Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Relevance Logic
Author(s): Heinrich Wansing
Reviewed work(s):
Source: Journal of Philosophical Logic, Vol. 31, No. 6 (Dec., 2002), pp. 591-612
Published by: Springer
Stable URL: http://www.jstor.org/stable/30226775 .
Accessed: 20/11/2012 16:16
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Springer is collaborating with JSTOR to digitize, preserve and extend access to Journal of Philosophical
Logic.
http://www.jstor.org
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
HEINRICH
WANSING
BESTFRIENDS
DIAMONDSAREA PHILOSOPHER'S
The KnowabilityParadoxand Modal EpistemicRelevanceLogic
Receivedin revisedversion16July2002
ABSTRACT.Theknowability
reasoningpattern
paradoxis an instanceof a remarkable
a
of
such
in the courseof whichan occurrenceof the possibility
(actually, pair
patterns),
thediamond,disappears.
Inthepresentpaper,it is pointedouthowtheunwanted
operator,
of the diamondmaybe escaped.Theemphasisis not laidon a discussion
disappearance
of thecontentious
premiseof theknowability
paradox,namelythatall truthsarepossibly
the conclusionis derivedthatall truthsare,in
known,buton how fromthis assumption
the solutionofferedis in thespiritof theconstructivist
attitude
fact,known.Nevertheless,
maintained
defenders
of
the
anti-realist
In
order
to
avoid
the
by
usually
premise.
paradoxical reasoning,a paraconsistent
relevantmodalepistemiclogic with strong
constructive
andshownto be complete.
Thesystemis axiomatized
negationis definedsemantically.
KEYWORDS:constructive
negation,epistemiclogic,knowability
paradox,modallogic,
relevance
paraconsistent
logic,
logic.
1. A DIAMONDLOST
Verificationists
holdthata sentenceA is trueif andonly if it is possible
to verify(prove)A. Fromthis anti-realist
pointof view,it seemsplausible thatif A is true,then A is knowable:if it is possibleto proveA, it
is possibleto knowthatA. Since its firstpublishedpresentation
in [11]
andits rediscoveryas an "unjustlyneglectedlogicalgem"by Hart[16],
the paradoxof knowability1has receivedconsiderableattention,see,
for example,[5,8,20,22,23,27-29,36,38,39,47-52].
The startingpointof
the construction
is the anti-realist
thesisthateverytruesentenceis knowclaimthateverytruesentenceis, in
able, andit leadsto the problematic
oversentencesis suppressed,one
fact, known.2If explicitquantification
obtainsthe followingreadingof the anti-realistthesis in propositional
modalepistemiclogicbasedon classicallogic:
(1)
AD OKA.
Intheliterature,
therearetwoversionsof theknowability
paradox.The
additionalassumptions
of VersionI are
(2)
ki
W1
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
592
HEINRICH
WANSING
of knowledgeoverconjunction),
(distribution
(3)
KA
(knowledgeimpliestruth),and
(4)
--O(KAA KA).
-(AA -- KA).
An applicationof an intuitionistically
unacceptableDeMorganlaw
eventuallygives
(provableby classicalreductioad absurdum)
(6)
A D KA.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITY
PARADOX
593
-,KA
assumption,
temporary
2. OK(AA-KA)
1,(1),
3. O(KAA K-KA) 2, (2),
4. O(KA
3, (3),
K--,KA)
5.
6.
-"O(KA A -'KA)
-n(A A --K A)
(4),
1, 4, 5 and negationintroduction.
--OK(A A - KA)
temporaryassumption,
1, (2),
2,
2,
4, (3),
1, 3, 5, negationintroduction,
6, necessitation.
Version(II)(see,forexample,[11,52]):
1. (A A --,KA) D OK(A A --KA)
2. -'OK(A A -'KA)
3.
