Filtering Friendship through

Phronesis: ‘One Thought too Many’?

KRISTJÁN KRISTJÁNSSON
Abstract
An adequate moral theory must – or so many philosophers have argued – be compat-
ible with the attitudes and practical requirements of deep friendship. Bernard
Williams suggested that the decision procedure required by both deontology and
consequentialism inserts a fetishising filter between the natural moral motivation
of any normal person to prioritise friends and the decision to act on it. But this
injects ‘one thought too many’ into the moral reaction mechanism. It is standardly
assumed that virtue ethics is somehow immune to this objection. The present
article explores this assumption and finds it wanting in various respects. Virtue
ethics filters friendship through phronesis and thus inserts an extra thought into
the mechanism in question. To escape Williams’s curse, the only way is to argue
that the extra thought required by virtue ethics is not ‘one thought too many’. The
article closes with an attempt to show that, contra deontology, the friendship motiv-
ation in virtue ethics is derived from the moral virtue, not the intellectual filter, and,
contra consequentialism, phronesis does not require the maximisation of value. The
presumed advantage of virtue ethics must lie in the content of its filter rather than
the filter’s non-existence.

1. Introduction

For most people, philosophers as well as non-philosophers, friend-

ship constitutes a significant – if not necessarily an unalloyed –
good, constitutive of human wellbeing. In Aristotle’s view, on
which most subsequent friendship discussions have drawn, ‘no one
would choose to live without friends even if he had all the other
goods’.1 This does not mean that friendship is the ‘highest good’
or, insofar as it is to be understood as a virtue, the ‘master virtue’,
trumping all others, for other goods, such as health, could well
have the same status. Possessing all other conceivable goods in life
is somewhat useless if one does not have the health to enjoy them.
Nevertheless, various philosophers have taken up the cudgel for
Aristotle and turned friendship into a theoretical construct of

1
Aristotle, Nicomachean Ethics, trans. T. Irwin (Indianapolis: Hackett
Publishing, 1985), 207 [1155a5–6].

doi:10.1017/S0031819119000433 © The Royal Institute of Philosophy, 2019

First published online 11 October 2019
Philosophy 95 2020 113
Kristján Kristjánsson

sufficient moral salience to serve as a benchmark for the evaluation of

general moral theories. An adequate moral theory must, for example,
on Badhwar Kapur’s understanding, be compatible with the attitudes
and practical requirements of friendship, correctly conceptualised.2
In a seminal paper by Bernard Williams,3 which blazed a trail of
numerous follow-ups in the next two decades, he argued that both
the reigning moral theories of the day, utilitarian consequentialism
and Kantian deontology, failed this adequacy test for a similar
reason: namely, by compelling us to subject our obvious natural
choice to prioritise the needs of a close friend (or a loved one) over
those of a stranger, in times of moral danger, to a theoretical decision
procedure before coming up with the ‘right’ reaction. Such require-
ments of reflective calculation rob us, according to Williams, of
psycho-moral reasons to live at all by attacking the source of any in-
tegrity-grounding prime motivation that makes us tick. The decision
procedure required by these two moral theories – be it the categorical
imperative or the utility calculus – inserts a filter (which Williams
seems to think of as an artificial gadget or fetish) between the
natural moral motivation of any normal person and the decision to
act on it. But this injects what Williams famously terms ‘one
thought too many’ into the moral reaction mechanism and fetishises
it in the service of a psychologically overbearing theory.
The problem identified by Williams has become known as ‘the
problem of integrity’, but as Williams uses ‘integrity’ in a somewhat
idiosyncratic sense with respect to either lay or philosophical uses,4 I
prefer to refer to it as ‘the problem of alienation’. Williams basically
urges that being held in thrall by the rationalist demands of the cat-
egorical imperative or the amoral assumption of the utility calculus
(which considers pleasure as the highest good) alienates us from
our most significant others and in the end from ourselves.
Juxtaposing this argument with Aristotle’s well-known one about
the predicament of the vicious who, because they are not capable of
loving themselves, also become incapable of loving others,5
2
N. Badhwar Kapur, ‘Why It Is Wrong to Be Always Guided by the
Best: Consequentialism and Friendship’, Ethics 101 (1991): 483–504,
p. 485.
3
B. Williams, Moral Luck: Philosophical Papers 1973–1980
(Cambridge: Cambridge University Press, 1981).
4
Cf. K. Kristjánsson, ‘Is the Virtue of Integrity Redundant in
Aristotelian Virtue Ethics?’ Apeiron 53 (2019): 93–115.
5
Aristotle, The Athenian Constitution, The Eudemian Ethics, On Virtues
and Vices, trans. H. Rackham (London: William Heinemann, 1935), 405
[1240b15–19]; op. cit. note 1, 246–247 [1166b1–25].

114
 Filtering Friendship through Phronesis

Williams turns the psychology upside down: because the moral

fetishisers are barred from forming integrity-grounding uncondi-
tional commitments to their beloved ones (including their closest
friends), they also become alienated from their core commitments
to themselves as moral agents; the commitments that give them any
reasons for living or acting at all. The ‘one-thought-too-many’ argu-
ment has become something of a mantra and I revisit it in more detail
in Section 2, albeit only insofar as it is relevant for the purposes of the
present article.
So what are the ‘present purposes’ then? My aim is, briefly put, to
explore a specific assumption that has emerged from the above-men-
tioned literature. While not explicitly elicited by Williams himself,
the lesson that most scholars seem have drawn from the problem of
alienation is that virtue ethics is somehow invulnerable to it and
hence better equipped to deal with the desirably realistic features of
common-sense morality that make us commit to it in the first
place. Needless to say, there are various other features that may
draw people towards virtue ethics as an alternative to the other two
moral theories – one commonly noted being its facility to make
sense of the role of emotions in the morally good life.6 However, it
is typically suggested or even stated without argument that the fact
that only virtue ethics makes do without a filter between deep friend-
ship – as a ‘virtue’ or relation between people that ‘involves virtue’7 –
and moral decisions provides a reason to abandon utilitarianism and
deontology and adopt virtue ethics as one’s moral theory.8 My aim is
to problematise the assumption about this unique advantage of virtue
ethics. There are many things to like about virtue ethics, but this is, I
submit, not one of them.
Notice some odd features about this assumption. First, it is not
clearly elicited by Williams himself, as already noted. Second,
despite the flurry of responses that followed Williams’s piece, none
developed in detail – to the best of my knowledge – the positive
side of the argument. In other words, the claim that virtue ethics
has unique resources to counter the one-thought-too-many argument
simply continued to be implicitly assumed rather than argued for.
Third, the assumption in question would work if Aristotelian

6
K. Kristjánsson, Virtuous Emotions (Oxford: Oxford University
Press, 2018).
7
Op. cit. note 1, 207 [1155a1–2].
8
See e.g. R. F. Card, ‘Consequentialism, Teleology, and the New
Friendship Critique’, Pacific Philosophical Quarterly 85 (2004): 149–172,
p. 149, although Card refrains from categorising virtue ethics as a ‘theory’.

