A revisionary Kantian view of moral dilemmas Michael Cholbi

A revisionary Kantian view of moral dilemmas
Michael Cholbi [[email protected]]
(November 4: appx 10,588 words)
A working single mother whose birth control fails must decide whether to abort her
recently conceived fetus for fear that she will have to leave the workforce otherwise; an
emergency room physician must determine in what order to perform triage on victims of a
natural disaster; a national leader must decide whether to attack a rogue nation’s recently
discovered nuclear weapons factories, even if this entails harming civilians; a teacher is torn
between upholding high academic standards and discouraging students who are at risk of
dropping out of school because of low grades; world leaders attempting to fashion a response to
global climate change consider how to balance the interests of present and future generations; a
sadistic concentration camp physician forces a mother to choose which of her two children will
be gassed to death.
Most every moral theorist concedes that these situations represent at least apparent moral
dilemmas. The more provocative question is whether there are genuine moral dilemmas: An
individual is in a genuine, rather than merely apparent, moral dilemma when she is morally
obligated to perform act A, and morally obligated to perform act B, but the contingent
circumstances of the world make it impossible to perform both A and B. Such a dilemma is
genuine because it is not the individual’s ignorance of any moral (or factual) consideration that
leaves her uncertain about what to do.1 It is rather that the moral universe itself is
Standard characterizations of genuine moral dilemmas found, e.g., in Christopher Gowans, Moral Dilemmas (New
York: Oxford UP, 1987), p. 3, or Terrance McConnell, “Moral dilemmas,” Stanford Encyclopedia of Philosophy
(http://plato.stanford.edu/entries/moral-dilemmas/, accessed 8/11/2009), tend not to index dilemmas to the agent’s
metaphysically constituted such that, despite an individual’s knowing what she is morally
obligated to do, it is impossible for her to do it. She is seemingly fated to moral failure.
In debates about the existence of genuine moral dilemmas, Kant is typically represented
not simply as denying the possibility of such dilemmas,2 but as the principal historical
spokesperson for a “rationalist” tradition in moral philosophy for whom genuine moral dilemmas
are anathema. As Christopher Gowans has characterized it, this rationalist tradition regards moral
practice as a species of human rationality, where reason is seen “as requiring system and order,
as necessitating commensurability and hierarchy, [and] as insisting on the importance of
generality and abstraction.”3 Because rationalists tend “regard whatever moral conflict might
appear in moral practice as mere appearance, as a betrayal rather than a manifestation of reason,
and as something that reason properly understood would reveal as such,”4 the rationalist tradition
is skeptical of genuine moral dilemmas. In contrast, what Gowans calls the “experimentalist”
tradition aims to “understand moral practice primarily from the standpoint of the moral
experience of persons,” giving priority to “observation and reflection on what it is like for a
person embedded in a particular social context to live a life constituted by values and
commitments, to encounter circumstances of perplexity and choice, to deliberate and determine a
epistemic or moral competence, knowledge, etc. Hence, it is possible for an agent both to be in a genuine dilemma
without knowing she is and possible for an agent to wrongly believe she is in a genuine moral dilemma.
As McConnell expresses it elsewhere, (“Moral residue and dilemmas,” in H.E. Mason (ed.), Moral Dilemmas and
Moral Theory (Oxford: Oxford U. Press, 1996)) “genuine moral dilemmas are ontological, not merely epistemic; the
truth of the conflicting ought-statements is independent of the agent’s beliefs.” (p. 36)
Gowans, Moral Dilemmas, pp. 6-7; Ruth Barcan Marcus, “More about moral dilemmas,” in Mason (ed.), p. 24;
Mary Mothersill, “The moral dilemmas debate,” in Mason (ed.), p. 69; Norman O. Dahl, “Morality, moral
dilemmas, and moral requirements,” in Mason (ed.), p. 90n8; and Thomas E. Hill, “Moral dilemmas, gaps, and
residues: A Kantian perspective,” in Mason (ed.), pp. 173-175.
Gowans, “Moral theory, moral dilemmas, and moral responsibilities,” in Mason (ed.), p. 200.
Gowans, “Moral theory, moral dilemmas, and moral responsibilities,” p. 202.
course of response, and to carry out this decision and live with its consequences.” Since moral
dilemmas do appear to arise within ordinary moral practice, the experimentalist tradition tends to
support the existence of genuine dilemmas.
My aim here is to defend a Kantian view concerning moral dilemmas that, while broadly
rationalistic, incorporates the experimentalists’ aim of accounting for ordinary moral experience.
At the level of theory, the revisionary Kantian view I shall defend is agnostic about whether
genuine moral dilemmas exist. However, at the level of practice, this view takes the denial of
genuine moral dilemmas as a regulative ideal that both guides and provides a goal for the
operation of practical reason in the moral realm. Moral inquiry and deliberation thus operate on
the implicit assumption that there are no genuine moral dilemmas, much in the fashion that
scientific inquiry and deliberation operate on the implicit assumption that our scientific
knowledge can be unified. Being agnostic about whether there are genuine moral dilemmas, this
revisionary Kantian view thus takes the denial of genuine moral dilemmas (or, correlatively,
their resolvability) as a methodological, rather than a metaphysical, hypothesis. Aside from its
interest as an interpretation of Kant, this view is superior to its metaphysical rivals — those that
either affirm or deny genuine moral dilemmas — in one crucial respect. Much of the recent
debate between these rival metaphysical views of moral dilemmas has focused on whether
certain features of our experience, most notably the “moral residue” identified by Bernard
Williams, show that there are genuine moral dilemmas. Proponents of genuine dilemmas claim
these features point to the existence of genuine dilemmas, whereas opponents of dilemmas reject
this inference. I shall argue that, by underscoring how apparently irresolvable moral dilemmas
represent failures to realize the aforementioned regulative ideal of practical reason, my
revisionary Kantian view more fully and more parsimoniously accounts not only for moral
residue but for other crucial phenomenological dimensions of our experience of moral dilemmas.
The revisionary Kantian view I defend is thus both historically significant and philosophically
attractive. Indeed, the plausibility of the view illustrates that the experimentalists’ commitment to
honoring ordinary moral practice, especially in “circumstances of complexity and choice,” is not
antithetical to the rationalists’ commitment to the elimination of moral conflict in the pursuit of
rational system and order. In fact, the latter is presupposed by the best account of the former.
My discussion unfolds as follows. Parts 1 and 2 argue that the passages on which the
standard interpretation of Kant on moral dilemmas rests have been read too hastily and that
Kant’s casuistical remarks about apparent moral dilemmas, as well as his views regarding the
general aspirations of reason, permit a methodological reading wherein the denial of moral
dilemmas is a regulative ideal instead of a metaphysical claim. On my reading, Kant sides
neither with those who accept, nor with those who deny, the existence of genuine moral
dilemmas. Instead, Kant’s apparent denial of moral dilemmas is an a priori principle of practical
reason we both use to guide our moral deliberation and as an ideal toward which such
deliberation strives. However, just as there may be as yet unnoticed evidence against even our
best scientific theories, evidence that confounds our effort to unify our scientific knowledge, so
too can never be certain that we have fully realized the regulative ideal according to which there
are no genuine moral dilemmas. For not only are there possible dilemmas we have not yet
encountered in our moral experience, we may also encounter apparent dilemmas that do not
appear resolvable despite our rational commitment to their resolution and despite our exercising
our deliberative capacities with full competence. This apparent irresolvability does not
demonstrate that there are genuine dilemmas, however, nor does it undermine the resolution of
dilemmas as a regulative ideal of practical reason.
After answering (in part 3) worries about the relation of theoretical and practical reason
and about the apparent incompatibility of this methodological view of moral dilemmas with core
elements of Kant’s moral philosophy, parts 4-6 show how this revisionary Kantian view
fruitfully explains three main features of the phenomenology of moral dilemmas. First, it can
explain the rationality of the self-reproach that agents often experience subsequent to acting
within apparent moral dilemmas. Second, it explains our attitudinal ambivalence concerning
which negative self-appraisals are appropriate in the wake of a moral dilemma. Lastly, it explains
an important asymmetry between our first personal and third personal standpoints on moral
This revisionary Kantian view, wherein the denial of moral dilemmas functions as a
regulative ideal of practical reason, is thus inviting both as an interpretation of Kant and as a
philosophical stance. If correct, this view permits us to retain many of our beliefs and attitudes
concerning moral dilemmas without hitching them to any metaphysical stance about whether
genuine moral dilemmas are real or not.