A -KA)
-,(A
instantiationin (1),
(7),
1, 2 and modus tollens.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
594
WANSING
HEINRICH
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITYPARADOX
595
it will not
It is well-knownthatforblockingtheunwantedderivations,
sufficeto replaceclassicallogic by intuitionistic
logic. The intuitionisticbe
inclined
to
avoid
theparadoxof knowaanti-realist
ally-minded
might
bilityby rejectingtheDeMorganlaw (orclassicalreductio)usedto derive
the problematic
(6). Accordingto Williamson[48] this may even be the
for
a
to escapefrom the knowabilityparadox.
verificationist
only way
However,the derivationof (5) uses only meansthatareintuitionistically
acceptable,andthe intuitionist,althoughrejectingthe derivationof (6),
wouldhaveto be willingto acceptthederivation
from(5)
of (5). Moreover,
the followingformulais intuitionistically
(understoodintuitionistically)
derivable:
(8)
--KAD --A.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
596
HEINRICH
WANSING
Thisobservation
seemsnotto havebeenmadein theliterature.7
Still,there
remainsa problem.If theconstructive
analogue
K (AA , K A)
(9)
(10)
canbe derived:
1. (AA ~
instantiationin (1),
,KA) D OK(AA iKA)
2. 0OK(AA--KA)
(9),
3. (A A -KA) D '-OK(AA KA) 2, empty implicationintroduction,
introduction.
1, 3, conjunction
4. (10)
~,,O(KAA ~~KA)
of modelssuchthatthese
withpurelystructural
properties
correspond
formulasmayor maynotbe postulatedas axioms,
butalsorelevant.
implicationis notonly constructive
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
PARADOX
KNOWABILITY
597
theneighbourhood
semanticsof classicalmodallogicsto interpret
the
knowledgeoperatorK, see [4],
* Routley'sandMeyer'sternaryrelationalsemanticsto interpretrelevantimplication,see [6,33-35],
* a binaryaccessibilityrelationto interpret
themodaloperators0 and ],
* thedistinctionbetweenverification
andfalsificationclausesto interprettheconstructive,
strongnegation.
Forthe relevanceandmodallogic partof the semantics,we shallfollow
thepresentation
in [6] andthepresentation
in [12,13].Note,however,that
all modalrelevancelogicsconsideredin [12,13]areaxiomatizedusingthe
rule.Whatis novelin thepresentsemanticsis thetreatment
contraposition
of strongnegationinsteadof the *-negationof relevancelogic andan inof the knowledgeoperatorthatallowscapturingformulas(9)
terpretation
and(11).
DEFINITION
1. A modalepistemicNelsonframe(orjust a frame)is
a structurea = (I, R, S+, S-, N+, N-, 0), where:
(1) I is a non-emptyset;
(2) 0 is a non-emptysubsetof I;
(3) R IxlxI x I;
(4) S+ andS- arebinaryrelationson I;
(5) N+ and N- are mappingsfrom I into YP(GP(I)).
Weshallreferto S+ in theverification
clauseforformulasOB anduse Sin theverification
clausefor formulas0B. TherelationsS+ andS- could
their
intuitive
exchange
readings,sinceit does not reallymatterwhether
we thinkof theverification
of OB as thefalsification
of D~ B or we think
of theverification
of OB as thefalsification
of 0~ B, see Definition4.
The componentsof a framemustsatisfya numberof conditions,all
of which are plausible(or at least not implausible)given the intuitive
of thecomponents.
Wedefine
understanding
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
598
*
HEINRICH
WANSING
Theconditionsare:
(identity)
s,
_
(Rstu and s'
(monotony)
(idempotence) Rttt,
(persistence 1) {(u, s) : 3w(u < w & S+ws} c S+,
(persistence2) {(u, s) : 3w(w < u & S-ws} c S-,
(persistence3) s 5 t implies N+(s) C N+(t),
(persistence4) s < t implies N-(s) g N- (t).
=
DEFINITION2. If a = (I, R, S+, S-, N+, N-, 0) is a frame,then M92
(M,v+, v-) is a modalepistemicNelsonmodel(orjusta model)iff
S-ut),
S+tu).
9M2
is called knowledgeconsistentif for every u, w E I and every X C I,
S+uw implies X e N+(w) or X e N-(w);
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITY
PARADOX
basis of informationstate t) and 032,t )-
599
BAC
+B
- BAC
91, t + BvC
B or9, t ~+ C,
- B and9N,t - C,
9A,t
- BvC
9, t
iff 9, t
then9A,s P+ C),
, t H- B D C
9, t H+ OLB
T9,t H- OB
)1, t H+ OB
u H+ B,
9), t =- OB
9, t 1+ KB
9)1,t H- KB
whereIBIB"(theverification
set of B) = {w : w e I and9A,w H+ B}.