115
Kristján Kristjánsson

virtue ethics were an intuitionist moral theory. However, any intu-

itionist readings of Aristotle tend to be misreadings (see Section 3).
Alternatively, it would work if friendship were an overriding
master virtue in Aristotle’s system, but it is not, with Aristotle
himself saying in the Eudemian Ethics that those who unreflectively
‘give everything to one whom they love [qua philia] are good-for-
nothing people’.9 In fact, there is no master moral virtue in
Aristotle on a par with, say, justice in Plato.
There is, however, unquestionably, a meta-virtue in Aristotle’s
virtue ethical system: namely, the intellectual virtue of phronesis. In
addition to helping individual moral virtues find means to their
ends, phronesis helps solve apparent virtue conflicts, and it also
informs the content of the virtues as they are understood by the
agent, by bringing them into harmony with an intellectually
grounded blueprint of the good life – for Aristotle’s is not a
Humean theory of moral motivation. I say more about this in
Section 4. At the present juncture, it suffices to note that phronesis
seems to provide a filter through which any virtuous considerations
needs to pass – including those of friendship – before they can justi-
fiably issue in either reason-imbued emotion (such as compassion
towards a friend) or action (such as helping a friend). It is therefore
hard to shake the impression that phronesis imports an ‘extra
thought’ between the motivational force of virtuous friendship as a
disposition and particular (re)actions of friendship. Does such filter-
ing of friendship through phronesis fall prey to Williams’s one-
thought-too-many argument against moral theories and undermine
the assumption about the unique advantage of virtue ethics – or do
not all extra thoughts count as ‘one thought too many’? These ques-
tions call for some sustained analysis in subsequent sections. To an-
ticipate, my conclusion will be that while Williams’s own
understanding of the one-thought-too-many argument is too
radical, it is worthwhile inquiring what sort of an extra thought
counts as detrimental to moral theorising and what sort does not. I
argue that although phronesis elicits an extra thought, it falls into
the latter category.
The kind of friendship I am interested in here, and to which
Williams was clearly referring, is what we would normally refer to
as ‘deep’ or ‘best’ friendship. For the sake of simplification, I
assume that ‘deep’ or ‘best friendships’, on a contemporary under-
standing, are close enough to ‘character friendships’ in Aristotle’s
well-known qualitative tripartite system (of friendships for pleasure,
9
Op. cit. note 5, 433 [1244a17–19].

116
 Filtering Friendship through Phronesis

utility and character) to speak of them in the same breath. I happen to

agree also with Aristotle’s controversial contention that the highest
and only ‘complete’ form of friendship, namely friendship for char-
acter, is grounded in moral qualities rather than, say, aesthetic ones.10
A fair number of caveats would ideally need to be entered here about
Aristotle’s friendship theory, but I will limit myself to observing that
Aristotle’s account of character friendships is unduly idealised and
insensitive to numerous potential difficulties.11 Be that as it may,
the aim of this article is not to defend an orthodox Aristotelian speci-
fication of friendship. The aim is rather to argue that the assumption
that (deep) friendship, on an everyday contemporary understanding,
is somehow an unconditional, unreflective virtue in Aristotelian
virtue ethics, is misguided; friendship requires ‘filtering’ in
Aristotelian virtue ethics just as in the other two major moral theor-
ies, albeit not filtering that necessarily imports ‘one thought too
many’.

2. Some reflections on the problem of alienation

As most of the responses that fuelled the lengthy debate about

Williams’s bombshell argument have come from consequentialists,
and consequentialism is logically closer to virtue ethics than deontol-
ogy (in being teleological), I will focus on those responses here. The
aim is not to rehearse this debate in any detail, but simply to fore-
ground the features that may be relevant for the discussion in the re-
mainder of this article.
The first thing to note is that the problem of alienation is logically
distinct from the other standard objection lodged against consequen-
tialist theories, especially of the traditional utilitarian kind: the
problem of repugnant consequences (aka ‘the problem of victimisa-
tion’). Whereas the alleged repugnant consequences (e.g. in well-re-
hearsed transplant and trolley cases) point to moral errors in
consequentialism, and the typical responses (about the need to
focus on long-term consequences, including those of precedents
set, or on the threat to overall utility incurred by sacrificing people
in lower moral risk zones for people in higher zones) seek to show

10
Contra e.g. A. Nehamas, On Friendship (New York: Basic Books,
2016).
11
K. Kristjánsson, ‘Ten un-Aristotelian Reasons for the Instability of
Aristotelian Character Friendships’, Journal for the Theory of Social
Behaviour 49 (2019): 40–58.

117
Kristján Kristjánsson

that consequentialism is not prone to those moral errors, the problem

of alienation is not first and foremost a moral problem. What sort of a
problem is it then? Williams’s own words often indicate that it is a
problem of rationality: namely, that it is ‘unreasonable for a man to
give up […] something which is a condition of his having any interest
in being around in the world at all’.12 However, a closer look at
Williams’s argument reveals an even more menacing and deep-
rooted source, for Williams also introduces the idea of a ‘categorical
desire’, the satisfaction or non-satisfaction of which settles the ques-
tion of whether the agent cares to stay alive or not.13 For all normal
people, Williams assumes, the desire to prioritise the needs of close
friends constitutes such a categorical desire. Hence, it is psychologic-
ally impossible at the same time to suppress this desire (in the interest
of a moral theory) and to continue to live. To succumb, say, to the
utility-maximising demands of consequentialist moral theories is,
therefore, not so much immoral and irrational as simply stretching
human psychology beyond its breaking point. It is, in Williams’s
sense, not alienating only vis-à-vis the significant other(s), but self-
alienating and self-destroying.
The standard interpretation of Williams’s argument is, as Wolf
correctly points out,14 that the psychological impossibility kicks in
at the moment of the moral decision. So it is the person who thinks
at the time of action about what would be morally permissible (e.g.
in a case involving a choice to save a close friend/spouse versus a
stranger) who falls prey to the problem of alienation, not the person
who prepares herself, for instance, as a moral learner for facing
moral dilemmas later in life by thinking through various possibilities
beforehand in order, say, to strengthen her commitment to her cat-
egorical desires or to figure out what those really are. Incidentally,
I agree with Wolf that this interpretation of Williams’s argument is
not radical enough, but before elaborating on that point, a few remin-
ders are in order about how consequentialists have tried to parry it, on
the standard interpretation.
Responding to Williams has turned into a whole cottage industry.
Many of these responses take the form of rejecting the claim that

12
Op. cit. note 3, 14, my italics; cf. M. Bernstein, ‘Friends without
Favouritism’, Journal of Value Inquiry 41 (2007): 59–76, 67–68.
13
Op. cit. note 3, 11.
14
S. Wolf,‘“One Thought too Many”: Love, Morality, and the
Ordering of Commitment, in U. Heuer and G. Lang (eds.), Luck, Value,
and Commitment: Themes from the Ethics of Bernard Williams (Oxford:
Oxford University Press, 2012): 71–92, 74.