Before proceeding, a final note concerning the Kantian pedigree of my view: The
revisionary Kantian view of moral dilemmas I defend here rests on Kantian claims concerning
the goals and commitments of practical reason as it is deployed in moral deliberation and
inquiry. Hence, I take pains to show that this view coheres with other substantive commitments
of Kantian morality (the standing of the Categorical Imperative as the supreme principle of
morality, etc.). However, I deploy a Kantian vernacular in part for rhetorical convenience.
Indeed, I believe that this view of moral dilemmas should be congenial to moral theorists with
other substantive philosophical leanings (consequentialists, virtue theorists, etc.). For, as best as I
can surmise, the view does not hinge on whether the substance of Kant’s moral philosophy —
the Categorical Imperative, humanity as end in itself, etc. — is correct. Different moral theories
will yield different conclusions concerning which situations represent apparent or genuine moral
dilemmas, and why. The Kantian pedigree of my view should not, I believe, raises suspicions
among moral theorists whose allegiances lie elsewhere.
1. Dilemmas in Kant’s Doctrine of Virtue
Readers of Kant’s Doctrine of Virtue might be surprised to learn that Kant is so commonly
interpreted as denying moral dilemmas. For there Kant offers5 series of casuistical questions in
conjunction with each main type of moral duty (whether it is morally permissible to end one’s
life to save one’s country, whether merely polite lies are wrong, etc). But strikingly, Kant does
not answer these questions. Now, if Kant truly thought there are no genuine dilemmas, then the
presence of these unanswered casuistical questions is a puzzle. Kant was presumably highly
competent at applying his own moral theory, so if these casuistical dilemmas are resolvable, we
would expect Kant to be able to resolve them in order to illustrate the power of his moral theory.
That these casuistical questions are merely mentioned without even an attempt at their resolution
is surprising if we assume that Kant believed there are no genuine moral dilemmas.
Still, that Kant leaves these casuisical questions unresolved is at most indirect evidence for
Kant’s views about moral dilemmas. In addition, advocates of the standard interpretation of
Kant’s view on the matter may well reply that Kant says nothing in connection with these
casuistical questions to suggest that these questions are not rationally resolvable even in
principle, i.e., nothing here indicates that Kant supposed these are genuine moral dilemmas
rather than simply hard cases for his moral theory to tackle.
Starting at 6:423.
The standard interpretation would be more problematic, however, if the passages central to
this interpretation can be plausibly read so as not to support that interpretation. Such is the case, I
contend, with the most widely cited passage given in favor of the standard interpretation, here
quoted in its entirety:6
A conflict of duties would be a relation between them in which one of them would cancel the other (wholly
or in part). — But since duty and obligation are concepts that express the objective practical necessity of
certain actions and two rules opposed to each other cannot be necessary at the same time, if it is a duty to
act in accordance with one rule, to act in accordance with the opposite rule is not a duty but even contrary
to duty; so a collision of duties and obligations is inconceivable. However, a subject may have, in a rule he
prescribes to himself, two grounds of obligation, one or the other of which is not sufficient to put him under
obligation, so that one of them is not a duty. — When two such grounds conflict with each other, practical
philosophy says, not that the stronger obligation takes precedence but that the stronger ground of obligation
Kant’s words certainly seem to deny the existence of moral dilemmas.7 Agents can
certainly be subject to competing “grounds of obligation,” Kant acknowledges. That is, an agent
may, even when subject to a “rule he prescribes to himself” such as the self-legislated
Categorical Imperative, recognize two moral “obligations” or considerations each of which, in
isolation, would be sufficient to impose upon him a duty. Yet in circumstances where both
grounds of obligation are present (and in accordance with the notion of a moral dilemma, the
agent cannot act on both such grounds), it is “inconceivable” in light of the “objective practical
necessity” expressed by the concept of duty that the agent could have a duty stemming from each
Metaphysics of Morals 6:224. I have excised Kant’s Latin renderings of certain phrases.
Kant is best interpreted here as concerned only with what are perfect duties in his taxonomy. Imperfect duties
prescribe only general ends and cannot, in a strict sense, come into conflict with one another or with perfect duties.
Perfect duties prescribe specific actions and therefore can generate dilemmas. Alan Donagan, “Consistency in
rationalist moral systems,” Journal of Philosophy 81 (1984), p. 294, notes these relationships as well.
such ground of obligation.8 It must be the case, Kant concludes, that the agent’s duty stems from
the “stronger ground of obligation.” Apparently then, there cannot be a circumstance in which an
agent both has a duty to perform, and to abstain from performing, one and the same action. In
cases of competing grounds of obligation, one such ground must not be sufficient to constitute
the agent’s duty. It would appear, then, that Kant is deploying modus tollens against the claim
that there are genuine moral dilemmas: If the imperatives of duty were to yield conflicting
obligations, then morality’s imperatives would lack “objective practical necessity.” But since
those imperatives are imperatives of duty, they must incorporate “objective practical necessity”
and so cannot yield conflicting obligations.
All the same, to attribute to Kant the denial of genuine moral dilemmas on the basis of
this passage is too quick. In general, philosophers ought to interpret passages from the work of
their historical predecessors so as to ensure consistency with the relevant texts, coherence with
those predecessors’ overall philosophical stances, and, when necessary, to opt for the weaker or
less ambitious interpretation of a passage if this interpretation has clear philosophical advantages
over stronger or more ambitious rivals. By these criteria, Kant is not best read as denying the
possibility of genuine moral dilemmas.
2. The denial of dilemmas as a regulative ideal
Kant is clearly concerned in this passage that the existence of genuine moral dilemmas
would somehow threaten the intelligibility of morality’s demands. In order to pinpoint the
Kant offers the same argument in the other main passage seeming to address dilemmas. In the Lectures on Ethics
(27:261), Kant states that “no two obligations can clash, because what is rendered morally necessary by one cannot
be made otherwise by another. …If the one duty is an obligation, the other cannot properly be termed so.”
threat, we can consider how the existence of genuine dilemmas might undermine the Categorical
Imperative, the principle which, in Kant’s ethics, gives expression to morality’s demands.
The Categorical Imperative plays two roles in Kantian morality. It is on the one hand a
theoretical principle, identifying those attributes of actions which render their performance
morally permissible or morally impermissible. Admittedly, how these attributes are identified is
a complex matter in Kantian ethics. Maxims — those principles that agents use to justify their
actions — are justified in a procedural manner under the Formula of Universal Law, but in a
more substantive manner under the Formula of the End in Itself. The Categorical Imperative is
morality’s supreme principle, in a theoretical sense, because no more fundamental principle
logically determines our moral obligations. On the other hand, the Categorical Imperative is also
morality’s supreme practical principle, directing human agents to what they must do, regardless
of their contingent inclinations. Even though we rarely deploy such general principles explicitly
in our moral deliberations, we nevertheless can, according to Kant, deploy the Categorical
Imperative to test our maxims and thereby judge what we morally ought to do.9
To interpret Kant as rejecting the possibility of genuine moral dilemmas in this passage is
to suppose that Kant’s concern here is with the theoretical standing of the Categorical
Imperative. In other words, if the Categorical Imperative were to yield genuine moral dilemmas,
its theoretical supremacy would be threatened because the Categorical Imperative itself would
yield contradictory moral demands. But I doubt the theoretical standing of the Categorical
Imperative is what worries Kant here. Granted, Kant may have naively believed that only a
monistic moral theory, i.e., one that offers a single supreme principle, can thereby avoid
My way of describing the Categorical Imperative’s two roles is indebted to Robert Stern, “Does ‘ought’ imply
‘can’? And did Kant think it does?” Utilitas 16 (2004), pp. 56-59.
generating genuine dilemmas.10 Yet philosophers now recognize that although a logically
inconsistent moral theory can yield genuine dilemmas, logical consistency is not sufficient to
forestall their existence either. The moral rules ‘defend your country against aggressors’ and
‘care for one’s parents in their old age’ are perfectly consistent but quite capable of generating a
moral dilemma, as Sartre’s famous example illustrated.11 Inconsistency can be generated by a
theory itself, but it can also be traced to the facts the theory takes as morally salient. 12 Since there
is simply no a priori guarantee that the world will cooperate with our best lain moral schemes,
the more cunning source of inconsistency, i.e., of genuine moral dilemmas, is the world itself,
which so often seems factually arranged to thwart even the most fastidiously consistent moral
theory. That the Categorical Imperative is logically consistent is of course a controversial
assumption. Doubts may be raised about, for instance, whether its various formulations yield the
same verdicts or whether contradictory imperatives result when we attempt to use it to determine
what we morally ought to do. But the important point for our purposes is that Kant certainly
supposed that his supreme principle of morality was logically consistent in these ways, and if so,
then the worries he expressed about the threat genuine moral dilemmas would pose to the
intelligibility of morality’s demands are not worries that the Categorical Imperative fails in its
theoretical role at identifying the moral attributes of actions.
