Aspects of Griceian Reasoning

By JLS
for the GC



Why is the elucidation of the notion of 'reason' important? Well, in Grice 2001, I.1, it just _is_. Grice suggests one good way to start: the notion of 'reason' qua
_verb_, and as evidenced in action.

2. Grice proposes as a 'stalking horse'
for reasoning the occurrence (entertaining, often acceptance) of a chain
of inferences. R reasons from P to C if R think that P & R intends that, in
thinking C, he should be thinking of something which would be the
conclusion of a formally valid argument the premisses of which are a suplementation
of P. In this occurrence, the premiss P (or antecedent sequence) is the
'reason' for the conclusion C.

3. Not all actual reasoning, however, is _good_ reasoning. ("Some is
actually downright appalling"). Hence the proviso: the conclusion is _thought_
by R to be derivable, rather than that it _is_ indeed derivable.
While there is, indeed, _bad_ reasoning, 'reasoning' is a
value-paradigmatic notion (unlike, say, 'climate').

4. 'Incomplete'. Most actual reasoning
is enthymematic.

5. Ee require, to ascribe reasoning to R, that R is trying
to solve _some_ problem. We are reluctant to call 'reasoning' a pointless
sequence of formally valid steps as in "I have 2 hands. If I had 3 more
hands, I would have 5 hands. If I were to have double 5, I'd have 10 hands. If
4 hands were removed, 6 hands would remain. Ergo, it MUST be the case that
I would have 4 more hands than I have now. The goal of resoning -- the
solution of a problem -- gives us a criterion for _successful_ reasoning.
Successful reasoning achieves its goal. Intention pervades reasoning. Merely
_judging_ that there _exists_ a formally valid supplementation does not
_count_ as reasoning. It may be that |- P -> C, but if that's not the object of
R's _intention_, there's no need to call it even a _possible_ reasoning. "We
think of the reasoner as intending the production of C to _be_ the
consequence of P (which clearly is different from R merely _judging_ that C is
derivable from P.

6. There are things logic can't teach. Consider the
philosopher who many months ago undertakes to give a set of 10 lectures. One month
before the starting date, he is asked for the titles of the individual
lectures, and reasons to himself as follows:

"Oh God. It's all a mess. I have piles of material. None of it seems worth
listening to. It isn't in shape. I am in a terrible muddle. IF I give them
the titles I had in mind I'm not sure they will fit what finally
emerges (if anything). Why do I do these things? Why don't I learn? If I
cancel the whole thing, though, my name will be mud. What I'll do is: Give
them those titles I had planned YET ASK for latitude to depart from them if
need be."

This lament illustrates what Grice calls the degree-variant notion of
'reasoning' (as opposed to the 'flat' notion). Reasoning is a gradational
notion.

7. We distinguish between 'rational' and 'reasonable'. While it would not
be 'irrational' to expect my wife to clean my football boots, it might be
_unreasonable_ for me to do son. Those boots, too, were not bought at a very
_rational_ price; only at a very _reasonable_ price. To cheat someone in a
business deal (as such) is not _irrational_ and is not _unreasonable_ It's
just repulsive. And, while it would not be irrational to cheat a man when
you knew you might be found out and as a result lose a valuable client, that
would be a rather unreasonable thing to do. Irrational would be to cheat
him when you knew it was quite likely that you would be found out, and when,
if you were, you would lose your job at a time when employment is very
difficult to obtain. Yielding to a tempting invitation to go out drinking when
you have already decided to spend the evening working on tomorrow's
lecture would be (as such) neither irrational nor unreasonable: just weak and
foolish. To yield to that tmptation when you have _not_ yet decided what do do
but know you ought to get on with that work for tomorrow would yet not be
irrational although it _might_ be unreasonable. Iff you have bungingly got
your firm into a difficulty, and you go and confess the matter to your
boss, your boss might be both rational and reasonable about it. Rational if he
cooly and in a reasoned way tells you what is the best course to take.
Reasonable in that he is not to hard on you. "Reasonable", but not "rational",
is, a _privative_ adjective, an excluder. It's _unreasonable_ the word that
wears the trousers: to be reasonable is to be relatively free from
unreasonableness. The 'rational' is the realm of what Aristotle in _Nichomachean
Ethics_ calls, the ratiocinative soul which is _intrinsically_ rational. The
'reasonable' is the realm of the volitive or desiderative: the appetites
and feelings, the pre- or sub-rational 'soul' -- which is only
_extrinsically_ rational as it heeds to the dictates of the ratiocinative. This
_privative_ notion of reasonableness is beneficial for the characterisation of a
rational being. Akrasia is not the stumbling block in the theory of
rationality, but, quite contrarily, something the possibility of which we must
provide from the very start.