118
 Filtering Friendship through Phronesis

thinking through moral possibilities at the time of action, in cases

such as the above, is bound to alienate the thinker from others and
then herself. The general complaint is that this argument romanti-
cises and de-intellectualises deep friendships overly and overlooks
the continually morally reflective and probing nature of at least
some such friendships.15 This general complaint can then be devel-
oped along various argumentative avenues, for instance by illustrat-
ing how, at least in the case of two devoted consequentialists, the
decision to honour the mutual friendship may be seen as even more
precious and noble by the friend if it involves, and is reached on
the back of, a lengthy reflection on the general happiness of human-
kind.16 It is even possible to envisage a conscious pact made between
two consequentialist friends that they will never favour each other
over others except as a result of rigorous deliberation about the
total state of the world, and that they admire each other the more
they hold to this pact, even when the friend’s decision goes against
them in the end. The trouble is that, while one can imagine certain
people deriving fulfilment from such considerations and such a
pact (say, someone like Harriet Taylor and John Stuart Mill), this
would hardly generalise to the rest of humankind. Williams’s argu-
ment does not require that all people have a categorical desire of
the sort he describes; it suffices that the majority of normal moral
agents do. Otherwise, the demands of consequentialism are prone
to fall foul of Flanagan’s criterion of ‘minimal psychological
realism’:17 of not being feasibly attainable for (most) beings like us.
A more promising line of response is to accept Williams’s claim that
thinking through moral possibilities at the time of action, in cases
such as the one about the friend and the stranger, is likely to be
self-alienating, and then to make sure somehow that one’s preferred
version of consequentialism accommodates this fact. One way of
doing that is to adopt rule utilitarianism, rather than act utilitarian-
ism, as one’s conscious moral theory, and to argue that many privi-
leged duties to friends are justified by their overall conduciveness
to the maximisation of the general good, even if they happen to

15
See e.g. P. Railton’s classic piece, ‘Alienation, Consequentialism, and
the Demands of Morality’, Philosophy and Public Affairs 13 (1984):
134–171.
16
See e.g. E. Conee, ‘Friendship and Consequentialism’, Australasian
Journal of Philosophy 79 (2001): 161–179; Bernstein op. cit. note 12.
17
O. Flanagan, Varieties of Moral Personality: Ethics and Psychological
Realism (Cambridge, MA: Harvard University Press, 1991), 32.

119
Kristján Kristjánsson

appear to be utility-reducing in a particular case.18 However, rule

utilitarianism imports problems of its own, both because of its ten-
dency to collapse logically into act utilitarianism19 and because
adopting it consciously as one’s preferred moral theory seems to
call for reflection at the time of action, which is exactly what
Williams’s argument debars us from doing. More plausibly, the con-
sequentialist could stick to act utilitarianism but augment it with the
psychological thesis that considering alternative possibilities to prior-
itising the needs of a close friend at the time of action is psychologic-
ally incompatible with a concern for utility, and such considerations
should therefore be forestalled. For if it is in fact true that (even con-
sidering the possibility of) not prioritising the friend robs the agent
psychologically of the will to live, then there are good consequential-
ist reasons for habituating oneself into prioritising the friend spontan-
eously (as distinct from adopting the prioritisation as a rule to follow).
The objection that the world would be better still if considerations of
overall utility could be engaged in at the time of action will not cut ice
with the consequentialist who has taken this psychological thesis on
board because consequentialism is not a theory about other possible
worlds, but just this world, and what ‘a consequentialist theory tells
us we ought to do is always actually possible’.20
At the risk of getting ahead of my argument in Section 3, let me
remark here that this response to the standard interpretation of
Williams’s argument may also seem to make Aristotelian phronesis
immune to it, at least on one reading of phronesis. Some scholars em-
phasise the developmental function of phronesis as an intellectual
virtue that prepares agents beforehand for taking the right decision
by ‘metabolising the past to simulate possible futures’.21 This will
then explain the facility of phronesis to get things ‘intuitively’ right
at the time of action. It is not because phronesis itself serves as a
vehicle of intuition, but rather because it has prepared us so well for
what could happen that once we enter into an already-reflected-upon
situation, all that phronesis needs to do is to activate our sensitivity

18
See e.g. E. Telfer, ‘Friendship’, Proceedings of the Aristotelian Society
71 (1970–1971), 223–241, 235.
19
D. Lyons, Forms and Limits of Utilitarianism (Oxford: Oxford
University Press, 1965).
20
Conee op. cit. note 16, 178–179.
21
P. Railton, ‘Intuitive Guidance: Emotion, Information and
Experience’, in M. E. P. Seligman, P. Railton, R. F. Baumeister and
C. Sripade (eds.), Homo Prospectus (Oxford: Oxford University Press,
2016): 33–85, 45–46.

120
 Filtering Friendship through Phronesis

to this being the sort of situation that calls for a certain reaction,
without the need for further deliberation at the time of action –
because perception happens before deliberation and preempts it.
On this reading, there does not seem to be any danger of phronesis im-
porting ‘one thought too many’. However, this reading does not show
virtue ethics to be superior to simple act utilitarianism in this respect,
provided we grant that the psychological thesis suggested in the pre-
ceding paragraph may also save act utilitarianism from the problem of
alienation.
What seems to be too good to be true is usually too good to be true.
The whole problem to which Williams alerted us seems to have dis-
appeared, simply because initial credibility has been granted to the
standard interpretation of his argument. Yet on that interpretation
the argument bewilders rather than enthralls, in particular if it is
meant to point towards the superiority of virtue ethics as a moral
theory. Snatching potential defeat from the jaws of victory, Wolf ex-
plains well, in an intriguing paper, the extent to which the standard
interpretation defangs Williams’s argument. Williams’s intention
was much more radical that standardly acknowledged: namely, not
only to show that reflecting on what to do, at the moment of action,
in cases involving friend–stranger conflicts, is psychologically impos-
sible, in the sense of being self-alienating, but rather that any reflec-
tions on the possibility of betraying the friend will be self-alienating,
even if engaged in ‘off stage’: be it prospectively, to prepare oneself for
proper decision making at the time of action, or retrospectively, to
justify to oneself what one has done. Removing the deliberations
from the emotionally charged scene of the action does not rescue
the extra thought from being one thought too many.22
On this radical interpretation, there is no way in which the strategy
invoked above can rescue consequentialism because it is impossible to
ask the consequentialist to hold all considerations regarding the
utility of prioritising friends over strangers (or vice versa) in abey-
ance, not only during but also before and after the relevant event.
That simply goes against the grain of the very idea of consequentialist
calculations of the overall good. There must be a time in which those
can be engaged in, no holds barred. The radical interpretation also
brings home to us, much more so than the standard interpretation,
the sense of ‘moral schizophrenia’ that Stocker23 famously attributed
to the endorsement of the two competing moral theories of the day:
22
Op. cit. note 14.
23
M. Stocker, ‘The Schizophrenia of Modern Ethical Theories’,
Journal of Philosophy 73 (1976): 453–466.