Verification
andfalsificationhereareto be understood
in a sensethatallowsforinconsistent
states:a possibleworldt takenas aninformation
state
verifiesformulaB if t supportsthe truthof B (in the sense of providing
theinformation
thatB is definitelytrue),andt falsifiesB if t supportsthe
B
of
the
thatB is definitely
(in senseof providingtheinformation
falsity
No
is
and
false). probabilistic
understandinginvolved, nothingspecificis
assumedaboutthe verificationandfalsificationconditionsof atomicformulas.Theseconditionsmaydependon thesubjectmatterorthediscipline
considered.
An information
stateneedneithersupportthetruthnorsupport
the falsityof a givenformulaB, andit mayalso happenthata stateboth
verifiesandfalsifiessomeB.
If thevalidityof KA D A doesnotemergefromtheverification
clauses
already(butis guaranteed
by condition(t+)),it makessense to postulate
thatKB is verifiedatstatet iff theverification
set of B is amongthesetsof
statesjustifiablybelievedat t (9m,t 1+ KB iff IBl" e N+(t)). Similarly
set of B is amongthe
then,t supportsthefalsityof KB iff theverification
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
600
HEINRICH
WANSING
E N-(t).
- KB iff IBIM"
andfalsificationclausesforformulasKB
Notethatwiththeseverification
a distinctionis drawnbetweenjustifiablybelievingthat ~-Bandjustifiably
disbelievingthatB. Thisis certainlywanted.Otherwise,9)1,t 1+ K-B
iff 92, t V+ ~KB. Butit is conceivable,for example,thatthereis a state
at whicheverybodyalwaysjustifiablydisbelievesthatpersona is happy,
butnobodyeverjustifiablybelievesthata is unhappy.The only available
to the effect thata is not happy,without
evidencemightbe information
to
effect
evidence
the
thata is unhappy.Perhapsthispoor
therebeingany
beingis alwaysneitherhappynorunhappy.9
5. Let 9)X = (I, R, S+, S-, N+, N-, 0, v+, v-) be a
DEFINITION
model.A formulaA is validin 9A1
(in symbols92 i= A) iff foreveryt E 0,
A.
A
entails
B
in
91
for every t E 1, 9)1,t j+ A implies
t
iff
9~2, +
B.
A
valid
with
is
9~2,t j+
respectto a class t of modelsiff A is valid
in every9)3QE . The theoryof a class of models(t is the set of formulas
validwithrespectto t.
6. The logic RN4is the theory(in the languageconsidDEFINITION
of
the
class
ered)
21of all reflexive,Euclidean,knowledgeconsistent,nice,
modalepistemicNelson models satisfying
and positivelysupplemented
condition (t+).
7. TheaxiomaticcalculusHRN4consistsof theadjunction
DEFINITION
rule (A, B/A A B), modusponens, the inferencerules A - B / K A
andA - B/ KA
KB
-K B, theinferencerules
D A
(Al A.--A
An)
RKto (OI
-A. A OAn) D OA
A A.
D(A1 v...
RKoAOA 3 (OAl v
v A,)
-v QAn)
(n > 0);
(n > O);
axiomschemata:
(AO) A D A,
(Al) (A D B) D ((C D A) D (C D B)),
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITY
PARADOX
(A2)
(A3)
(A4)
(A5)
(A6)
(A7)
(A8)
(A9)
(A10)
(All)
(A12)
(A13)
(A14)
(A15)
(A16)
(A17)
(A18)
(A19)
601
(A D (A D B)) D (A D B),
(A D (B D C))D (B D (A D C)),
((A D B) A (A D C)) D (AD (BA C)),
(A A B) D A,
(A A B) D B,
A D (A v B),
B D (A v B),
(A D C) D ((B D C) D ((A v B) D C)),
(A A(B v C)) D ((AA B) v (AA C)),
~-(AD B) - (A A - B),
~-(AA B) - ('-A v -B),
~
-~(A v B) A
(~-A -B),
~-'A - A,
~-OA- 0~A,
-OA
A -El-~A,
-O(KA A ~KA),
A
r-KA),
~.OK(A
DA D A,
(A20)A D OA,
(A21)
(A22)
(A23)
(A24)
OA D OOA,
OOA D OIA,
KA D A,
K(A A B) D (KA A KB).