II. 1. Reasons can be justificatory and explanatory.

2. Flat and variable
reason. Grice features five features of _flat_ 'reason': It's not variable,
basic, non-valuational, essential to the idea of rational being, and the
source for degree-variant reason. He lists thirteen manifestations
("excellences") of this degree-variant reason: clear headedness, critical acumen,
thoroughness, tenacity in argument, fleixibility, orderliness, breadth, a
sense of relevance, intellectual caution, nose (intuitiveness), inventiveness,
subtlety, and memory. Some of these are truly specificatory (e.g.
orderliness, critical acumen). Some are just ancillary (memory) -- just as in tennis
a good eyesight is while indispensable is not itself a part of excellence
as having a powerful service is.

3. Grice proposes to focus on what he
calls the 'justificatory' use of the word 'reason'. He distinguishes it from a
purely 'explanatory' use by a number of criteria:

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
| | Distinctive Features |
| |_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|
| Canonical Form |factive|explanatory |causal|mass/count|relative |
|_ _ _ _ _ _ _ _ _|_P _C_ |_ _ _ _ _ _ |_ _ _ |_ _ _ _ _ |_ _ _ _ _|
|i. Justificatory:|No Yes |P justifies C|P ->X | mass | Yes |
|P is a reason for| | | | | |
|x to C. | | | | | |
|_ _ _ _ _ _ _ _ _| _ _ _ | _ _ _ _ _ _ |_ _ _ | _ _ _ _ _| _ _ _ _ |
|i. Explanatory: |Yes Yes|P explains C | P->C | count | No |
|P is a reason | | | | | |
|why C | | | | | |
|_ _ _ _ _ _ _ _ _| _ _ _ | _ _ _ _ _ _ | _ _ _| _ _ _ _ _| _ _ _ _ |


A justificatory 'reason' always allow the qualification "good" or "bad"
(or "little" or even "no" reason). Not so explanatory 'reason' "There was a
bad reason why the bridge collapsed" sounds terrible. A justificatory reason
may become explanatory. This happens when the reasoner _thinks_ that P is
a _justificatory_ reason for C, and acts accordingly. 4. Alethic and
practical reason. Justificatory reasons can be, to
borrow from von Wright, "alethic" or "deontic". With this Grice introduces
Grice's main thesis in the lectures, the Univocality (or Aequivocality, or
Equivocality) Thesis. 'Reason' is univocal in the collocations "alethic"
and "deontic"? His approach to conversational implicature suggests so. Grice
discusses Davidson's arguments in 'How is weakness of the will possible'
viz:
"If the barometer falls, it will probably rain. The barometer falls. Ergo,
ceteris paribus it will probably rain.
-- underlying form: Prob (h, p) = Good (h; a, b). More prob (h; p, q) =
Better (h; a, b).
If, prima facie, Act I would be a lie and Act II would not, P is better
than Q. Act I, but not Act II, would be a lie. Ergo, ceteris paribus, P is
better than Q.
Grice notes one problem with _deontic_ or non-alethic modality: for anyone
else, though, a non-alethic conclusion is _nothing_ that one can draw,
either logically or decently or sensibly. What is a reason for a reasoner R1
to do something may _not_ be a reason for R2. If Tommy has been tormenting
my cat, it's best that I ambush him on his way home from school. Tommy _has_
been tormenting my cat. Principle of Total Evidence. Ergo, I shall ambush
Tommy. 5. Mode operators. Grice introduces the operator "Acc": "It is
reasonable that...". "Acc" is followed by a neustic (or mood operator: .and !)
and a phrastic or radix. The scheme for Procedure Specifiers for Mood
Operators being:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
| mood | sub- | | |
| operator | mood | differential | complement |
|_ _ _ _ _ |_ _ _ _ _ |_ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _|
| |exhibitive| | |
| | !1 | none | |
|imperative|_ _ _ _ _ |_ _ _ _ _ _ _ _ _| wills that p |
| ! |protreptic| | |
| | !2 | wills addreseee | |
|_ _ _ _ _ |_ _ _ _ _ |_ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _|
| |exhibitive| | |
| | .1 | none | |
|indicative|_ _ _ _ _ |_ _ _ _ _ _ _ _ _| judges that p |
| . |protreptic| | |
| | .2 | wills addressee | |
|_ _ _ _ _ |_ _ _ _ _ |_ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _|