If the theoretical standing of the Categorical Imperative is not what is threatened by genuine
moral dilemmas, then Kant’s anxiety must stem from the fear that the Categorical Imperative’s
practical supremacy is threatened by genuine moral dilemmas. And there are subtle hints in this
See Marcus, “More about moral dilemmas,” p. 25. Moral dilemmas need not result from the presence of distinct
rules at all. An agent who promises two distinct parties that she will be at different locations at the same hour
appears to face a dilemma rooted in the very same rule to honor one’s promises.
See Ruth Barcan Marcus, “Moral dilemmas and consistency,” Journal of Philosophy 77 (1980): 121-136.
Mark Timmons, Moral Theory: An Introduction (Lanham, Md.: Rowman & Littlefield, 2002), p.13.
passage that this is Kant’s intent. Note first that Kant does not literally deny the existence of
genuine dilemmas in this passage. Rather, the existence of moral dilemmas is, Kant says,
“inconceivable” in light of the objective practical necessity expressed in the Categorical
Imperative, morality’s supreme principle. The existence of moral dilemmas would in effect deny
the a priori practical authority of the Categorical Imperative, an authority which, Kant argues in
the Groundwork, is evident to common moral understanding. For such a principle to yield two
incompatible ‘musts’ need not cast into doubt its theoretical supremacy, but it would stymie our
ability to deploy it in identifying what we must do. The “inconceivability” of a “collision of
duties” is presumably not the inconceivability of a principle that literally directs us to perform
two metaphysically incompatible actions. For the principle in question, the Categorical
Imperative, is definitely conceivable. What is inconceivable is that this principle could, in
conjunction with the factual circumstances in which we attempt to deploy the principle, could
yield metaphysically conflicting imperatives. Throughout the passage, Kant’s language is
epistemological or methodological, not metaphysical. For it is the possibility of “certain actions”
colliding, actions each of which bear the imprint of objective practical necessity, that piques
Kant’s worries. The resolution of moral dilemmas is an requirement emanating from our
practical reason. This is why Kant says that “practical philosophy” insists that a stronger ground
of obligation must prevail in any apparent moral dilemma.
I propose, then, that the thesis of this passage is not that acknowledging the existence of
genuine moral dilemmas would be a metaphysical error, a false belief about the nature of moral
reality. Rather, accepting the existence of genuine moral dilemmas would be a practical error, in
effect denying morality its distinctive normative capacity to guide action. But what is the source
or rationale for this practical Kant is clearly indicating that we should accept that there are no
dilemmas — that their denial should be treated as true for practical purposes. But why should
this be so?
We see in the aforementioned passage glimmers of Kant’s characteristic transcendental
argumentation: We possess uncontroversial knowledge of a certain kind (of the Categorical
Imperative as morality’s supreme principle, in this case), knowledge of which is explicable only
if some other claim (that there are no genuine dilemmas) is acknowledged as a necessary
condition of the first. But this cannot be quite right, for again, Kant does not seem to be arguing
that the Categorical Imperative would not be morality’s supreme principle if there were genuine
dilemmas. The problem is instead that we could not coherently treat the Categorical Imperative
as the principle that ought to guide our moral deliberations if we assume there are genuine
dilemmas. Regardless of whether the Categorical Imperative is morality’s supreme principle,
acting as if there are genuine moral dilemmas would make deliberating as if the Categorical
Imperative is morality’s supreme principle unintelligible.
Perhaps Kant’s denial of moral dilemmas is a practical conclusion akin to his infamous
“postulates of practical reason,” conclusions supported not by theoretical reason but by demands
of “moral faith.” But this too cannot be quite right. For we are to accept the postulates of God’s
existence and the immortality of the soul because these ensure the congruity of happiness and
moral virtue that Kant believed constituted the highest good. The acceptance of such a congruity
provides us hope in the face of the challenges presented by the evident divergence between
happiness and moral virtue in the actual, empirical world. But Kant’s denial of moral dilemmas
does not sustain our commitment to morality in the same way. The denial of moral dilemmas
assures that morality’s own principle can intelligibly guide choice and action. The practical
necessity of this denial is internal to the moral point of view, whereas the practical postulates are
necessary to assure us of the harmony of the moral point of view with something only
contingently related to it, happiness.
Kant’s denial of moral dilemmas is better understood as what Kant termed a “transcendental
idea,” a concept originating in reason, the faculty that seeks to unify our understanding under
principles.13 As such, transcendental ideas cannot be objectively deduced as the basis of the
principles of possible experience, as the transcendental categories can. Nor are there objects in
our experience corresponding to such ideas. Transcendental ideas can only be subjectively
deduced from the nature of reason itself, and in this case, the denial of moral dilemmas is
deduced from the nature of practical reason, and more specifically, from the nature of the
demands imposed by the Categorical Imperative as the supreme principle of practical reason.
However, the denial of moral dilemmas cannot be known as a matter of synthetic a priori
knowledge, nor can it be empirically verified. It is rather an assumption or aspiration that
practically rational creatures must bring to their moral deliberation and inquiry in order for their
understanding of morality as prescribing categorically to be intelligible.
Kant’s denial of moral dilemmas thus reflects reason’s interest in the systematic unity of its
knowledge. Reason guides our empirical inquiry, postulating as an ideal the increasing
subsumption of our scientific knowledge into fewer and fewer principles so that the necessary
connections among the items of our knowledge are apparent.14 In so doing, reason must both
assume and aspire to such a unity. Hence, the unity of science sets the philosophical agenda for
the conduct of science. The denial of moral dilemmas plays a similar role in the conduct of
moral deliberation. It is both an assumption and an aspiration of practical reason that sets the
agenda for moral deliberation. Reason, Kant reminds us, always seeks the unconditioned to
Critique of Pure Reason, A399/B355-56, A303/B358, A311/B368
Critique of Pure Reason (A646-651/B674-80)
unify what is conditioned and thus assumes that the unconditioned can be found. The Categorical
Imperative is, Kant argues, an unconditional practical demand, but it can only unify or
“complete” the various particular moral demands (morality’s innumerable categorical
imperatives) if its prescriptions are free of contradiction, i.e., if the non-existence of moral
dilemmas is assumed.
Kant’s denial of moral dilemmas thus reminds us that Kant believed that the proper use of
reason is always “regulative,” not “constitutive.” The principle that there are no genuine moral
dilemmas would be constitutive if it provided a real object with which it is assumed to
correspond. Yet the point of Kant’s Transcendental Dialectic is that attempts to provide such
objects for certain ideas of reason (God, freedom, etc.) inevitably result in contradiction and
illusion, thus dooming traditional metaphysics. So, too, presumably for the denial of moral
dilemmas in a metaphysical sense: Not only does such a thesis transcend reason’s proper
application, it cannot be empirically verified or disconfirmed either.
Yet reason must retain a regulative use such that the transcendental ideas orient inquiry and
deliberation. Our empirical inquiries would, Kant concludes, be little better than fumbling and
groping absent the aspiration for theoretical unity in the sciences. But we cannot in advance, or at
any given point, determine whether our scientific understanding has achieved this unity, for there
is no standpoint outside our best scientific practices from which we inspect them for this unity.
Similarly, the most we can hope for is that the denial of moral dilemmas may be true. But it
cannot be legitimately treated as a knowledge claim in a metaphysical sense. Furthermore, there
may be no standpoint outside the proper deployment of our moral powers from which we could
determine whether an apparent moral dilemma has been fully resolved. The denial of moral
dilemmas thus serves as a regulative, rather than constitutive, principle of reason: not a claim
known to be true but a condition for the rational application of our knowledge of morality as
issuing in categorical rational demands.