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
602
HEINRICH
WANSING
Case 1. Suppose91, 5s
+
~,B
w
9), t [+ A D B iff forall u, w, if u < w and9A,u 1=+A, then9, w=+
B. If 9)M,
u k+ A andu < w, thenby (persistence),
9), w + A. Since
A entailsB, 9A,w 1-+ B. Conversely,supposet E 0, 9)1,t + A D B,
and9), s k+ A. Thenfor all u, w, if u < w and 91, u )+ A, then
O
T1,w -=+B. Sinces < s, we have9A,s j+ B.
THEOREM1 (Soundness).I- A implies1=2 A.
Proof.By inductionon the lengthn of proofs.We considerhere,by
way of example,onlytwo subcasesof n = 1.
(A17):
~KA)
9)1,t 1 +
i(KAA
iff 93, t =- O(KAAr ,KA)
02, 1u - (KAA -KA)
iff (Vu E I) S+tu
iff (Vu E I) S+tu = (09, u k- KA or9N, u =- ~KA)
iff (Vu E I) S+tu 4 (9X, u =-KA or 9A, u + KA)
iff (Vu E I) S+tu = (JA19"N-(u) or IAl9x e N+(u)).
~
~OK(AA ^KA)
(Vu e
(Vu e
(Vu e
(Vu E
I) S+tu
I) S+tu
I) S+tu
I) S+tu
=9 t, u - K(AAriKA)
A
e N-(u)
= I(A --KA)j~
A
e
= IIAIn I--KAIT' N-(u)
e N-(u).
n
=A j1Aj~ {v: IA:9"E N-(v)}
The latteris the case because9) is nice. Note thatwe have shownthat
of (A17) and(A18) (alias(11) and(9)) are verifiedat any
instantiations
stateof anyknowledgeconsistent,respectivelynicemodel.
o
Also thecompletenessprooffollowsfamiliarlines.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
603
KNOWABILITY
PARADOX
DEFINITION
8. A set of formulasA is calleda theoryiff
(1) if A E A andB E A, thenA A B E A (closureunderadjunction),
(2) if A E A and F- A D B, then B E A (closureunderprovable
implication).
A theoryA is saidto be primeiff
A, thenA E or Be A.
(3) if AvB
A theoryA is saidto be saturated
if it is primeandregular:
(4) F-AimpliesA E A.
The soundnessof HRN4guarantees(usingclassicalreasoning)the existenceof sets of formulasA, F suchthatA rF. Theexistenceof prime
theoriesis guaranteed
by thePrimeExtensionLemma.
LEMMA3 (PrimeExtension). If A is a theory and A YLF, then there
is a prime theory A' such that A c A' and A' Yr F (and therefore
A'n r = o).
= A,
J An U {An+l}
A+
/},
A'
if
A,, {A,,+I}
otherwise,
r,
ULA.
i<ao
The model '9* = (I*, R*, S+*, S-*, N+*, N-*, 0*,
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
604
HEINRICH
WANSING
Anotherlemmathatis usedis:
LEMMA 6 (Inclusion) (Vt, u E I*) t < u (i.e., (3s E 0*) R*stu) if
t C u.
To verify(persistence3),
(i), OB w, andby (ii) B V s, a contradiction.
Thecaseof (persistence
4) is analogous.Clearly,S+*andS-* arerelations
on I*, butit hasto be shownthatN+* andN-* arewell-definedfunctions.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITY
PARADOX
605
KB.E
Completenessfollows almostimmediatelyfrom the 'CrucialLemma', stating that for every formulaA and every w E I*, 9*, w -+ A iff A E w,
and 9R*,w --- A iff -A e w. To provethe CrucialLemma,another
lemmais used,namely:
LEMMA8. (1) Let s I* and t' = {B : OB e s}. Then(i) t' is a theory
and (ii) thereexists t D t' such that t E I* and A t if A i t'.