121
Kristján Kristjánsson

the sense that those theories propose determinable methods for con-
sidering when goods like deep friendship ought to be outweighed by
more general requirements of duty or the overall good.24 However,
this creates a schizophrenia in the psychology of the normal moral
agent to the extent that what she is meant to value and calculate as
valuable jars with what in fact moves her most. Notably, the radical
interpretation also casts serious doubts on the facility of
Aristotelian virtue ethics to escape from the clutches of Williams’s ar-
gument, for phronesis is surely presented by Aristotle as a method of
moral deliberation, and a determinable one at that, although admit-
tedly not codifiable in exactly the same sense as the utility calculus
or the categorical imperative (see further in Section 3).
The radical interpretation of Williams’s argument presents us with
the following dilemma, given the aim of the present article. Either we
accept the argument with respect to phronesis, but then we implicitly
concede that every extra thought of the kind envisaged above will
count as one thought too many, and that seems to do away with phron-
esis altogether (unless we understand phronesis as some sort of intui-
tive artistry, outstripping conscious thought, but that is a misguided
understanding of phronesis, as I argue in Section 3). Or we reject the
argument with respect to phronesis, but then we need to show either
that phronesis does not import an extra thought into the decision-
making process (which I think is impossible), or that although phron-
esis imports an extra thought qua filter, it is not one thought too many
for some substantive reasons and hence the ‘filter’ is not ‘fetishising’
in the same way as, say, the utility calculus.
While I propose to go down the second avenue in Section 4, we can
safely conclude at this juncture that there are no obvious escape routes
in Williams’s argument (on the radical interpretation) that would
prevent Aristotelian virtue ethics from potentially being sent to the
gallows along with the other two dominant moral theories.
However, before that sentence phronesis needs a fair and thorough
hearing.

3. Phronesis as an intellectual filter

The important role that the intellectual virtue of phronesis plays in all
Aristotelian or quasi-Aristotelian forms of virtue ethics was noted in
Section 1. In order to answer the question of whether phronesis
24
Cf. S. Woodcock, ‘Moral Schizophrenia and the Paradox of
Friendship’, Utilitas 22 (2010): 1–25, 14.

122
 Filtering Friendship through Phronesis

imports one thought too many, various considerations need to be ad-

dressed (see the current and next sections), beginning in this section
with some reflections on what phronesis really is. To cut a long story
short, the easiest way to show that phronesis does not import one
thought too many about deep friendship would be to argue, with
regard to the (a) nature and (b) content of phronesis, that (a) phronesis
does not add an extra thought, but simply serves as an intuition
pump, and moreover (b) that this pump motivates the unconditional
prioritisation of deep friendship as a moral concern. I have already
provided a citation from Aristotle that seems to rule out (b).
However, remember that virtue ethics is a naturalistic moral theory,
answerable to empirical findings on how people actually flourish or
wilt, and it could well be the case that (b) needs to be revised in the
light of new empirical evidence to accommodate a primacy-of-friend-
ship intuition. After such revision, phronesis could still potentially be
considered to retain the spirit, as distinct from the letter, of
Aristotelian virtue ethics. Alternatively, there are other variants of
virtue ethics than the Aristotelian one and some of those could ac-
commodate friendship as a master virtue.25
It is more difficult to revise (a) by eliciting other variants because
the guidance that all leading virtue ethicists in the West tend to
follow on phronesis has been wrenched from Aristotle, and there is,
to the best of my knowledge, no completely un-Aristotelian phronesis
theory out there (although many theorists rely on MacIntyre’s ren-
dering which departs from Aristotle’s in some respects26). Not all
hope is lost, however, of escaping Williams’s curse, for there are
almost as many variants of ‘Aristotelian’ phronesis as there are
Aristotelian exegetes, and some of those understand the workings
of phronesis first and foremost in terms of intuitive artistry rather
than as an extra thought, let alone a thought too many. I will consider
two of those variants later in this section, but first some brief rehear-
sals of Aristotle’s own account are in order.
Aristotle’s phronesis is an intellectual virtue (virtue of thought) that
serves the purpose of living well by monitoring and guiding the moral
virtues. Building on emotional dispositions cultivated through early-
years habituation, phronesis re-evaluates those dispositions critically,
allowing them to truly ‘share in reason’, and provides the agent with

25
Cf. D. Cocking and J. Kennett, ‘Friendship and Moral Danger’,
Journal of Philosophy 97 (2000): 278–296; Cicero, How to Be a Friend,
trans. P. Freeman (Princeton: Princeton University Press, 2018), 35 and 45.
26
See K. Kristjánsson, Aristotle, Emotions and Education (Aldershot:
Ashgate/Routledge, 2007), chap. 11.

123
Kristján Kristjánsson

proper justifications for them. In addition to latching itself on to

every ‘natural’ moral virtue, and infusing it with systematic reason,
the function of phronesis is to ‘deliberate finely’ about the relative
weight of competing values, actions and emotions in the context of
the question of ‘what promotes living well in general’. A person
who has acquired phronesis has thus, inter alia, the wisdom to adjudi-
cate the relative weight of different virtues in apparent conflict situa-
tions and to reach a measured verdict about best courses of action.27
This is, more or less, where the consensus ends on what phronesis
really involves and the controversial interpretations begin, interpreta-
tions which haunt the landscape of Aristotelian scholarship. While I
want to avoid begging controversial questions about what phronesis is,
simply for the sake of moving on with the discussion of Williams’s ar-
gument, I do not think that ‘anything goes’ in Aristotelian exegesis
and I reserve the right below to reject interpretations that are bla-
tantly un-Aristotelian.
Let me highlight here the oddity that although phronesis is under-
going a revival, not only within contemporary virtue ethics, but also
in social scientific circles and in various areas of applied professional
ethics, no psychological instrument currently exists to measure
phronesis (although the measure designed by Brienza and colleagues
perhaps comes close to it28). This is even more remarkable given
the current burgeoning of so-called wisdom research in psychology.29
Serious efforts are now afoot, however, to remedy this shortcoming
by an interdisciplinary team of philosophers and psychologists.30
The advantage of instrument design is that relevant conceptual
nuances need to be elicited and the components of the construct
under examination identified in detail. According to the fairly min-
imalist reading by Darnell and colleagues of Aristotelian phronesis,
it serves at least four distinct functions and thus constitutes what

27
Op. cit. note 1, 153, 154, 159, 164, 171 [1140a26–29, 1140b4–6,
1141b30–31, 1143a8–9, 1144b30–32].
28
J. P. Brienza, F. Y. H. Kung, H. C. Santos, D. R. Bobocel and
I. Grossmann, ‘Wisdom, Bias, and Balance: Toward a Process-Sensitive
Measurement of Wisdom-Related Cognition’, Journal of Personality and
Social Psychology 115 (2018): 1093–1126.
29
See I. Grossman, ‘Wisdom in Context’, Perspectives on Psychological
Science 12 (2017): 233–257.
30
C. Darnell, L. Gulliford, K. Kristjánsson and P. Panos, ‘Phronesis
and the Knowledge–Action Gap in Moral Psychology and Moral
Education: A New Synthesis?’ Human Development 62 (2019): 101–129.