3. Two worries
One worry here stems from the seemingly ambiguous rational status of the denial of moral
dilemmas on my revised Kantian view. I have argued that the denial of moral dilemmas is a
regulative principle of practical reason whose authority rests on its being a consequence of our
recognition of the Categorical Imperative as practical reason’s supreme principle. But whether
there are genuine dilemmas appears to be a theoretical question, a question about whether moral
considerations supervene on non-moral ones to generate incompatible imperatives. Thus, it may
seem that the position I have attributed to Kant is orthogonal to recent philosophical debates
about moral dilemmas. To a large extent, this worry is on target, for the position I have outlined
is not a position about how the world is, morally speaking, but a position about the stance we
ought to take toward the world in our moral deliberation and inquiry. Yet for Kant, not all
metaphysical beliefs or attitudes are settled by theoretical considerations. Practical reason has
primacy, according to Kant, such that practical considerations should sometimes guide our
beliefs and actions even in the face of evidence against those beliefs or actions. Moreover, since
the Categorical Imperative is the supreme principle of practical reason, the interests of theoretical
reason may be subordinated to it. In the case of moral dilemmas, then, this primacy mandates
that the non-existence of genuine moral dilemmas be assumed, even absent a theoretical
demonstration of that thesis.
A second worry is that this view of moral dilemmas is far more damaging to Kant’s overall
project in ethics than I have acknowledged. In particular, this methodological view weakens
Kant’s claim that the Categorical Imperative is the supreme principle of morality. Allowing that
we ought to utilize the Categorical Imperative as a principle of deliberation through which we
aspire to resolve moral dilemmas, critics may contend that this principle’s supremacy is in doubt
if the existence of moral dilemmas is even possible. Yet it is not the case that the Categorical
Imperative’s supremacy is threatened by this view of dilemmas. The Categorical Imperative is
typically understood as articulating necessary and sufficient logical conditions for the moral
permissibility of acting on a maxim, and my position does not cast that into doubt. Nor does it
weaken this principle’s claim to having any of the features that Kant seemed to have in mind
when he sought morality’s supreme principle: It does not undermine its a priori necessity; its
application to all rational agents; its being such that we can abide it motivated by respect for the
moral law; its ability to derive a plausible set of moral duties; or its standing as fundamental (i.e.,
there is not a more basic principle than the Categorical Imperative).15 Admittedly this view of
moral dilemmas inserts what Thomas Hill has called “gaps” into Kantian theory, since it
concedes that there is a respect in which the Categorical Imperative is not always sufficient,
“even given all pertinent facts about a case,” to determine “for all acts whether they are
obligatory, forbidden, or neither.”16 For this regulative ideal may be unrealized when we
confront an apparent dilemma where the options end up in a tie after full moral deliberation. But
Kantianism’s inability to eliminate all such gaps does not show that the theory or its supreme
principle fail to be action-guiding. Not to be uniquely action-guiding is not to be non-actionguiding simpliciter, and nothing in this revisionary Kantian view necessitates our relinquishing
the Kantian thesis that the “primary function of moral judgments … is to express rational
These are among the conditions that Samuel Kerstein, Kant’s Search for the Supreme Principle of Morality
(Cambridge: Cambridge U. Press, 2002) argues Kant intended to capture in a supreme principle of morality.
Hill, “Moral dilemmas, gaps, and residues,” p. 179.
demands on our wills as deliberating agents, telling us which among our (perceived) options to
choose to take.”17
In sum then: In terms of Kant’s famous three central questions for reason — what can I
know, what ought I do, and what may I hope— the answers with respect to moral dilemmas are:
1. We can know a priori that reason assumes and aspires to the denial of genuine moral
dilemmas, but we cannot know whether that denial is in fact true. Nor can we come to
know with certainty that an apparent moral dilemma has been correctly resolved, for we
have no standpoint outside our practical reason from which to appraise such resolutions.
2. We ought to deliberate and inquire in moral matters on the assumption that there are no
genuine moral dilemmas – that every apparent dilemma is resolvable.
3. We may reasonably hope that the world is not fundamentally dilemmatic in character.
The revisionary Kantian position I have outlined has a number of modest, and thereby
attractive, features. It neither asserts nor denies the existence of genuine moral dilemmas, but
instead takes the resolution of moral dilemmas as a methodological ideal. However, advocates of
non-agnostic views of genuine moral dilemmas — those philosophers who defend or deny their
existence— may be happy to accommodate my methodological view. For its very modesty does
not impugn, and is in fact consistent with, either non-agnostic view. Hence, my revisionary
Kantian view may appear to complement non-agnostic views rather than being a philosophical
rival to them.
However, this would not be the case if there were independent reasons to favor the
revisionary Kantian position over its non-agnostic rivals. And if we can explain a body of
phenomena solely in terms of our epistemic or rational commitments rather than reverting to a
Hill, “Moral dilemmas, gaps, and residues,” p. 175.
controversial metaphysical hypothesis, we should do so. In the next three sections, I argue that
the revisionary Kantian position has the advantage that it can better explain several central
dimensions of our experience of moral dilemmas than can its non-agnostic rivals.
4. Explaining the phenomenology, part I: Moral residue and rational self-reproach
As one might expect, my revisionary Kantian position is skeptical of arguments, whether
empirical or a priori, intended to establish whether or not genuine moral dilemmas exist. Such
arguments can nevertheless appeal to premises that reflect central dimensions of our experience
of moral dilemmas. The best-known example of such an argument is Bernard Williams’
argument from ‘moral residue.’18 This argument attempts to demonstrate that in order for the
reactive attitudes we often have after we act in dilemmas to be rational, there must in fact be
genuine moral dilemmas. More specifically, when an agent acts in response to an apparent moral
dilemma, she is rightly subject to self-reproach in the form of various negative self-appraisals
(regret, guilt, remorse, etc.), regardless of which of the two acts she performed. But these selfappraisals are rational or justified only if the agent would have been equally rational or justified
had she acted to fulfill the other apparent obligation that bound her in the dilemma. Since these
self-appraisals are appropriate only when an agent acts wrongly, it follows that the agent could
not have failed to act wrongly and was therefore in a genuine moral dilemma. Put more formally:
1. Suppose S is in an apparent moral dilemma, compelling her either to do A or to do B.
2. Regardless of whether S does A or B, S will undergo rational self-reproach.
3. Self-reproach is rational only if S acts wrongly (i.e., S violated a moral obligation).
4. Thus, regardless of whether S does A or B, S will have acted wrongly.
“Ethical consistency,” Proceedings of the Aristotelian Society supp. 39 (1965): 103-24.
S’s situation is a genuine moral dilemma (since she cannot avoid acting wrongly).
Skeptics about moral dilemmas typically object to the moral residue argument by
rejecting premise 3 on the grounds that the rationality of these negative self-appraisals does not
entail that an agent acted wrongly. For one thing, much will depend on the exact negative selfappraisals in question. Some such self-appraisals (e.g., guilt) imply wrongdoing, but others
(regret) do not. Furthermore, it can be rational for agents to experience negative self-appraisals
that imply wrongdoing even when agents accept they did not act wrongly.19 Such self-reproach
could motivate agents to compensate (or seek the forgiveness of) those they injure as a result of
acting in a dilemma or motivate agents to fashion social institutions and circumstances that do
not tend to put individuals in dilemmas in the first place. Hence, considerations of social welfare
make self-reproach rational even when agents do not in fact violate any obligations in the course
of acting to resolve a dilemma.
I follow opponents of genuine dilemmas in concluding that these objections to the moral
residue argument show the moral residue argument to be unsound: premise 3 is not the only
credible explanation for the truth of premise 2, and in general, using our experience of moral
dilemmas to diagnose the existence of genuine dilemmas looks like a hasty leap from
phenomenology (or epistemology) to metaphysics. (And if my revisionary Kantian view is
correct, all such leaps will be illicit.) However, rather than resting satisfied with an apparent
refutation of the moral residue argument, I wish to shift focus to the common ground that unites
advocates of the moral residue argument and its critics, namely, its second premise. Both accept
See among other responses, Terrance McConnell, “Moral dilemmas and consistency in ethics,” Canadian Journal
of Philosophy 8 (1978): 269-87; Barcan Marcus, “Moral dilemmas and consistency”; Earl Conee, “Against moral
dilemmas,” Philosophical Review 91 (1982): 87-97; Philippa Foot, “Moral realism and moral dilemma,” Journal of
Philosophy 80 (1983): 379-98; and McConnell, “Moral residue and dilemmas,” p. 38.
that after acting in an apparent moral dilemma, an agent may be subject to rational self-reproach,
regardless of which act she performs. Even after doing what we believe we ought to have done
in resolving a moral dilemma, we rarely feel pleasure or a sense of emphatic selfcongratulation.20 Granting that the existence of genuine moral dilemmas is a problematic
explanation of this rational self-reproach, what view of moral dilemmas would explain this fact?