(2) Let s E I* and OB E s. Thenthere exists t E I* such that B e t and
for all A, A E t implies OA E s.
observation
is thatforeveryA, AE Cl({B})implies0A e s. If F-B D A,
by RKOwe haveF OB D OA, andsinceOB E s ands is closedunder
provableimplication,OA e s. Since T is non-emptyand closed under
unionsof non-emptychains,by Zorn'sLemma,T has a C-maximalelementt. Clearly,t is closed underadjunctionandprovableimplication.
For primeness suppose D v C E t. Then O(D v C) e s. OD e s or
OC e s, because F-O(D v C) D (OD v OC) and s is closed underprovableimplicationandprime.Suppose0D e s. ButthenCl(t U {D}) e T, a
contradiction
withthe C-maximality
of t in T.
O
LEMMA9 (CrucialLemma). For everyformula A and every w E I*,
(1) 9)*, w
(2) 91*, w
+ A iffA e w, and
A iff -~AEw.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
606
HEINRICH
WANSING
(2) followsusingtheinductionhypotheses.Wehereconsideronlythecase
A = OB.
(1) (=)
iff
iff
iff
only if
92*, w + OB
(3u E I*) S-*wu & 9R1*,u V+ B
(3u E I*) S-*wu & B E u
(3u E I*) {B :B E U} w &B E u
OB w.
=-
OB.Then
with(ii).
It remainsto be shownthatthecanonicalmodelbelongsto 9.
is reflexiveand Euclidean.
LEMMA 10. 9RJ*
Aet =40A es
and (ii)
E=s
LOA
A Eu.
D OAE s
s (A~)
OA E
A E u,
LEMMA 11. Thecanonical model 9)1*is knowledgeconsistent,nice, positivelysupplemented,and satisfies condition (t+).
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITY
PARADOX
607
contradiction.
THEOREM2.
2 A = H-A.
Assume
thatY'A. Thenthereis a saturated
set A notcontaining
Proof
A and9T*,A K+ A. Sincef9* E 2(,we have
A.
o
=t
A three-valued
constructivemodalepistemicrelevancelogic RN3is obtained,if it is requiredthatin everymodelfor everyatomp, v-(p) n
v+(p) - 0.
5. A DIAMONDPRESERVED
We can, finally,returnto the knowabilityparadox.If one follows Version (I) of the paradox,in RN4the reasoningcomesto an end afterfive
steps:
1. A A -KA
temporary
assumption,
2.
3.
4.
5.
OK(AA
~i-KA)
O(KA A KrKA)
O(KAA
r -KA)
0 (KA -KA)
1, (1),
2, (2),
3, (3),
(11).
Thederivation
cannotbe continuedusingnegationintroduction.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
608
HEINRICH
WANSING
in RN4,it breaksdown
If Version(II) of the argumentis represented
afterthreesteps:
1.
A A OKA
(1),
2. -,OK(AA~KA)
(9),
3. (A r KA) D OK(AA ~KA) instantiation
in 1.
Modustollens and contraposition,
both as a rule and as an axiom,fail
to hold in RN4 (and N4 and N3), and therefore--OK(A A^ ,KA) D
to detachtheunwanted
(12)
~,(A A ~~KA).
Intuitionistic
negation
logic hasbeencriticizedforits non-constructive
construcwith
Nelson's
for
reasons
and
working
by manyauthors, many
see
tive logics with strongnegationhavebeenassembledin theliterature;
therein.
and
references
Perhapsthe
[1,2,7,15,18,19,24-26,30-32,37,40-46]
most appealingrouteto constructivelogic with strongnegationfor the
of the
[21] supplementation
previousverificationistis L6pez-Escobar's
and
intuitionistic
of
conjunction,disjunction, impliproof-interpretation
A canonicalproofof the constructive
cationby a disproof-interpretation.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
KNOWABILITY
PARADOX
609
of A andnota proofshowing
negation-A of A is adisproof(orrefutation)
that A impliesabsurdity.A canonicaldisproofof ~-Ais a proofof A.