"Willing" and "judging" are treated as primitive concepts in a
functionalist theory of psychology ('Method in philosophical psychology'). For the
neustic, Grice draws from the idea of "direction of fit", as
distinguishing alethic and deontic or boulomaic. Perceptual beliefs are
_caused_ by the word (the world affects the doxastic system), but in the case
of the boulomaic subsystem it is the system (the will) which _affects_ the
world. Grice remarks that he does not limit the scope of the phrastic for
!-utterances: "_They_ shall not pass" is a legitimate intentional
(indicative exhibitive) utterance. "The sergeant is to muster the men at dawn" --
uttered by a captain to a lieutenant -- is a legitimate imperative. There is
no need to restrict the realm of the intentional to the first person, or
the imperatives to the second.

6. Conditional and Unconditional Rationality:
relative vs. absolute. Further to his Equivocality thesis, Grice considers
"S must get the Oxford Chair of Pastoral Theology" (i.e. "There is (some)
reason for S to get the Chair"). Unlike 'alethic' modality, there seems to
be a agency-relative necessity. It's either S or _us_ who must do something
about that. How are we to deal with this apparent asymmetry? Grice examines
general introduction and elimination rules a la natural deduction system:
"Derive Nec.P from P" and "Derive P from Nec.P". This rules may be
interpreted as syntactic (= 'provabilitiy') or semantic (= validity or
demonstrability). In the alethic realm, for example, Fermat's conjecture is not provable
yet valid. The apparent asymmetry ("[]p -> p" vs. "ObligatoryP ~-> P")
Grice judges a matter of interpretation only. If R says that R must
eat his hat, R is committed to his saying that he shall eat it. The
agency-relativity of "must" is just _system_-relativity. Consider: "It is
necessary for R1 that R2 becomes his ambassador". We mean: R2's becoming an
ambassador will be advantageous to R1, and that it's R1 who shall bring about
that R1 becomes so. It is as if the introduction rule were: If it is
established in R's system that P, "Nec (relative-to-R) P" is establishable (i.e.
satisfactory-for-R). The elimination rule would read: If it is necessary
(relative-to-R) that R shall eat his hat, 'R shall eat his hat' will be
establishable. A further problem concerns utterances which contain a _double_
agency-relative modality. E.g. "It is necessary for Joe Gumb that the American
public retains an interest in baseball". Here, we must distinguish between
the agent R1 for whom something is a reason and the agent R2 the utterance
is _about_ when it is said what there is a reason for. Grice proposes a
general universalisability thesis to deal with these
complex cases: It is necessary that, if it is necessary for R1 that p
should be the case, then let there be some condition C such that R1 satisfies &
and necessarily for any R2 who satisfies it is also necessary
for R2 that p should be the case".

III. 1. Justificatory reason can become explanatory: If R judges that P is
a justificatory reason for accepting that Q, and if R does accept Q on
account of P, then P _explains_ his accepting Q. One must distinguish three
readings of "Acc", though: i. The Kantian "must" (full, indefeasible,
acceptability). E.g. "A bishop must get fed up with politicians". Associated
conditional: "If x is a bishop, (unreservedly), x will get fed up with
politicians". ii. The weaker "ought" (ceteris paribus acceptability). "To keep his
job, a bishop ought not to show his irritation with politicians". Associated
conditional: "If one is to keep one's job and if one is a bishop then
ceteris paribus, one is not to show one's irritation with politicians". iii.
Degree-Variant acceptability: it is to such-and-such a degree acceptable
that..." There is _another_ distinction
to be made, a progress as it were, from 'technical' (Kant's "rules of
skill"), via (ii) 'prudential', to (iii) 'categorical' modality. 2. A fuller
exposition of the initial idea. Given akrasia, we can't accept that if R
concludes, via practical reasoning, that it is acceptable that R shall go
home', he would _ipso facto_ _will_ to go home. Similarly, if R concludes, via
alethic reasoning, that it is acceptable that it snows, would, _ipso facto_
_judge_ that it snows. We should allow for the conditionality involved here
formally. Grice proposes: Acc (Given that S has a red face, S has high
blood pressure) and this from "Acc (Given that x has a red face, x has high
blood pressure)". S has a red face. Ergo, Acc S has high blood pressure.
While for indicative-doxastic reasoning we qualify the conclusion with the
phrase "with some degree of probability", for imperative-boulomaic reasoning we
use "with some degree of _desirability_: "Let it be that A. It is the case
that B. Ergo, with some degree of desirabiity, let it be that C"). Consider
the following:

It is acceptable that, given that R is to keep dry and that it the the case
that R is such that it rains, that R is to take an umbrella. R is to keep
dry. It rains. Ceteris paribus, it is desirable that R takes an umbrella.

In symbols:

Acc (Given !A & .B, !C). (via Acc (Given !Fx & .Gx, !Hx). !A. .B.
(Principle of Total Evidence). Ergo, it is desirable that !C

Or:
Acc, given that x is to one survive that x is to eat. by Universal
Instantiation, Acc, given that R is to survive that R is to eat. R is to survive.
Ergo, by Det, Acc, R is to eat.
_Ceteris paribus_ generalisations permeate boulomaic reasoning.

Or:
Ceteris Paribus, Acc (given that R1 likes R2, R1 wants R2's company. R1
likes R2. No DEFEATER (e.g. R is ill). Ergo R1 wants R2's company.

Or:

Ceteris Paribus Acc (given that R is to leave USA, & R is an alien, that R
is to obtain a saling permit from the Internal Revenue. R is to leave USA
and is an alien. No Defeater (e.g. R is a close friend of the President and
R arranges a travel in Air Force I). Ergo, R is to obtain a sailing permit.

3.The principle of total evidence. Consider R, the owner a firm which
makes & sells ornaments from sea-shells. Concerned if business will improve
during the coming year, R reasons:

These days, every beachcomber is collecting seashells like mad (so as to
sell them to firms such as mine), so I can get seashells more cheaply. It is
thus likely that given that I will get seashells more cheaply, the business
will improve. Yet, my not easily replaceable craftsmen are getting restive
for higher pay. I accept that given that the craftsmen are restive, the
business will _not_ improve. Ornaments from seashells are all the rage at the
moment, so I may be able to put my prices up though & make more money.
Ergo, it is pretty likely, given that I will get seashells more cheaply, that
my employees are restive, & that everyone is eager to buy sheashell
ornament, my business will improve.

Grice provides a correlative boulomaic version. Now R is head accountant of
a firm in Redwood City (it's accounting time) & gets an invited from his
mother to visit her in Milwaukee. Further, his wife has hada head a bad car
accident and is lying in a hospital in Boise, Idaho. We can represent R's
reasoning as:

Acc (Given that R is to give his mother pleasure & that R is her favourite
son, R is to visit her in Milwaukee next week). Acc (Given that R is to
get ready his firm's accounts -- he's head accountant & it's
accounting time, R is to spend next week in his office). Acc, given that R
is to give his mother pleasure and he's to get ready the firm's accounts,
& that he's the favourite son, & head accountant and it's accounting time,
that R visit his for a long weekend & return to the office on Tuesday. Acc,
given that R is to sustain his wife & she is lying, after with two broken
legs, internal injuries, & much pain, that R is to spend next week in Boise
Idaho. Acc, given that R is i. to give his mother pleasure and ii. get
ready the firm's accounts and iii. sustain his wife, and that iv. R is the
favourite son and v. he's head accountant at accounting time, and vi. R is a
husband with a wife lying in Boise, Idaho, that R is to spend next week in
Boise, Idaho, and telephone the mother and the office daily.

Ergo: Acc (R is to spend next week in Boise, telephoning his mother and
office daily.

A slightly different case concerns a doctor examining a patient: should he
give him electromixosis?