There are two specific features of premise 2 that require explanation. The first is that the
rationality of this self-reproach is intrinsic to it inasmuch as it is, or results from, retrospective
judgments of moral dilemmas themselves and how agents acted in those dilemmas. Skeptics
about moral dilemmas, as I mentioned above, attempt to explain how such rational self-reproach
is justified, regardless of which course of action a person takes in an apparent dilemma, by
claiming that making amends to those injured, trying to reform social institutions so as to reduce
the frequency of dilemmas, etc., are rational responses to the experience of dilemmas. A
physician who experiences negative self-appraisals after finding herself in too many moral
dilemmas concerning end of life care may then advocate for better palliative care so as to
minimize future dilemmas of this kind. Doubtless, responses such as making amends and the like
can be supported by moral reasons, and in this sense, the dilemmas have a causal role in the
history of these responses. But this is quite different from a rational response to dilemmas
themselves, and it is the rationality of self-reproach itself, not the rationality of how we act in the
wake of such self-reproach, that advocates of the moral residue argument seek to explain in terms
of the existence of genuine dilemmas. Advocates of the moral residue argument thus see the
Daniel Markovits, “The architecture of integrity” in D. Callcut (ed.) Reading Bernard Williams (London:
Routledge 2009), p. 114, makes this same observation in connection with how consequentialists analyze Bernard
Williams’ infamous ‘Jim and Pedro’ example — that “having killed one, Jim might congratulate himself on saving
nineteen lives.”
rationality of self-reproach as justified retrospectively, as involving the proper cognition or
appreciation of dilemmas and how best to resolve them. What makes guilt, remorse, etc.,
rational, they argue, is that they incorporate accurate cognitions both of the moral considerations
that constitute moral dilemmas and of how to act when facing moral dilemmas. In contrast,
skeptics about genuine dilemmas explain the rationality of self-reproach prospectively, in terms
of how agents respond, rationally-cum-morally, to having been in apparent moral dilemmas. But
this is the wrong kind of explanation (moral, rather than epistemic) of the rationality of this selfreproach and it targets the wrong explanandum (our moral responses to our judgments
concerning moral dilemmas, rather than the judgments themselves).
Skeptics about genuine dilemmas may reply that they can explain this rational self-reproach in
other ways, however. Regardless of how agents react to having been in dilemmas (whether they
try to avoid such dilemmas in the future, etc.), agents undergo rational self-reproach when they
believe they judged a dilemma wrongly. If, as skeptics believe, there are no genuine dilemmas,
then an agent who concludes that she opted for the wrong course of action ought to feel selfreproach. Suppose that agent S does B in an apparent dilemma, when in reality, she was
obligated to A, and S later comes to realize this fact. In such a case, rational self-reproach would
clearly be warranted. But this reply is inadequate, for in endorsing premise 2, skeptics about
dilemmas are conceding that this rational self-reproach does not depend on whether an agent
believes she erred in evaluating a dilemma. The self-reproach is supposed to be rational
irrespective of whether an agent believes she acted properly in resolving the dilemma. Thus,
even if believing one erred might be part of the explanation for rational self-reproach in some
cases where agents were compelled to act in a dilemma, it cannot be the full explanation.
Furthermore, this only deepens the puzzle about providing a comprehensive explanation of
rational self-reproach: If the self-reproach is rational in the sense that if involves a cognition or
judgment of the dilemma and how to resolve it, then it would appear that agents’ self-reproach
must rest on their judgments that they erred in evaluating or responding to the dilemma (or both).
But self-reproach does not appear to hinge on agents’ judgments about whether they erred in
evaluating or responding to a dilemma. Although the self-appraisals we experience after acting
in a dilemma are to some extent sensitive to our judgments about whether we engaged in
wrongdoing, there often are, as Patricia Greenspan has suggested, agents who feel guilt without
having a corresponding judgment of fault or wrongdoing.21 But the rationality of such selfreproach is precisely the fact in need of explanation, and for agents to suffer an “uneasy
conscience” even when they correctly judge both that their situation was not a genuine dilemma
and that they acted correctly in that situation is actually irrational.
We will return to this first feature in a moment, but the perplexity about rational selfreproach only deepens when we consider a second important feature of it: The reproach is
reproach of the self.22 But neither non-agnostic view of moral dilemmas illuminates how the
agent herself is an object of reproach after she acts to resolve a moral dilemma. There is
something insufficiently reflexive about explaining these self-appraisals by proposing that there
are moral dilemmas. For these are negative self-appraisals, and there need be no failure
(epistemic, moral, or otherwise) on the part of an agent if these negative attitudes reflect a
disjointed moral world populated by genuine moral dilemmas. That an agent is, in a sense,
trapped by a genuine moral dilemma is not a rational basis for negatively appraising the agent
(aside from dilemmas that are the agent’s own making). Indeed, an agent who concludes that she
Practical Guilt: Moral Dilemmas, Emotions, and Social Norms (Oxford: Oxford U. Press, 1995), p. 151.
Greenspan is one of the few philosophers to take much stock of the specifically self-directed character of
dilemmatic moral residue. See Practical Guilt: Moral Dilemmas, Emotions, and Social Norms, p. 135.
was in a genuine dilemma may in fact experience the opposite of such negative self-appraisals.
She may instead feel a sense of relief that there was, in an important sense, no way not to have
done wrong.
On the other hand, we have already observed that skeptics about moral dilemmas can argue
that reproach directed at the self is appropriate if an agent concludes she erred in evaluating or
responding to a dilemma. But even here it is not entirely clear that an agent who believes she
erred in evaluating or responding to a dilemma ought to feel self-reproach. For a failed effort to
resolve what is assumed to be in principle resolvable need not reflect negatively on an individual.
Suppose that arithmetic is such that all its theorems are provable. Nevertheless, there are
theorems whose proofs cannot be readily identified. The proof of Fermat’s last theorem eluded
generations of brilliant mathematicians, for example. Now imagine such a mathematician felt
self-reproach because she inferred that because of (a) the in principle provability of all theorems,
and (b) her own inability to prove Fermat’s theorem despite her using sound disciplinary
methodology, etc., that she must therefore have erred in some way she is unable to detect. Such
self-reproach would not be a rational response to her failed efforts. It would instead amount to
self-flagellation. Some problems in mathematics are sufficiently daunting that erring in
evaluating or responding to them does not impugn mathematicians who so err. After all, even
Andrew Wiles, the mathematician who ultimately proved Fermat’s last theorem, said it was to
some degree due to luck. So too for morality: There are some moral dilemmas, apparent or
otherwise, that are sufficiently daunting that erring in evaluating or responding to them does not
impugn agents who so err. If Sophie in Styron’s novel Sophie’s Choice chose the wrong child to
send to the gas chambers, would anyone hold this against her? If this is correct, however,
regardless of whether there are no genuine moral dilemmas, rational self-reproach does not hinge
on whether agents believe they succeeded in resolving a dilemma.
Thus, the rationality of self-reproach is therefore not contingent on whether genuine moral
dilemmas exist23 or even whether agents believe they deliberated correctly in their efforts to
resolve dilemmas. Thus, the rationality of this self-reproach cannot be located either in the
metaphysical facts about the existence of genuine moral dilemmas or in the properties of agents’
specific deliberative or epistemic performances. It seems as if this rational self-reproach is not
the product of substantive irrationality, since it does not depend on agents’ getting moral reality
right. But nor is it the product of procedural irrationality, since it does not necessarily depend on
how well agents deliberate or act in their efforts to get moral reality right.