This gives a naturalinterplaybetweenthe notionsof proofanddisproof.
Moreover,the ideathata proofof the negationof A is a proofrevealing
thatA impliesabsurdityis problematic,
becauseabsurdityhas no proof,
and a proofthat A impliesabsurditywouldrequirea constructionthat
convertsanyproofof A into a non-existingproofof absurdity.
Nelson's
four-valuedlogic is soundwithrespectto the proof-and-disproof
semanand
Nelson's
three-valued
is
sound
under
the
further
tics,
logic
assumption
thatno formulaA is bothprovableanddisprovable.
It is thenclearthat
modustollensmustfail. SupposethatA D B and ~-Bare valid:there
is a construction
convertinganyproofof A into a proofof B andthere
is a disproofof B. This does not guaranteethe existenceof a refutation
of A. Also contraposition
mustfail. Considercontraposition
as an axiom.
Thereis no guaranteethatthereexists a constructionthatconvertsany
construction
anyproofof A into a proofof B into a contransforming
structiontransforming
any disproofof B into a disproofof A. For the
thatconverts
rule,assumethatthereexists a construction
contraposition
of
A
into
a
of
B.
Then
there
is
no
any proof
proof
guaranteethatthere
also existsa construction
thatconvertsanydisproofof B intoa disproofof
A. Thefarewellto negationintroduction,
modustollensandcontraposition
with
the
use
of
relevant
is a farewellto theparadoxical
together
implication
of thediamond.
disappearance
like
mathematical
theoriesin
Curiously, Nelson,alsoFitchinvestigated
whichtrueandfalsestatements
aredefinedsimultaneously
by constructive
means,see [9,10].In fact, Fitch[9] consideredpropositional
logics that
amountto the systemsinvestigatedby Nelsonmerelyuponthe addition
of the conditionsforfalsifyingimplicationsat the stateof evaluation,see
also [32].
6. CONCLUSION
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
610
WANSING
HEINRICH
ACKNOWLEDGEMENT
I amgratefulto StenLindstr6m,
HelgeRiickert,YaroslavShramkoandan
anonymousrefereeforcriticalcommentson earlierversionsof thispaper.
Moreover,a previousversionof thepaperwaspresentedattheuniversities
andUtrecht.I wouldlike to acknowledgeusefulquestions
of Amsterdam
Dick de Jongh,and
and remarksby GiovannaCorsi,Paul Harrenstein,
Maartende Rijke.Thanksalso to Neil Tennantfor makingavailablethe
[39].Thecolleaguesmentionedare,of course,notresponsible
unpublished
fortheviews advocatedin thepresentpaper.
NOTES
1 The knowabilityparadoxis also often called the Fitch paradox.Since Fitch [11], however, attributesthe paradoxto an anonymousreferee, the term 'knowabilityparadox' is,
perhaps,more appropriate.
2 In some presentationsof the knowability paradoxit is assumed that necessarily all
truths are knowable. We shall neglect this complication, since it is not essential for the
derivationsto be scrutinized.
3 For the sake of the argument,assumingthat0 and0 are K-typemodal operatorswould
be enough. However,logical necessity and possibility are usually representedby necessity
and possibility in S5 (= KT5).
4 If the reader is convinced that (1) is also valid, note that (1) corresponds with the
following semantical condition: w E X implies 3t Swt and X e N(t), where S is the
accessibility relationused to interpretthe diamond.
5 The discussion of (1) or its naturallanguage counterpartis really a different matter.
If K is understoodas provabilityin arithmetic,for example, G6del's first incompleteness
theoremshows that for the languageof arithmetic,(1) in untenable.
6 The logic N3 can be seen as a conservativeextension of intuitionisticlogic by strong
negation. However, intuitionisticnegation need not be taken as primitive,because the intuitionisticnegation --A of A can be definedin N3 as AD -~A. Negation introductionand
modus tollens fail with respect to strongnegation.
