Normativity. Cambridge: Cam-
bridge University Press.
LaFollette, Hugh (ed.) 2000: The Blackwell Guide to Ethical Theory.
Oxford: Blackwell.
McNaughton, David 2000: 'Intuitionism', in LaFollette 2000.
Mill, J. S. 1959: Utilitarianism. Oscar Piest (ed.). Indianapolis: Bobbs-
Merrill.
Moore, G. E. 1903: Principia Ethica. Thomas Baldwin (ed.) Cambridge:
Cambridge University Press. Rprnt. 1993.
Outka, G. and Reeder, J. P. (eds) 1993:Prospectsfor a Common Morality.
Princeton: Princeton University Press.
Prichard, H. A. 1949: Moral Obligation. Oxford: Oxford University
Press.
Rawls, John 1971:A Theory ofJustice. Cambridge, MA: Harvard Univer-
sity Press.
Ross, W. D. 1930: The Right and the Good. Oxford: Oxford University
Press.
- 1939:Foundations of Ethics. Oxford: Oxford University Press.
Scanlon, T. M. 1998: What We Owe to Each Other. Cambridge, MA:
Harvard University Press.
Sidgwick, Henry 1907: The Methods of Ethics. Chicago: University of
Chicago Press. Rprnt. 1962
- 1909: Practical Ethics. London: Swan Sonnenschein.
Sinnott-Armstrong, Walter 1992: 'Intuitionism'. The Encyclopedia of
Ethics. New York and London: Garland Publishing Co., pp. 638-
630.
Stratton-Lake, Philip, 'Can Hooker's Rule-Consequentialist Principle
Justify Ross's Prima Facie Duties?' Mind, 106, pp. 751-58.
Swanton, Christine 1987: 'The Rationality of Ethical Intuitionism'. Aus-
tralasian Journal of Philosophy, 65, pp. 172-81.


Audi - A Kantian Intuitionism PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Audi - A Kantian Intuitionism PDF

Uploaded by

Copyright:

Available Formats

A Kantian Intuitionism

Kantfamouslysaid that one could not do moralitya worse disservicethan to derive

Mind,Vol. 110 . 439 . July2001 ? Oxford University Press 2001

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

points' in ethics, as Mill calledthem, that are so salient in practicaleth-

1. The possibility of systematizing Rossian principles

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

priate to mediating between 'deeper' axioms and other principles. In a

4 Ibid., pp. 29-30. Cf. H. A. Prichard's

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

Another pertinent distinction is between proving and evidencing,in

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

the 'ideal utilitarianism' of Moore,'1 as attempting to state one. He

2. Normative completeness, epistemic completeness, and con-

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

kind of adequacycondition for a comprehensivemoral theory; it ena-

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

knowledge both of what we should do and of why we should do it. A

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

A theory exhibitingfirst-ordernormativecompleteness,even if fully

3. A Kantian approachto conflicts of duty

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

clearly if we first consider the kind of deliberation appropriate to decid-

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

formulation is highly appropriateto testingacts for permissibility,but

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

implying that it is at least a candidate for self-evidence. Whether Ros-

4. The beneficence problem

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

Ross heightened the beneficence problem by describing the duty of

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

The personal-relationship reading of the duty to treat others as ends

5. Kantian intuitionism as providing a plausible grounding of

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

this context,34rectification of injustice one discovers, as opposed to

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

the principle of utility, as is suggestedby Sidgwick'sappealto his utili-

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

only principle-expressing in very general terms what sort of action

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

6. Kantian intuitionism as a plausible development of Kantian

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

dignity. The principles are constraints on the use of the imperative:

41 The second and third

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

reflectingon the categoricalimperativein the light of wide experience

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

43McNaughton(2000) might havebeen thinking of such passagesin saying,'I am more skepti-

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

casesof treatingsomeone merelyas a means. These points interconnect

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

justification for accepting the principles of duty it systematizes; and our

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

Adequately interpreted, then, the categorical imperative and the set

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

7. Between the middle axioms and moral decision: grounds of

This content downloaded from 128.235.251.160 on Thu, 15 Jan 2015 00:11:57 AM

tionism, by contrast, provides for the searchfor reflectiveequilibrium

50Sidgwickheld, for example,that 'Utilitarianism... must be acceptedas overrulingIntuition-