124
 Filtering Friendship through Phronesis

psychologists would call a four-component construct. Here is a quick

overview of those functions.31

(i) Constitutive function. This is the ability, and eventually cognitive

excellence, which enables an agent to perceive what the salient fea-
tures of a given situation are from an ethical perspective, and to see
what is required in a given situation as reason(s) for responding in
certain ways. (ii) Integrative function. This component involves inte-
grating different components of a good life, especially in dilemmatic
situations where different ethically salient considerations or virtues
appear to be in conflict. This function is highly situation-specific,
which means that traditional wisdom research in psychology, which
homes in on more global capacities, is mostly irrelevant to the deriv-
ation of a phronesis construct. (iii) Blueprint function. Phronetic
persons possess a general conception of the good life (eudaimonia)
and adjust their moral identity to that blueprint, thus furnishing it
with motivational force. This does not mean that each ordinary
person needs to have the same sophisticated comprehension of
the ‘grand end’ of human life as a philosopher or an experienced
statesperson might have, in order to count as possessing phronesis.
Rather the sort of grasp of a blueprint of the aims of human life
informing (and informed by) phronesis is within the grasp of the
ordinary well-brought-up individual and reflected in ordinary acts.
It draws upon the person’s standpoint of life as a whole and deter-
mines the place that different goods occupy in the larger context
and how they interact with other goods. This blueprint is ideally
‘on call’ in every situation of action. (iv) Emotional regulation func-
tion. Phronesis requires, and contributes to, the agent’s emotions
being in line with her construal of a given situation, moral judgement
and decision, thereby also offering motivation for the appropriate re-
sponse. Notice that emotional regulation must not be understood
here in terms of emotional suppression or policing, but rather as
the infusion of emotion with reason, which calibrates the emotion
in line with the morally and rationally warranted medial state of
feeling, and the subsequent harmony between the two.
To be sure, this identification of the four core components of
Aristotelian phronesis does not dissolve all exegetical disputes about
the concept. However, it does help fend off seriously aberrant inter-
pretations, including those which consider phronesis a mere intuition
pump. Indeed, I do believe the idea of ‘phronetic intuitionism’ (as

31
See further in ibid.

125
Kristján Kristjánsson

espoused e.g. by Kaspar in this journal32) involves something of an

oxymoron. Let me briefly mention two attempts at ‘intuition-ising’
phronesis, one from the current education literature and the other
from contemporary moral psychology.
There is a powerful approach in recent educational theory (harking
back at least to Dunne33) which offers an (allegedly Aristotelian)
anti-realist, non-foundationalist, perspectivist and particularist
account of education: most felicitously described as a ‘phronesis-
praxix appoach’. I have criticised this approach in detail elsewhere34
and will not repeat that critique here except insofar as it relates to an
intuitionist reading of phronesis. According to Dunne, we need to avoid
seeing phronesis in terms of ‘the application of theory to particular
cases’.35 In praxis, as the sphere of phronesis, ‘practical-moral universals
cannot unproblematically cover or include particular cases’ precisely
because the former contain ‘an element of indeterminateness’.36 This
is, in Dunne’s words, so far from being a defect that it is, rather, ‘the
great merit’ of phronesis:37 best captured by terms such as ‘particularist
discernment’, ‘intuitive artistry’, ‘perceptual capacity’, ‘illative sense’
or ‘situational appreciation’. Phronesis is, in other words, the eye of
moral experience: the discernment of particular situations that
enables us ‘to see aright’ every time, but which remains ultimately ex-
periential rather than universal ‘since the universals within its grasp
are always modifiable in the light of its continuing exposure to par-
ticular cases’.38 This intuitionist reading of phronesis then allows
Dunne and his followers to make sweeping generalisations about
the essential uncodifiability of Aristotelian phronesis-guided ethical
and educational decision making.
The snag is that Aristotle’s much-cited assertion that phronesis is
about particulars and therefore needs perception39 says nothing
about the epistemological priority of perception. A simpler interpret-
ation is that Aristotle considered universal moral beliefs that would be

32
D. Kaspar, ‘How We Decide in Moral Situations’, Philosophy 90
(2015): 59–81.
33
J. Dunne, Back to the Rough Ground: ‘Phronesis’ and ‘Techné’ in
Modern Pilosophy and in Aristotle (Notre Dame: University of Notre
Dame Press, 1993).
34
Op. cit. note 26, chap. 11.
35
Op. cit. note 33, 157.
36
Ibid., 311.
37
Ibid., 314.
38
Ibid., 280, 293, 297 and 361.
39
Op. cit. note 1, 160–161 [1142a12–30].

126
 Filtering Friendship through Phronesis

fully capable of taking into account every possible situation to be so

complicated – although logically possible – that they would in fact
be impossible to learn and apply. Think, for instance, of all the com-
parisons that would need to be made between individuals with sim-
ultaneous, yet diverse, interests. Instead of trying to achieve such a
super-human feat, it would be better to acquire a perceptual aware-
ness that guides us to the right answer in the greatest number of
factual situations – as we, more realistically, define only ‘as far as
we can’.40 Phronesis, while not unproblematically codifiable,
because of its ‘practical’ as distinct from ‘theoretical’ subject
matter, is thus not necessarily (but merely contingently) uncodifi-
able. A perfect moral theory, which resolved once and for all every
question of application, would be possible only for a perfect being.
Yet what remains is the ‘blueprint function’ of phronesis, which
applies a general conception of the human telos to diverse, complex
ethical situations and furnishes the agent with theoretical tools – an
intellectual ‘filter’ if you like – to think through complex practical si-
tuations, rather than having to rely simply on hunches.
A less sophisticated attempt to co-opt Aristotle to the intuitionist
camp has recently been made by social intuitionists: the proponents
of the currently fashionable two-system (dual-process) theories of
moral decision making.41 According to social intuitionism, people
typically experience a moral intuition about a given state of affairs –
an emotion-driven hunch or an implicit sense of what is the appropri-
ate reaction. Such intuitions normally do not require explicit, effort-
ful reasoning; indeed they seem to persist in the face of contrary
rational judgement or of the lack of any rationally grounded convic-
tion. They often arise non-voluntarily and are not fully articulable.
Most importantly, they motivate spontaneous action, uninformed
by conscious deliberation, although people exhibit a tendency –
through motivations such as peer pressure and canonical norms of
discourse – to justify their actions retrospectively. Here would be
the proposed Aristotelian corollary, then. Human beings typically
act upon motivations provided by general traits of character:
vicious, virtuous or somewhere in between. We are essentially crea-
tures of habits (qua traits). These traits include emotions ( pathe)
which are the most immediate motivators of action. However, we
are not really responsible for our episodic emotions, such as our

40
Ibid., 243 [1165a35].
41
See e.g. J. Haidt, ‘The Emotional Dog and its Rational Tail: A Social
Intuitionist Approach to Moral Judgment’, Psychological Review 108
(2001): 814–834.