7 In [51], Williamson considers N3. Williamson discusses a formalization of "Goldbach's Conjecture will never be decided". Its formalization in intuitionistic logic as
-n(KA v K -A) leads to a contradiction,if VersionII of the knowabilityparadoxis accepted
as a proof of -'KA D -"A. Williamsonthen considers '-(KA v K~-A) as a formalization
of "Goldbach'sConjecturewill neverbe decided"in N3. Note thatWilliamson'streatment
of strong negation differs from the development in the present paper also insofar as in
Williamson'ssystem, formulasKA arenot falsifiedat the stateof evaluation.Instead-K A
is taken to mean --A.
8 Also Johansson's minimal logic [17] does not help. If (7) is assumed, modus tollens
must fail, but it holds in minimallogic.
9 Theremay well be otherfalsificationclauses for formulasK B worthyof investigation.
The presentclause keeps belief and disbelief separate,enables capturing(9) by the frame
condition of niceness, and avoids 911,t *+K B iff 9)J,t *+ K B.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
PARADOX
KNOWABILITY
611
REFERENCES
[1] Akama,S.: Curry'sparadoxin contractionlessconstructivelogic, J. Philos. Logic 25
(1996), 135-150.
[2] Almukdad,A. and Nelson, D.: Constructiblefalsity and inexact predicates,J. Symbolic Logic 49 (1984), 231-233.
[3] Bozi6, M. and Dogen, K.: Models for normal modal intuitionistic logics, Studia
Logica 43 (1984), 15-43.
[4] Chellas, B.: Modal Logic. An Introduction,CambridgeUniversityPress, Cambridge,
1980.
[5] Dummett,M.: Victor'serror,Analysis 61 (2001), 1-2.
[6] Dunn, J. M.: Relevancelogic and entailment,in F. Guenthnerand D. Gabbay(eds.),
Handbookof Philosophical Logic, Vol. 3, Reidel, Dordrecht,1986, pp. 117-224.
[7] Dunn, J. M.: Partialityand its dual, StudiaLogica 66 (2000), 5-40.
[8] Edgington,D.: The paradoxof knowability,Mind 94 (1985), 557-568.
[9] Fitch, F.: SymbolicLogic. An Introduction,RonaldPress,New York, 1952.
[10] Fitch, F.: The system CA of combinatorylogic, J. SymbolicLogic 28 (1963), 87-97.
[11] Fitch, F.: A logical analysis of some value concepts, J. SymbolicLogic 28 (1963),
135-142.
[12] Fuhrmann,A.: Relevant logics, modal logics and theory change, Ph.D. thesis,
Departmentof Philosophy,AustralianNational University,Canberra,1988.
[13] Fuhrmann,A.: Models for relevantmodal logics, StudiaLogica 49 (1990), 501-514.
[14] Gettier,E.: Is justified truebelief knowledge?Analysis 23 (1963), 121-123.
[15] Gurevich,Y.: Intuitionisticlogic with strongnegation,StudiaLogica 36 (1977), 4959.
[16] Hart,W.:The epistemology of abstractobjects:Access andinference,Proc. Aristotelean Soc. 53 (1979), 152-165.
[17] Johansson, I.: Der Minimalkalkiil,ein reduzierterintuitionistischerFormalismus,
CompositoMathematicae4 (1936), 119-136.
[18] Kracht,M.: On extensions of intermediatelogics by strongnegation,J. Philos. Logic
27 (1998), 49-73.
[19] von Kutschera,F.: Ein verallgemeinerterWiderlegungsbegrifffiir Gentzenkalkiile,
Arch.fiir Math.Logik Grundlag.12 (1969), 104-118.
[20] Lindstr6m,S.: Situations,truthand knowability:A situation-theoreticanalysis of a
paradoxby Fitch,in E. EjerhedandS. Lindstrim (eds.), Logic,Action,and Cognition
- Essays in Philosophical Logic, Kluwer Academic Publishers,Dordrecht, 1997,
pp. 181-209.
[21] L6pez-Escobar,E.: Refutability and elementary number theory, Indag. Math. 34
(1972), 362-374.
[22] Mackie, J.: Truthand knowability,Analysis40 (1980), 90-92.
[23] Melia, J.: Anti-realismuntouched,Mind 100 (1991), 341-342.
[24] Nelson, D.: Constructiblefalsity,J. SymbolicLogic 14 (1949), 16-26.