Given that the patient is to be relieved of cephalalgia (symptom: headache)
& he is of blood group G, the patient ought to take an aspirin. Given that
the patient is to be relieved of cepahlalgia _and gasteroplexis_ (symptom:
stomach cramp) and that the patient is of blood group G, the patient ought
to be treated by electromixosis. Given that the patient is to be relieved
of cephalalgia and gasteroplexis and that he is of blood group G & that his
blood has an abnormally high alcohol content, the patient ought to be
given gentle message until his condition changes. R does not find an abnormally
high alcohol content in the patient's blood. Ergo (via the Kantian switch
on the face of unqualified indefeasibility), the patient not just ought,
but _must_ be given electromixosis.

The principle of total evidence here seems to be: If R accepts-at-t an
acceptability conditional C1, the antecedent of which _favours_ to degree d the
consequent of C, and R accepts-at-t the antecedent of C1, and C is
optimal-at-t for R (i.e. after _due_ (or proper -- sic valuational) search by R
for such a further conditional, there seems to be _no_ conditional C such
that R accepts-at-t C2 & its antecedent, & the antencent of C2 is an extension
of the antecedent of C1, & the consequent of C2 is a _rival_ of the
consequent of C1, & the antecedent of C2 favours the consequence of C2 more than
it favours the consequent of C1, R may accept-at-t, to degree d, the
consequent of C1.

4. Satisfactoriness, embedding & mixed mood markers. "Satisfactory" is the
term chosen by Grice to do general duty for both "truth" -- the radical of
an indicative utterance -- and "goodness" -- the
radical of an imperative utterance. It may be possible to define goodness
in terms of truth as follows: It is acceptable that !p if 'It is good that
!p' is true. Satisfactoriness provides general versions for the
truth-conditions of usual operators: "p AND q" is satisfactory if p is
satisfactory AND q is satisfactory. "p OR q" is satisfactory if p is
satisfactory OR q is satisfactory. "IF p, q" is satisfactory if p is not
satisfactory or q is satisfactory. This presents complications with
"mixed-modal" utterances.

'The best is filthy & don't touch it'

is alright, but it's reverse,

'Don't touch the beast & it is filthy'

is dubious. 'Touch the beast & the beast bites you' is not the
_conjunction_ that "and" suggests it is. And, while 'Either he is taking a bath or
leave the bathroom door open' is alright, 'Either leave the bathroom door open
or he is talking a bath' seems less so. Grice proposes that in case of
these utterances, we take the boulomaic operator as having broader scope.
There's also problem with the operation of negation. With doxastic modality
there is really no problem: ".~p" and "~.p" are truth-conditionally equivalent.
With "!" the issue is more complex. It looks as if "~!~A" may read as "you
may (permissive) do A" (one signifies one's refusal to prohibit the
addressee's doing A. Yet another complication is brought by utterances like

"The bicycle is to face north",

which he regards as value-indifferent. None of this complications is
insoluble, though.

IV. 1. Crossing the divide. Inferential relations between alethic &
boulomaic modality. Is boulomaic modality reducible to doxastic modality.
Consider:

"To preserve a youthful complexion, if one has a relatively insensitive
skin, one should smear one's face with peanut butter before
retiring at night".

More formally:

It should be, given that R is to preserve a youthful complexion & that S
has a relatively insensitive skin, that S is to smear the face with peanut
butter before retiring" ("SHOULD
(!p, .q; !r).

Now, the boulomaic acceptability here seems to be based on the flat
doxastic acceptability of "SHOULD (.q, .r; .p)":

"It should be, given that R smears the face with peanut butter skin before
retiring & that R has a relatively insensitive skin, that R preserves a
youthful complexion".

There is one problem with the reducibility thesis: in the boulomaic
version, the phrastic or radix "p" features in the antecedent; in the doxstic
version, it features in the consequent. Some defeater may be so for the former
but not but not for latter.

This, Grice notes, is what concerned Kant in the _Grundlegung_
("Foundations"). Kant's example was:

"It's fully acceptable, given that R is to bisect a
line on an unerring principle, that R is to draw from its extremities two
intersecting arcs".

Such a conditional is vouched for by the analytic claim of geometry:

"If R bisects a line on an unerring principle, R does so as a result of
having drawn from its extremities two intersecting acts".

In its more general terms, we need to explore the basis for

"He who wills the end, wills the indispensable means".

Grice proposes seven steps in the derivation.

Step I:

It is a fundamental law of psychology that, ceteris paribus, for any
creature R, for any P and Q, if R wills P & judges if P, P as a result of Q, R
wills Q.