127
Kristján Kristjánsson

bouts of anger or pangs of jealousy; those happen to us rather than

being chosen by us. Hence, Aristotelian pathe are quite similar to
what the social intuitionists such as Haidt understand moral intui-
tions to be. Indeed, those theorists love the idea of ‘automaticity of
virtue’ in Aristotle.
The problem with this analogy is that it is over-simplified to the
brink of being blatantly wrong. To be sure, Aristotle does not deny
that we may be driven by knee-jerk reactions to events: conditional
reflexes and non-cognitive feelings. However, those would not be
pathe on his understanding, and the claim that pathe are not within
our responsibility elides important complexities. Let it suffice to
say that Aristotelian moral intuitions (qua pathe) are part of a learning
system that is infused with reason – be it good or bad reason. Moral
judgement is in essence an exercise of reason. There is no ‘brute’
moral intuition in Aristotle, and even what he calls ‘natural virtue’
is not ‘natural’ as in either ‘genetically pre-programmed’ or ‘condi-
tioned by the nature of one’s society’. ‘Natural virtue’ in Aristotle
is actually a somewhat infelicitous name for a stage of habituated
but non-phronetic virtue. True, there is both quick and slow moral
decision making in Aristotle, but the difference between the two
does not correspond to that between the non-rational versus rational
or to emotion versus reason. Aristotle was simply not a two-system
dualist, full stop.42
Friendship is indeed a good example here. Early-years habituation,
ideally followed by years of autonomous, critical honing of one’s dis-
positions through the exercise of phronesis, enables us to react quickly
in uncomplicated situations where a close friend needs help. Phronesis
guides us towards the helping behaviour, almost automatically.
However, as soon as the situation becomes more complicated, the de-
cision process slows down, as phronesis needs time to kick in and
evaluate the situation. Do the needs of the friend conflict with
those of another friend, or perhaps a large group of strangers? Has
the friend’s character changed so dramatically for better or for
worse (although Aristotle himself only considered the latter in any
detail) that the virtue of friendship does not apply anymore?
Furthermore, once we are acting through the mediation of phronetic
rather than just natural virtue, the filter becomes much more de-
manding, as it requires not only that we comply with the demands
of the most immediate virtue relevant to the given situation (in our
case, friendship) but that it also takes account of claims proper to
42
See further in K. Kristjánsson, Flourishing as the Aim of Education:
A Neo-Aristotelian View (London: Routledge, 2020), chap. 8.

128
 Filtering Friendship through Phronesis

other ethical virtues – say, compassion and justice.43 Friendship as a

natural virtue may be compatible with your helping the friend for an
unjust or foolish cause. However, from the perspective of phronesis-
guided virtue, the critical dimensions of the virtue of friendship are
not determined by the architectonic of that particular virtue only
but also by the demands of other virtues. All these different require-
ments need to be synthesised through phronesis, and although that
synthesis may appear to proceed fairly quickly and reliably in the
case of an experienced moral agent, to get things right the agent
still needs to apply the filter of phronesis correctly to the concrete situ-
ation. There is no room in Aristotelian theory for a phronetic decision
that is, in principle, unfiltered.
Someone like Stocker44 could still argue that because of the essen-
tial motivational unity of the phronetic agent in Aristotle’s virtue
ethics, this procedure does not involve the same sort of schizophrenia
as in the other moral theories and, hence, some (or perhaps all) of the
thrust of Williams’s argument can be averted. This consideration is,
however, bound to touch even the greatest of Aristotelian aficionados
for if there is any psychological claim in Aristotle that seems to jar
with common intuitions and empirical evidence, it is the one about
the motivational unity of the phronimoi being such that they never ex-
perience regret.45 Even those contemporary virtue ethicists who go
furthest in sticking to the Aristotelian script, such as Hursthouse,
admit that there are tragic situations from which even the most virtu-
ous agents cannot escape with their lives unmarred.46 Our ethical
outlook as a whole, as well as individual moral virtues such as friend-
ship, need indeed, as Hursthouse points out, to be validated over and
over again, ‘plank by plank’,47 by appealing to the blueprint of the
good life as it comes into confrontation with complex life situations.
This is bound to be a painful process, although perhaps not ‘patho-
logical’ in the strict sense that Stocker’s term ‘schizophrenia’ may in-
dicate.48 The pain cannot be averted by avoiding to apply the filter of
43
Cf. A. W. Müller, ‘Aristotle’s Conception of Ethical and Natural
Virtue’, in J. Szaif and M. Lutz-Bachmann (eds.), Was ist das für den
Menschen Gute?/What is Good for a Human Being? (Berlin: De Gruyter,
2004): 18–53.
44
Op. cit. note 23.
45
Op. cit. note 1, 246 [1166a27].
46
R. Hursthouse, On Virtue Ethics (Oxford: Oxford University Press,
1999), 74; cf. also. D. Carr, ‘Virtue, Mixed Emotions and Moral
Ambivalence’, Philosophy 84 (2009): 31–46.
47
Hursthouse, ibid., 165.
48
Cf. op. cit. note 24.

129
Kristján Kristjánsson

phronesis and simply relying on some raw ‘intuitions’. By blocking

out the ‘extra thought’ needed to remain and to continue to
develop as a virtuous agent – and a trusted friend – one proceeds
to trivialise the message handed down to us by Aristotelian virtue
ethics.

4. Is the ‘extra thought’ in virtue ethics also ‘one thought too

many’?

The preceding section demonstrated the futility of the assumption

that, because of the alleged facility of virtue ethics to motivate
correct friendship-instantiating moral action directly without the me-
diation of a theoretical filter, there is something unique about virtue
ethics as a moral theory that provides immunity from Williams’s one-
thought-too-many argument. It turned out that there is no such
direct motivation in (Aristotle-inspired) virtue ethics. However,
sensing potential defeat at this juncture, it is still possible for the
virtue ethicist to argue that although there is an ‘extra thought’ encap-
sulated by the phronesis filter, it does not involve – as opposed to the
deontological or consequentialist filters – one thought too many.
What could there be unique about the phronesis filter that would
leave it untouched by the argument that hits at the other two
filters? I explore four possible responses below. The first response
builds on the thought that phronesis does not offer a filter in the
same sense as the utility calculus or the categorical imperative
because phronesis is just about methods of implementation, not
content. The other three responses suggest that although the phron-
esis filter is essentially of the same kind as two competing ones, in-
forming the content of moral decision making (e.g. in the case of
deep friendship, on which Williams fastens), there is something
about its content that escapes or at least mitigates the charge of one
thought too many.
(1) ‘The phronesis filter is only about means-end reasoning; it does
not impose substantive one-thought-too-many constraints on the
moral content of the decision, as do the other two filters; and no
moral theory can conceivably work, in any case, without practical
advice on how to find the best means to actualising what has been
decided.’
This response turns Aristotle-inspired virtue ethicists into pure
Humeans about moral motivation. For pure Humeans, reason is ir-
relevant to the choice of ultimate ends, which is based on non-delib-
erative desires only. This thesis seems to follow naturally if we take at
130
 Filtering Friendship through Phronesis