Is the rationality of self-reproach just inexplicable then? No. The Kantian position I have
outlined, despite taking no stand on the metaphysics of moral dilemmas, nevertheless provides
an explanation of why such negative self-appraisals may be rationally warranted regardless of
whether genuine moral dilemmas exist and regardless of whether agents believe they have
successfully resolved a dilemma. As a regulative ideal, the denial of dilemmas is implicitly
accepted by all moral agents as a norm of reason. It is thus the agents’ own deliberative moral
ideal. Thus, when an agent finds herself in what appears to be an especially confounding
dilemma, it would not, I contend, be irrational for such an agent to later experience the negative
self-appraisals allegedly associated with the existence of moral dilemmas. Moral agents aspire to
have their conduct guided by the categoricity of moral demands, but hard cases can make it
unclear in which acts that categoricity resides. We are, Kant claimed, capable of the distinctly
moral emotion of reverence for the moral law. To worry that we have not identified what the
Walter Sinnott-Armstrong, Moral Dilemmas (Oxford: Basil Blackwell, 1988), pp. 46-47.
moral law asks of us thus sparks a rational anxiety that manifests itself as self-reproach. Moral
residue thus induces the sense that we have let morality let down. Yet Kant’s conception of
morality is that it emanates from our own autonomous, practically rational nature. Thus, to
worry that one may have acted immorally by choosing the wrong horn of an apparent moral
dilemma is to let oneself down. It is the moral law within and not the starry heavens above that
generates this self-reproach. A negative moral appraisal is thus implicitly a negative selfappraisal. Hence, the regulative ideal implicit in agents’ deliberation explains how agents are
subject to negative appraisals directed specifically at the self as a moral agent.
Furthermore, the regulative ideal pinpoints what makes this self-reproach rational. It is neither
that, because there are no genuine dilemmas, agents must have fallen short in the resolution of
apparent dilemmas despite their best deliberative and epistemic efforts. Nor is that agents must
have deliberated or reasoned badly. Instead, as dilemmas increase in their apparent intractability,
concern that the regulative ideal will not be realized increases in proportion. This concern is not
rooted in the firm belief either that genuine dilemmas do not exist or that agents have not on a
particular occasion actually resolved a dilemma. Agents’ self-reproach is rational because
especially daunting dilemmas put the realizability of this regulative ideal in doubt.
5. Explaining the phenomenology, part II: Attitudinal ambivalence
As noted above, the negative self-appraisals individuals may experience once they act within
an apparent moral dilemma are numerous. They include guilt, shame, regret, and remorse, as
well as attitudes or intentions that flow from these self-appraisals, such as desires to make
amends to injured parties and to seek their forgiveness and understanding. Furthermore, only
some of these carry the suggestion that an individual violated a moral obligation. A person
ought to feel guilt, for example, only on the condition that she violated an obligation, but if she
did not, an appraisal such as regret is more appropriate. Some of these (a desire to make amends
to injured parties, for instance) may be appropriate regardless of whether an obligation was
But taxonomizing these self-appraisals and attitudes in terms of their relationship to
moral wrongdoing is one matter. Deciphering which ones are appropriate to attribute to oneself
in a given context is another. For one thing, they do not necessarily have sufficiently distinct
phenomenological profiles to enable us to identify which of them we are in fact undergoing in a
given instance. Guilt and shame, on the one hand, and regret on the other, are all species of
‘feeling bad’ for what one has done. But the anguish associated with these is often so inchoate
that introspection cannot tell us which of these self-appraisals is most fitting in the wake of our
having acted in an apparent moral dilemma. Our emotional discernment is simply not that finetuned.24 This phenomenological coarseness may be inevitable. Perhaps it serves an important
developmental or evolutionary function. (The analogous phenomenon of survivor guilt, despite
not being a form of guilt, could be an evolutionary byproduct, an emotion that enhances
solidarity and social responsibility even when blame is not justified.) Yet when we act in a
situation that seems to us like a moral dilemma, we often do not have confidence after the fact
how we feel, or ought to feel, about our actions. Normative and descriptive self-interrogation
become intertwined here, as we simultaneously attempt to discern what we feel and evaluate the
fittingness of what we feel.
We can of course have recourse to the dilemma itself, again sifting through the facts that
gave rise to it, the moral considerations speaking in favor of each course of action, etc. But as I
McConnell, “Moral residue and dilemmas,” p. 39.
have argued, an agent thrown back on the dilemma may still find it troubling even in retrospect,
a fact which generates the moral anxiety that grounds these negative self-appraisals. In that case,
intractable ambivalence about the dilemma not only occurs, but is a warranted response. My
revisionary Kantian view has a ready explanation for the rationality of this ambivalence. It stems
from ambiguity about the dilemmas themselves. The regulative ideal that denies the existence of
genuine dilemmas, and to which we are committed just insofar as we are practically rational
agents, collides with situations in which dilemmas are difficult to resolve. Our agential
aspirations clash with moral reality, but with no obvious path indicating whether it is the
aspirations or moral reality that must give way.
In contrast, non-agnostic views of moral dilemmas — those claiming that there are or
there are not genuine dilemmas — must ultimately dismiss this ambivalence. For on either nonagnostic view, additional information or moral insight can, in every instance and in principle at
least, be mustered to answer the metaphysical question of whether an individual was in a genuine
dilemma. And once mustered, this information or insight directs the agent to the appropriate
appraisals or attitudes (guilt, say, if the dilemma stands, mere regret if it does not). The
resolution of the metaphysical uncertainty in turn resolves the phenomenological ambivalence.
But my revisionary Kantian view honors the rationality of this ambivalence and does not seek to
argue it away prematurely.
6. Explaining the phenomenology, part III: Self-other asymmetry
This revisionary Kantian view also helps explain a third feature of our experience of
Our reactions to apparent dilemmas exhibit an asymmetry. As described earlier, from the
first-personal standpoint of the person whose dilemma it is, the aftermath of a dilemma often
brings self-reproach, a fact that my revisionary Kantian view is able to explain. However, for
third parties to hold similarly reproachful attitudes about how other agents acted in response to
moral dilemmas seems less appropriate. The proper responses to other people having confronted
apparent moral dilemmas include sympathy and solidarity, rather than condemnation or emphatic
judgment. Sympathetic third parties may well have opinions concerning how the dilemma was
best resolved, and it would not be inappropriate to provide those opinions if, for example, the
agent who was in the dilemma sought advice or insight after the fact. But the impulse to
evaluate the conduct of the agent in the dilemma is properly subordinated to less epistemically
grounded responses such as sympathy and solidarity.
There is, then, an asymmetry between first-personal responses to moral dilemmas and
third personal responses to dilemmas. Let us call the appropriate first-personal response the
verdictive standpoint and the appropriate third-personal response the sympathetic standpoint.
This attitudinal asymmetry is difficult to explain. One possible explanation for this
asymmetry is epistemic: ‘Outsiders’ to a dilemma may believe that they have insufficient
understanding of the situation to form a reasoned moral judgment about the dilemma (or at least
not the level of understanding that the individual in the dilemma has simply by virtue by being in
it). Hence, the inappropriateness of outsiders’ adopting the verdictive standpoint, and the
corresponding appropriateness of the sympathetic standpoint, reflect outsiders’ reasoned
suspension of judgment, i.e., they simply do not know enough to form a reasoned moral
judgment (or their judgments cannot be as well-grounded as the judgments of the individual in
the dilemma). But this too cannot be the explanation of this asymmetry. For these third-party
attitudes do not spring from a outright suspension of judgment. Third party sympathy and
solidarity are not contingent on third parties’ own judgment concerning how the dilemma is best
resolved. An outsider may have just as much information or moral probity as the agent actually
in the dilemma and come to conscientiously agree or disagree with the actions taken by that
agent. But sympathy or solidarity are called for in either case. That this asymmetry does not
stem from third-parties’ epistemic perspective on the dilemma does not mean that others’
judgments concerning the dilemma place no limits on how they respond. Third parties must at
least think that the individual is in at least an apparent dilemma — that she is confronting a ‘hard
case,’ so to speak — in order to feel sympathy or solidarity with her. A dogmatic moral
opponent of abortion, for instance, may not feel sympathy for an agent who, caught in an
apparent dilemma, opts for an abortion. After all, for the opponent of abortion, the situation does
not even present an apparent dilemma.25
Another tempting explanation is that this asymmetry is only superficial. While it may not
be appropriate for individuals to assert or express critical or judgmental attitudes about how
others respond to dilemmas, one might argue, it is not inappropriate for third parties to have such
attitudes. Rather, the appropriate attitudes to express are those of sympathy or solidarity,
regardless of whether these are one’s true attitudes. But this cannot be the correct explanation
either. For one thing, a person who has unsympathetic attitudes may experience self-reproach of
her own simply for having those attitudes. In other words, a person may subject herself to selfcriticism because she wishes she felt sympathy or solidarity in response to another person’s
In any event, our sympathy or solidarity for those who must act in difficult moral dilemmas does not immediately
follow upon an epistemic suspension of judgment concerning such dilemmas. For one can lack both the verdictive
response and the sympathetic response.. Hence, whatever considerations militate against the verdictive response,
those considerations do not suffice for the sympathetic response.
dilemma, while in actuality she is indifferent or even harshly judgmental. This indicates that
whatever may be wrong with expressing verdictive responses, its wrongness supervenes on
having those responses in the first place.