Step II:

Place this law within the scope of a "willing" operator:
R wills for any P & Q, if R wills P & judges that if P,
P as a result of Q, R wills Q.

Step III:
"wills" turns to "should":

if rational, R will have to block unsatisfactory (literally) attitudes:

R should (qua rational) judge for any P & Q, if it's satisfactory to will
that P & it's satisfactory to judge that if P, P as a result of Q, it's
satisfactory to will that q.

Step IV:

Expliciting mode specifications:

R should (qua rational) judge for any P & Q, if it's satisfactory that !P &
that if it .P, .P only as a result of Q, it's satisfactory that !Q.

Step V:

via (p & q -> r) -> (p -> (q -> r)):

R should (qua rational) judge for any
P & Q, if it's satisfactory that if .P, .P only because Q, it's
satisfactory that, if let it be that P, let it be that Q.

Step VI:

R should (qua rational) judge for any P & Q, if P, P only because p
_yields_ if let
it be that P, let it be that Q.

Step VII:

For any P & Q if P, P only
because Q _yields_ if let it be that P, let it be that Q.

2. 'Counsels of prudence'. Prudential acceptability can be dealt with in a
similar fashion, i.e. as analytical consequences of indicative statements
to the effect that so-and-so is the means to such-and-such, with
the proviso that there is a special end: "let R be happy".

Grice regards that end as what Hume would call a "natural disposition",
i.e. a matter of _natural_ necessity.

One complication is drawn from Austin: one thing is to know what is to be
done (the realm of "wisdom" qua _administrative_ rationality), another how
to do what is to be done (the realm of
_phronesis_ as the executive rationality).

A further complication is brought by ends which are not only _finitely_ non
specific.

I may want a large fierce dog to guard my house, and don't care which kind.

But antecedently _indeterminately_ non specific -- i.e. not yet
deliberated:

I may want a large fierce dog to guard my house, but may have not yet
decided which kind.

Is the boundary between phronesis and wisdom clear-cut?

Consider the case of ever dilligent Secretary.

If a boss empowers his secretary to make determinate the boss's
indeterminately non-specific desires not on behalf of his boss, but as _she_ thinks
best, she just ceases to be a secretary.

She's best described as an administrative assistant.

A further problem is that ends relating to prudence may be
non-propositional (just as an amimal desires food, and not _to eat an
apple_).

Grice explores at this stage an extremme scheme of "Crazy-Bayesy"
deliberation which rests on a mechanical model of the vectoring of forces (with
desires -- or animal spirits -- as such drives).

In the scenario, R has two ends E1 & E2.

Each has a different degree of desirability: d1 and d2.

Now for any action A1 which will realise E1 or E2 -- to this or that degree
-- there is a probability p1 that A1 will realise E1 & a probability p2
that A1 will realise E2.

The desirability of the action, relative to E1 & E2 is a function of the
desirabiity of E1 and E2 & the probability that A1 will realise E1 and E2.
Only if A1 scores higher (in
action-desirability relative to E1) than any alternative action, should R
should choose to do A1.

Provisions should be made for, e.g. long-term (standing) vs. situational
ends.

R may priorise family over business, and his children over Aunt Jemima
(who's been living with R all these years). On a particular occasion, though, R
may priorise Aunt Jemima (to get her
out of one of her tantrums) over taking R's son to see the hippopotami at
the zoo.

While prudential rationality is perhaps _not_ just means-end rationality,
the determination over desires certainly does involve means-end rationality.

V1. Ends, systems of ends, and 'happiness'. Grice proposes to explore
the_inclusive_ end of "counsels of prudence", viz. let R be happy. Grice tries:
to be happy is to be well-off, and this is analytic. Qua
inclusive end, this not only connects with the phenomenon of diminishing
marginal utility, but of _vanishing_ marginal utility. It may be, as Grice
puts it, that that, on occasions, the bucket of happiness is filled, and no
further inpouring of a desirable has any effect on its contents. One feature
to be preserved in the analysis is that what gives rise to the end of
happiness is not, strictly, a _means_ to happiness. Being a inclusive end --
and a _rationally_ inclusive end at that -- it means it's rather, a a set, or
better, harmonic system, of ends, an end for the sake of which certain
desirables are desirable. 2. We need to analyse the contributiveness relation,
i.e. the predicate "... is contributive in way w to ...". Grice proposes:

R wills to do P for the sake of Q just in case R wills to do P because R
judges his doing P as somethting which is contributive in way w to the
realisation of Q, and R _wills_ Q.