face value Aristotle’s repeated claims about phronesis only constitut-

ing reasoning about means to ends, not about the ends themselves
that seem rather to be formed non-deliberately through the cultiva-
tion of (habituated) virtue.49 There are other places in Aristotle’s
texts that do challenge this understanding, however, and there is
good reason to take those seriously, because they read as more accur-
ate elaborations, or even corrections, of the general thesis about
phronesis being concerned with means only. There we are told not
only that non-intellectual habituation is insufficient for full virtue,
but that full virtue requires a decision to choose virtue for itself,
and that decision requires phronesis. So, although phronetic virtue
grasps the right ends because the virtuous person has the right
desires, those desires require phronesis for their creation precisely in
order to count as the right desires in the first place.50 In other
words, the transition from habituated to phronetic virtue is one of
essence: the previously non-intellectually founded desires become
deliberative desires, and they are no longer the same desires as
before, simply dressed up in fancy intellectual clothes, but rather
new desires, created by phronesis. Hence, I agree with Irwin that
Aristotle cannot be categorised as a Humean with respect to
Hume’s thesis that all practical thought depends on non-deliberative
desires.51 Insofar as Response (1) is meant to defend virtue ethics, in-
cluding the intellectual virtue of phronesis, and insofar as phronesis is
an essentially Aristotelian concept, Response (1) fails to show that
phronesis, in virtue of its proposed exclusive instrumentality, does
not incorporate one thought too many – simply because it is not ex-
clusively instrumental.
(2) ‘What makes consequentialists import one thought too many
into their thinking about deep friendship is their assumption that
the only intrinsically valuable good in the world is pleasure, and
that non-intrinsically valuable goods such as friendship need to be
measured against it. Similarly, for deontologists, the ultimate moral
motivation is derived from respect for the categorical imperative
rather than from an intrinsically valuable virtue such as friendship.
In contrast, for virtue ethicists, virtues such as friendship are intrin-
sically valuable.’
In Kantian deontology, the moral motivation to pursue friendship
(or any other virtue) does not have its source in the emotional

49
Op. cit. note 1, 168, 171–172 [1144a6–9, 1145a4–6].
50
Ibid., 39–40, 168–169 [1105a28–32, 1144a13–22].
51
T. H. Irwin, ‘Aristotle on Reason, Desire, and Virtue’, Journal of
Philosophy 72 (1975): 567–578, 571.

131
Kristján Kristjánsson

component of the relevant virtue, as in Aristotle-inspired virtue

ethics, but in principles of practical reason, encapsulated by the cat-
egorical imperative. If the substantive problem identified by
Williams’s argument is that the crucial moral motivation is derived
from the filter rather than the original source of moral concern
(here the friendship), then it is indeed true that virtue ethics
escapes his charge. However, I do not think the same applies,
mutatis mutandis, to virtue ethics versus sophisticated forms of conse-
quentialism. Badhwar Kapur complains that ‘consequentialist tele-
ology defines intrinsic value in morally neutral terms and morality
as a means to intrinsic value’.52 While that is, strictly speaking,
true, the implications are not as radical as Badhwar Kapur makes
them out to be. For Mill,53 for example, virtues such as justice (and
arguably friendship) constitute essential goods that are parts of the
sole source of intrinsic value, happiness as pleasure and the absence
of pain, rather than just being instrumentally conducive to it.
Those are goods that ought to be valued, whether or not we happen
to value them or not, and also goods whose moral value remains
intact even in the rare cases when they are outweighed by other,
more salient, essential goods.54 Thus, for the happiness pluralist
Mill, the motivation to help a friend would derive from the essential
goodness of friendship, rather than from the utility calculus itself.
That calculus is only necessary because there are cases where the mo-
tivations of different essential goods clash.
There is, in fact, not much to choose between (Millian) consequen-
tialism and virtue ethics here. To love a friend ‘as an end’, Badhwar
Kapur says, ‘is to place a special value on her – to believe that her
value is not outweighed, say, simply by the greater needs of others
or the needs of a greater number of others’.55 But this only holds if
friendship is the sole intrinsic value in her axiology56 or if she consid-
ers friendship as a master virtue, like Cicero who insisted that ‘you
should place friendship above all other human concerns’.57 But
those are clear departures from Aristotle and from most contempor-
ary forms of virtue ethics, according to which the intrinsic value of
friendship can in principle be overridden by another competing

52
Op. cit. note 2, 503.
53
J. S. Mill, Utilitarianism (Oxford: Oxford University Press, 1998).
54
K. Kristjánsson, Justice and Desert-Based Emotions (Aldershot:
Ashgate/Routledge, 2006), 146–148.
55
Op. cit. note 2, 484.
56
Op. cit. note 8, 157.
57
Cicero op. cit. note 25, 35.

132
 Filtering Friendship through Phronesis

source of intrinsic value. Phronesis can be defined as excellence in

moral deliberation precisely because of its capacity to adjudicate cor-
rectly in cases where two such sources seem to clash. In that particular
sense, it serves the same purpose as the utility calculus. So if there is
something about the content of the intellectual filter that separates
virtue ethics from consequentialism and protects it against the
charge of one thought too many, Response (2) has not identified
what that unique content is.
(3) ‘Phronesis allows for preferential treatment but the other filters
force us to treat all persons equally. It is the psychologically impos-
sible requirement of non-preferential treatment that makes the
other two filters, but not phronesis, succumb to the error of one
thought too many.’
It is helpful to remember that while the great medieval Christian
thinkers retrieved, accommodated and ‘infused’ most of Aristotle’s
basic tenets about the virtues, they remained sceptical of any insights
that implied elitism, favouritism or differential treatment based on
people’s allegedly unequal claims to moral worth. Christianity is,
after all, at its core an egalitarian moral system, within which one is
meant to love all one’s ‘neighbours’ equally through agape rather
than through the favouritism-tainted lens of philia. This assumption
was then subsumed within the two great secular systems of morality
that developed during the Enlightenment. Is this perhaps what
makes their two filters import one thought too many, on Williams’s
understanding (as suggested e.g. by Bernstein58)? After all,
Aristotle says, in contrast, that it is ‘more shocking’ to ‘rob a compan-
ion of money than to rob a fellow-citizen’ and to ‘fail to help a brother
than a stranger’.59
While it is true that impersonality and impartiality are foreign to
the spirit of Aristotelian virtue ethics, agent-relativity is in no way a
random, subjective variable in Aristotle. In addition to the somewhat
pedantic specific advice that Aristotle gives here (about returning
favours to benefactors before favouring a friend and returning debts
to creditors before making loans to friends60), it is crystal clear that
Aristotle’s partiality allowances are meant to be strictly calibrated ac-
cording to demonstrated levels of moral virtue. Firstly, character
friends are chosen precisely because of their ethical excellence, and
they are to be discarded if they turn bad beyond redemption.
Secondly, Aristotle discusses in detail conflicts that arise between
58
Bernstein op. cit. note 12, 67.
59
Op. cit. note 1, 224–225 [1160a4–6]; cf also 256 [1169a18–34].
60
Ibid., 241 [1164b25–1165a4].