[25] Nelson, D.: Negationand separationof conceptsin constructivesystems, in A. Heyting (ed.), Constructivityin Mathematics,North-Holland,Amsterdam,1959, pp. 208225.
[26] Pearce, D.: n Reasons for choosing N, TechnicalReport 14/91, Gruppefiir Logik,
Wissenstheorieund Information,Free Universityof Berlin, 1991.
[27] Percival,P.:Fitch and intuitionisticknowability,Analysis 50 (1990), 182-187.
[28] Rabinowicz,W.: Intuitionistictruth,J. Philos. Logic 14(1985), 191-228.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions
612
HEINRICH
WANSING
[29] Rabinowicz, W. and Segerberg, K.: Actual truth, possible knowledge, Topoi 13
(1994), 101-115.
H.: An Algebraic Approachto Non-classical Logic, North-Holland,AmsRasiowa,
[30]
terdam,1974.
[31] Rautenberg, W.: Klassische und nichtklassische Aussagenlogik, Vieweg Verlag,
Braunschweig,1979.
[32] Routley,R.: Semanticalanalyses of propositionalsystemsof Fitch andNelson, Studia
Logica 33 (1974), 283-298.
[33] Routley, R. and Meyer, R. K.: The semantics of entailment II, J. Philos. Logic 1
(1972), 53-73.
[34] Routley, R. and Meyer, R. K.: The semantics of entailmentIII, J. Philos. Logic 1
(1972), 192-208.
[35] Routley,R. andMeyer,R. K.: The semanticsof entailment,in H. Leblanc(ed.), Truth,
Syntaxand Modality,North-Holland,Amsterdam,1973, pp. 194-243.
[36] Riickert,H.: A solution to Fitch's paradoxof knowability,in D. Gabbay,S. Rahman,
J. TorresandJ. P. vanBendegem (eds.), Logic, Epistemologyand the Unityof Science,
KluwerAcademic Publishers,Dordrecht,2002, to appear.
[37] Shramko,Y., Dunn, M., and Takenaka,T.: The tri-latticeof constructivetruthvalues,
J. Logic Comput.11 (2001), 761-788.
[38] Tennant,N.: The Tamingof the True,ClarendonPress, Oxford, 1997.
[39] Tennant,N.: VictorVanquished,Analysis, to appear.
[40] Thomason,R.: A semanticalstudy of constructiblefalsity, Z. Math. Logik Grundlag.
Math. 15 (1969), 247-257.
[41] Wagner, G.: Logic programming with strong negation and unexact predicates,
J. Logic Comput.1 (1991), 835-859.
[42] Wansing,H.: TheLogic oflnformationStructures,LectureNotes in AI 681, SpringerVerlag,Berlin, 1993.
[43] Wansing,H.: Tarskianstructuredconsequencerelationsand functionalcompleteness,
Math. Logic Quart.41 (1995), 73-92.
[44] Wansing,H.: Semantics-basednonmonotonicinference,NotreDame J. FormalLogic
36 (1995), 44-54.
[45] Wansing,H.: Negation as falsity: A reply to Tennant,in D. Gabbay and H. Wansing
(eds.), Whatis Negation?, Kluwer Academic Publishers,Dordrecht,1999, pp. 223238.
[46] Wansing, H.: Negation, in L. Goble (ed.), The Blackwell Guide to Philosophical
Logic, Basil Blackwell, Cambridge,MA, 2001, pp. 415-436.
[47] Williamson,T.: Intuitionismdisproved?Analysis 42 (1982), 203-207.
[48] Williamson,T.: On the paradoxof knowability,Mind 96 (1987), 256-261.
[49] Williamson,T.: Knowabilityand constructivism,Philos. Quart.38 (1988), 422-432.
[50] Williamson, T.: On intuitionisticmodal epistemic logic, J. Philos. Logic 21 (1992),
63-89.
[51] Williamson,T.: Never say never, Topoi13 (1994), 135-145.
[52] Williamson,T.: Tennanton knowable truth,Ratio 31 (2000), 99-114.
This content downloaded by the authorized user from 192.168.72.226 on Tue, 20 Nov 2012 16:16:23 PM
All use subject to JSTOR Terms and Conditions