Lastly, one might think that the third-personal sympathy, etc., occur because third parties
recognize that agents in apparent dilemmas often have their own interests or concerns at stake in
those dilemmas. For example, one reason to sympathize with Sophie is that one of her children
had to die, regardless of how she responded to her dilemma. Certainly sympathy for other
agents’ anguish or suffering is usually justified. But this cannot be the whole picture, for not all
moral dilemmas are self-affecting in this way. Physicians, military leaders, or educators may
confront moral dilemmas in which their own concerns or interests are not implicated, and yet
sympathy and solidarity still seem to be appropriate response by third parties. Furthermore, our
sympathies for those in self-affecting dilemmas extend beyond what happens to them. They
extend to what they must do. For imagine if Sophie were not made to choose between her
children, but instead, a Nazi officer simply killed one of her children. Sympathy and solidarity
for her loss would of course be justified in that case. But if Sophie herself is made to choose
which of her children should die, there is an additional set of reactive attitudes we ought to feel
as a response to her having been placed in a dilemma. Our sympathy, solidarity, etc., reflect not
simply what happened to her, but also how her agency became entangled in a dilemma such that
what happened became the product of her agency.
Of course, a number of factors may rightfully mitigate our sympathy with those in moral
dilemmas. For instance, our sympathies lessen if the dilemma is of the agent’s own making (if,
by having chosen differently in the past, she might have avoided the dilemma altogether). But
explaining this asymmetry is challenging, and the non-agnostic rivals to my revisionary Kantian
view succeed in explaining only half of the asymmetry. The thesis that there are no genuine
moral dilemmas helps to explain the first-personal verdictive standpoint. Since there are no
genuine dilemmas, agents in apparent moral dilemmas strive to resolve them, and this
necessarily involves arriving at justifiable verdicts concerning how one acts (or acted) in a
dilemma. However, the thesis that there are no genuine moral dilemmas can only explain thirdpersonal sympathy in the very epistemic terms I have already suggested are inadequate. For if
there are no moral dilemmas, why should we have sympathy for another person’s being in a
dilemma aside from our being impressed with the difficulty of adjudicating the dilemma?
Conversely, that there are genuine dilemmas can explain third-personal sympathy but not the
first-personal verdictive standpoint. Those who are not party to a dilemma justifiably experience
a sense of its injustice to the individual whose dilemma it is. But if a situation is a genuine
dilemma, the agent whose dilemma it is has comparatively little basis for self-criticism. Being
concerned to get a dilemma right is intelligible only if there is a unique something to be gotten
right, which is not the case in a genuine moral dilemma.
Thus, the non-agnostic views about moral dilemmas can only account for different halves
of this attitudinal asymmetry. However, if the denial of dilemmas is merely a regulative ideal,
then an explanation for this asymmetry itself comes into view. I have argued that practically
rational agents, as Kant understood them, accept the resolution of moral dilemmas as a
constitutive aim of their practical deliberation. However, this claim is ambiguous. The notion of
a moral dilemma can be analyzed in terms of situations where individuals function as variables,
i.e., a situation is a moral dilemma iff any agent S in that situation would be subject to
incompatible claims of obligation. But this characterization is clearly inadequate. For one thing,
not just any agent S will be in a dilemma if placed in such a situation. In some cases, features
specific to S (her profession, her loyalties, etc.) determine whether or not the situation is a
dilemma in the first place. More fundamentally, every dilemma is someone’s dilemma. If there
are simply no possible agents for whom a situation would appear to generate incompatible
obligations, that situation is not a moral dilemma. At root, moral dilemmas are relations between
individuals and their circumstances so that talk of a situation’s being a dilemma must ultimately
be analyzable in terms of someone’s being in a dilemma. Moral dilemmas are thus
fundamentally local. And just as we can speak of headaches in the generic, so too can we speak
of situations as dilemmas, but we should not be deceived into believing that dilemmas exist
independently of the agents whose dilemmas they are anymore than we should be deceived into
believing that headaches exist independently of those individuals whose headaches they are.
In claiming that moral dilemmas are fundamentally someone’s dilemmas, I do not intend
to deny that some moral dilemmas present themselves to agents in concert. Many quite ordinary
moral decisions (for instance, how much to provide others by way of charity) invite dilemmas for
many agents. In fact, some of the most unsettling dilemmas involve multiple agents because
they are symmetrical. Individuals in the stereotypical ‘desert island’ scenario may each be torn
between their morally justifiable interest in survival and their moral reservations about the use of
violence to secure the means to that survival. Nor am I suggesting that moral dilemmas should be
approached solipsitically, as if each apparent dilemma is so novel that agents should not draw
upon moral insights learned from prior dilemmas. Nevertheless, this picture of the ontology of
dilemmas — that dilemmas qua relations are prior to dilemmas qua generic situations — bears
on how we should understand the rationality of the prerogative associated with the regulative
Kantian ideal I have defended.
I have argued that the denial of moral dilemmas is a regulative ideal that both structures
and motivates moral deliberation and that this is an ideal for practically rational agents who seek
to govern their conduct in accordance with morality’s categorical demands. If moral dilemmas
were fundamentally generic situations, then the reasons that underlie this regulative ideal might
appear to be agent-neutral. As Phillip Pettit describes it, an agent-neutral reason is a reason that
can be fully specified without irreducible indexical reference to an individual.26 If the rationality
of the regulative ideal consisted in providing agents with agent-neutral reasons to resolve moral
dilemmas, then an agent’s relation to the reasons she has for seeking to resolve dilemmas is
strictly incidental. She happens to find herself in moral dilemmas due to contingent facts, and
her reason for seeking to resolve these dilemmas is that moral dilemmas ought to be resolved. In
contrast, since moral dilemmas are fundamentally relational, an agent has agent-relative reasons
to seek to resolve those dilemmas she finds herself in, where an agent-relative reason is “one that
cannot be fully specified without pronominal back-reference to the person for whom it is a
reason.” In other words, that a moral dilemma is her dilemma is part of the rational force behind
her aspiration to resolve the dilemma. The reasons generated by the regulative ideal that denies
genuine moral dilemmas are agent-relative. That is, aside from the substantive moral
considerations associated with the dilemma itself (that, depending on which course of action the
agent takes, a person may be injured, a promise may be broken, etc.), which may provide either
agent-relative or agent-neutral reasons for action, an agent aspires to resolve a moral dilemma for
reasons that are irreducibly hers, and more exactly, because the dilemma is normatively realized
by her involvement in it.
“Universality without universalism,” Mind 72 (1987), p. 75.
At the same time, however, each moral agent has the resolution of the dilemmas in which
she finds herself as an aim of her practical deliberation. Hence, each agent has agent-relative
reasons to seek the resolution of her dilemmas. Thus, the regulative ideal that denies, for
practical purposes, the existence of genuine moral dilemmas is a universal but agent-relative
norm. It is a norm grounded in the common, constitutive agential features of rational humans
who acknowledge the categoricity of morality’s demands.27
This way of understanding the normativity of the denial of moral dilemmas allows us to
explain the aforementioned asymmetry between the verdictive first-personal standpoint and the
sympathetic third-personal standpoint on moral dilemmas. The verdictive quality of our firstpersonal responses to the moral dilemmas we find ourselves in is a reflection of the rational
aspiration to resolve all such dilemmas in accordance with the Kantian regulative ideal. Yet
when we encounter others in moral dilemmas, our acceptance of this regulative ideal elicits our
sympathy because we see other agents in morally fraught situations structurally analogous to
those we know or can imagine. To focus attention principally on the epistemic or moral aptness
of the verdicts others reach in their dilemmas is to adopt, however implicitly, the stance that their
agent-relative reasons that animate their desire to resolve their dilemmas are our reasons too —
that their reasons are not simply analogous to ours, but are ours, with the result that we should
adopt the deliberative first-personal standpoint that strives to resolve the dilemmas. Thus, we
engage in a kind of bracketing of our epistemic and moral judgment of others, without fully
renouncing our own judgment.