133
Kristján Kristjánsson

the claims that a friend can have on us, versus a virtuous non-friend,
and his conclusion is that if the friend is just a utility friend, then her
claims on us are strictly limited, whereas the issue becomes more
complicated if the clash is between the claims of a character friend
and another ‘virtuous man’, for the former is then also, ex hypothesi,
virtuous. The overall conclusion seems to be that this has to be as-
sessed on a case-by-case basis but, in any case, simply ditching
agent-neutrality and giving everything spontaneously to the friend
is what only ‘good-for-nothing people’ do.61
These Aristotelian considerations are far removed from the roman-
ticised view of friendship as based on spontaneous preferential treat-
ment that Bernstein,62 for one, attributes to Williams. There is no
hint in Aristotle of an aestheticised view of friendship as lying
beyond the limits of moral justification (as e.g. espoused by
Nehamas63) and beyond the scope of phronesis. The idea that the
non-reflective partiality of friendship holds the key to how phronesis
escapes the thrust of Williams’s argument seems to be a non-
starter. To Williams’s credit, that idea is not his in the first place,
for he explicitly admits that the notion of an attachment to a particu-
lar person as a psychological integrity-grounding project, in his
sense, protected by a categorical desire, would have appeared ‘mys-
terious or even sinister’ to Aristotle.64
(4) ‘The uniqueness of phronesis as a filter vis-à-vis the utility cal-
culus is that it does not require the maximisation of value. That re-
quirement is the main reason why consequentialism imports one
thought too many. Hence, this criticism does not hit at phronesis-
guided reflections on what friendship requires in particular cases.’
To be sure, although Aristotle places fairly strict moral constraints
on the scope of agent-relativity that virtue ethics affords us, there is
no hint of the idea of maximisation in his ethics. That is not to say
that moral value comes without any quantification, for generally
speaking, ‘the greatest virtues are necessarily those most useful to
others’, such as justice and courage.65 Yet virtue comes to us indi-
viduality-adjusted according to Aristotle. There is no description of
an individual’s virtue repertoire available to us that abstracts from
its instantiation in that person in all her psycho-social uniqueness.

61
Op. cit. note 5, 431–433 [1244a1–19].
62
Bernstein op. cit. note 12.
63
Op. cit. note 10.
64
Op. cit. note 3, 15–16.
65
Aristotle, On Rhetoric, trans. G. A. Kennedy (Oxford: Oxford
University Press, 2007), 76 [1366b3–6].

134
 Filtering Friendship through Phronesis

Some of the well-known things that Aristotle says about the golden
mean of action and emotion may seem to indicate that there is an
ideal imitable agent whose virtue consists in hitting this mean accur-
ately on each occasion. However, on closer inspection, there is no
unique blueprint of the perfectly virtuous person per se to aim for.
For example, emulousness is a virtue for young people whereas
adults do not need to emulate role models. Magnificence and magna-
nimity are virtues for people blessed with unusually abundant mater-
ial resources but not for ordinary folks. Temperance in eating is not
the same for Milo the athlete as for the university professor,
because what is intermediate in virtue is relative to the individual,
‘not in the object’.66 And, from an educational perspective, a
boxing instructor will not ‘impose the same way of fighting on every-
one’.67 There is thus no one best way across individuals to be, say, vir-
tuously generous as opposed to be being stingy or wasteful. It all
depends on your individual circumstance (are you poor or
wealthy?) and your own natural inclination towards either extreme,
away from which you should try to drag yourself – with the help of
friends. When I love the generosity of my friend, I do not love it as
matching well or less well the repeatable generosity of the perfectly
generous person, for there is no such generosity simpliciter. There
is an endless plurality of traits that all make the grade as virtuous gen-
erosity, as instantiated in different persons, and there is no way to
choose between them in ways that satisfies some ideal condition of
virtue-exposure maximisation.
The non-existence of a maximisation requirement (because there is
no common currency to maximise) gives us considerable leverage in
making wise phronetic choices regarding conflicting claims by differ-
ent friends, or friends and (virtuous) non-friends, and some may even
legitimately come down to mere tastes when there is no difference in
demonstrated levels of moral worth.
Out of all the responses canvassed so far, this one comes closest to
explaining the potential advantage of Aristotle-inspired virtue ethics,
in general, and its phronesis filter, in particular, in escaping the thrust
of Williams’s one-thought-too-many argument. Yet I am not con-
vinced that this response would have satisfied Williams himself,
given the radical (and probably accurate) interpretation of his argu-
ment explained in Section 2.

66
Op. cit. note 1, 43 [1106b1–7].
67
Ibid., 295 [1180b9–11].

135
Kristján Kristjánsson

5. Concluding remarks

This article has explored the common assumption that there is some-
thing about Aristotle-inspired virtue ethics that makes it immune to
Williams’s infamous one-thought-too-many argument. I have shown
that the idea that virtue ethics has no filter, and hence imparts no
extra thought into the moral decision process, is untenable. The
reason is the simple one that Mill pointed out to us a long time ago:
‘There exists no moral system under which there do not arise un-
equivocal cases of conflicting obligations’.68 So far is it from being
true that Aristotle suggests an exception to this rule that he explicitly
provides us with a filter, called phronesis, to sort out how to strike a
morally justifiable balance between competing sources of intrinsic
value. To argue that this filter does not import an extra thought, it
would have to be shown that phronesis is an intuition pump, rather
than an intellectual virtue of adjudication and, furthermore, that
the intuition to favour friends trumps all other considerations.
However, neither happens to fit Aristotle’s system. In any event, it
is difficult to imagine Williams asking for blind attachment to friend-
ship, to which any extra thought would be inimical.
To escape Williams’s curse, the only remaining way is to argue that
the extra thought required by virtue ethics is not ‘one thought too
many’. The presumed advantage of virtue ethics must, in other
words, lie in the content of its filter rather than the filter’s non-exist-
ence. The article closed with an attempt to show that, contra deontol-
ogy, the friendship motivation in virtue ethics is derived from the
moral virtue, not the intellectual filter, and, contra consequentialism,
phronesis does not require the maximisation of value. This argument
goes some distance in shielding phronesis from Williams’s complaint.
Yet the work that still remains for phronesis to do, and the way it is
meant to do it, would probably still count as a fetish on Williams’s
understanding, as well as falling under Stocker’s sarcastic description
of decision filters as ‘mental alarm clocks’.69 So there is no way to
avoid a substantive disagreement between a neo-Aristotelian, such
as the present author, and Williams on what exactly counts as one
thought too many.
I happen to agree with Woodcock70 that balancing our broad
ethical obligations with authentic personal motives is bound to
remain a non-trivial psychological challenge for any moderately
68
Op. cit. note 53, 71.
69
Op. cit. note 23, 458.
70
Op. cit. note 24.

136
 Filtering Friendship through Phronesis

demanding moral theory. I salute Aristotle for having tried to offer us

an extra thought to guide our reflections on this challenge, and I
maintain that the phronesis filter does not deserve the sardonic desig-
nation of ‘one thought too many’. Filtering friendship through it is a
morally justifiable, and indeed necessary, enterprise.
KRISTJÁN KRISTJÁNSSON ([email protected]) is Professor of
Character Education and Virtue Ethics at the University of Birmingham, U.K. He
works in Moral Philosophy, Moral Education and Moral Psychology. His latest publica-
tions are Virtuous Emotions (OUP, 2018) and Flourishing as the Aim of Education
(Routledge, 2020).

137