This involves all sorts of sub-modalities. Take 'causal'.

The causal element may be an initiating factor

("I stop Jones talking by knocking him cold")

or a sustaining one

("I stop Jones's talking by keeping my hand over his mouth").

Take 'specificatory':

A host's seating someone at this right hand side at dinner may be a
specification of treating him with respect.

Take 'inclusive proper':

R may wish to take a certainly advertised cruise because it includes a
visit to Naples. R may be hospitable to someone today because R desires to be
hospitable to that
someone throughout his visit to R's town.

Perhaps Aristotle's considerations on the practical syllogism may help us
here.

Grice notes two versions of such a syllogism.

R wills to realise Q, he enquires what would lead to Q & decides that
doing P would. So, R comes to will, & do, P.

Or:

R conceives his doing P, enquires what doing P would lead to, sees that it
would lead to Q, which he finds himself willing. So R comes to will, and
do, P.

There are complications, with regard to inclusive ends with _special
circumstances_.

For one, a man wrecked on a desert island where he has thus to spend some
time (fixed scenario, not chosen) decides to study the local flora and
fauna.

A second complication is with reason versus rationalisation:

a man wants to move to Ipswhich, but he decides it's because of the
salubrious climate.

Here the reason (though not, consciously, _his_ reason) why he desires to
move to a salubrious climate is that such a desire will justify the desire
to move to Ipswich.

A third type of case is illustrated by the tyrant who punishes a minister
by conferring to him the organisation of the disposal of the garbage of the
palace. Now, just to frustrate the tyrant's plan to humiliate the minister,
the minister decides to take pride in the discharge of his duty.

Here a higher-order desire is involved: the minister wants to discharge his
office efficiently, "for its own sake" as it were, and he wants to want
this because he wants by so wanting to frustrate the tyrant.

Grice thus notes that "wanting p for the sake of q" covers indeed two
different cases:

R wants P because R judges it to be a means for R, and (the minister case),
R wants P because he judges that _wanting_ P will help to realise Q.

This is important: with the inclusive end, "that R be happy", its
components are not, strictly, the realisation of a specific end but, rather, the
desire for that realisation.

Wanting p for the sake of q, where q is happiness does not strictly require
that R judges P is a means for Q; only that R judges that _wanting_ P will
help realise Q; or, in other words, judging P
is one of a set of items which collectively exhibit the open feature
associated with happiness. 3. "R is to be happy" being a rationally inclusive
end, it follows that happiness is a higher-order desire, i.e. a desire to
have, and satisfy, a set of desires which exemplifies some open feature.
Willing is crucial here. R's _decides_ that certain items are constitutive of
his happiness. One may still wish to list some features which are conducive
to stability and flexibility, which will characteristic of happiness qua
realisation of a system of ends. Such features are: Feasibility: the adopted
system of ends should be workable. Autonomy: or self-sufficiency. It's
better, for example, not to have to rely on government grants. Compatibility or
'harmony' of component ends: i.e. the practical analogue of consistency or
coherence. Comprehensiveness: the practical analogue of completeness.
Supportiveness of component ends: one's devotion to one's wife, for example, may
inspire one to heightened endeavour in one's
business of selling encyclopaedias. Simplicity: related to feasibility.
Agreebleness, understood not as the mere satisfaction of a desire, but as the
idea of an activity being a source of delight, or an attraction
supporting an otherwise not very desirable principle. Grice is aware of
limitations. One may wish to add ideals such as maximal development of one's
natural talents, or the provision of scope for outstanding or
distinctive achievement. Perhaps a more serious problem is the closed
systematicity of the listed features. Except for "agreebleness" they seem all
internal, and thus prone to objections such as those directed to the
coherence theory of truth. It will thus be difficult, in the present scheme, to
decide between the real happy life, and, say, the life of a hermit, a
monomaniacal stamp collector, an unwavering egoist, and a
well-balanced, kindly country gentleman. A way out of this objection
concerns the importation of the notion of value. Rationality is after all, man's
metier of man, the capacity with which the Genitor has endowed us in order
to make us maximally viable in our living condition, i.e. in the widest
manageable range of different environments. Or something.