The sort of reason is similar to what Christine Korsgaard has called a “mixed reason.” See her “Reasons we can
share: An attack on the distinction between agent-neutral and agent-relative values,” in her Creating the Kingdom of
Ends (Cambridge: Cambridge U. Press, 1996), p. 288.
In this respect, though the regulative ideal I have defended is methodological rather than
being a substantive moral principle of its own, it nevertheless has moral implications. When
agents in moral dilemmas adopt the verdictive standpoint toward their own decisions and
conduct, and when third parties adopt the sympathetic standpoint toward those decisions and that
conduct, they respectively exhibit the Kantian virtue that Jeanine Grenberg has called “proper
humility.” Agents with proper humility are aware of their limitations as moral agents, and thus
manifest humility. Yet at the same time, they enjoy self-respect that grounds their recognition of
their worth as moral agents and their entitlement to claim morally decent treatment from others.
Proper humility combines these attributes resulting in a “meta-attitude” wherein the rational
agent perceives herself as “dependent and corrupt” but “capable and dignified”28 in striving to
honor the moral principles she takes as her own. The regulative ideal that denies moral
dilemmas is characteristic of Kantian agents with proper humility. Such agents are committed to
resolve moral dilemmas because they are committed to the often daunting task of according their
conduct with the supreme principle of morality. Agents with proper humility therefore see other
agents entangled in moral dilemmas not as their inferiors but as engaged in a common, and
humbling, human endeavor. Indeed, it would be wrong for them to judge others’ reactions too
harshly, for that would run afoul of their duty “to sympathize actively in [others’] fate.”29
In so doing, they judge themselves not in relation to one another but in relation to the
supreme principle of morality they and other rational agents accept. Having developed the
instinct of conscience, the instinct to “to direct oneself according to moral laws,”30 such agents
Kant and the Ethics of Humility (Cambridge: Cambridge U. Press, 2005), p. 133.
Kant, Metaphysics of Morals 6:457
Lectures on Ethics 27:351
thereby spurn “moral egoism”31 and enjoy “moral self-esteem”32. As Grenberg explains it, this
proper appreciation of moral principles transcends the self in that it transcends particular agents’
perspectives. But we do not transcend our selves entirely since moral principles, including the
denial of moral dilemmas as a practical implication of the categoricity of the moral law, are
autonomously given. The denial of moral dilemmas thus provides a standard that instead of
being “foreign, other, inaccessible, and truly not one’s own,” is a standard to which we aspire
because it is a consequent of the moral law. The Kantian regulative ideal is thus a transcendent
and “simultaneously local” standard whose authority is rooted in the “self-evaluation of a person
claiming status as a rational being.”33
The regulative ideal I have defended does not logically preclude non-agnostic views of
moral dilemmas. Nevertheless, the phenomenology of moral dilemmas suggest that the
revisionary Kantian view more fruitfully and parsimoniously accounts for central features of our
experience of moral dilemmas than do non-agnostic views.
7. Conclusion
Since my thesis is unorthodox both as an interpretation of Kant and as a position
concerning moral dilemmas, its practical implications could easily be misunderstood. I do not
claim that every apparent dilemma is a genuine one. Many apparent dilemmas will turn out to be
resolvable once we come into possession of additional facts, refine the maxims we are
considering acting upon, etc. Some apparent dilemmas could in fact be genuine dilemmas. Some
will prove irresolvable in practice even under optimum epistemic and deliberative conditions.
Lectures on Ethics 27:359
Lectures on Ethics 27:349
Grenberg, Kant and the Ethics of Humility, p. 142.
We simply cannot know in advance and without careful attention to contextual detail whether or
not an apparent dilemma will prove resolvable. For that fact will, in these instances, not be one to
which we finite rational agents have access.
Remaining agnostic about the metaphysics of genuine moral dilemmas while treating the
denial of dilemmas as a methodological precept permits us to steer clear of two conclusions that
Kantians ought to hold in suspicion. If genuine moral dilemmas exist, then the enterprise of
conforming our conduct to the Categorical Imperative’s categorical demands is undermined.
The very goal of a “metaphysics of morals” becomes untenable. On the other hand, a sweeping
denial that genuine moral dilemmas exist requires an unwarranted confidence in our rational and
moral powers, a kind of arrogance that flies in the face of our own moral experience. Thus, if
my revisionary Kantian view is at all plausible, it represents an intriguing hybrid of the
experimentalist and rationalist orientations. In treating the denial of dilemmas as a regulative
ideal rooted in practical reason, my view reveals a commitment to the rationalist aspiration for
systematize our moral judgments. Yet the revisionary Kantian view also honors the “the
standpoint of the moral experience of persons” and the “circumstances of perplexity and choice.”
Indeed, it provides a more accurate account of this standpoint and of the circumstances of
perplexity and choice than does its more metaphysically ambitious rivals.
Conversely, despite its Kantian heritage, my position is not one which moral theorists in
other traditions ought to reject ab initio. Moral inquiry and moral deliberation presuppose, and
aspire to, the resolution of any apparent dilemmas that arise in the course of these endeavors.
This claim is relatively uncontroversial and should not invite hostility from non-Kantians. Of
course, some moral theories may be inhospitable to the possibility of moral dilemmas. Actconsequentialism, for instance, permits numerical ties between alternative courses of action, but
does not treat these actions as representing a dilemma but as representing equally defensible
options. And while I believe that such non-agnostic theories become less plausible in light of the
phenomenological advantages that my revisionary Kantian view offers, even these non-agnostic
theories need not reject the regulative ideal. Moreover, although the claims that moral inquiry
and moral deliberation presuppose and aspire to the resolution of moral dilemmas, and that this
regulative ideal stems from a rationalist aspiration to systematize our moral understanding, are
claims that are most at home in a Kantian framework, they are intelligible within other theories
too. The rational aspiration to resolve moral dilemmas is rooted in a yearning to govern our
conduct by morality’s demands and is hence an expression of the autonomy of practical reason.
However, this does not necessitate that the content of morality’s demands be grounded in that
autonomy, as Kantians do. Nor am I making the implausible claim that moral agency as such
requires acknowledgment of the moral ideal. Dogmatists and ideologues will deny the
possibility of such dilemmas, and I do not deny that dogmatists and ideologues are moral agents.
However, if moral agents acknowledge even the appearance of moral dilemmas, they shall find
themselves compelled by the regulative ideal I have identified to deploy their moral powers in
the service of these dilemmas being resolved.
Most importantly, my revisionary Kantian view explains why the appearance, if not the
actuality, of moral dilemmas causes us to undergo a feeling of rationally justified dismay
directed not at the world but at ourselves insofar as morality is an enterprise rooted in our own
rational self-governance. As a prominent Kantian has written:
Life does prey on life; nature is a scene of suffering; if those things are repugnant to human moral
standards, then the world is set up in a way that we must deplore, but in which we must nevertheless
participate. But on a Kantian conception of morality, this objection is not to the purpose. For Kant
believed that moral standards, like all rational standards, are essentially human standards, and there is no
guarantee that the world will meet them, or make it possible for us to do so. 34
At the same time, my account permits a guarded optimism about our ability to make moral
reality tractable by resolving each and every moral dilemma. In this regard, the revisionary
Kantian view I have defended reflects the sentiments that our knowledge of morality as issuing
in categorical demands should guide our action, so that reason compels us to exercise our moral
capacities to the utmost and thereby seek to resolve the apparent moral dilemmas we confront,
regardless of whether the world is ultimately constituted so as to thwart our rational aspirations.
Christine Korsgaard, “Fellow creatures: Kantian ethics and our duties to animals,” Tanner Lectures on Human
Values (delivered at the University of Michigan, 2004), p. 35.