Professional Documents
Culture Documents
Abstract
In this work we introduce a new approach to the study of Semantical
Systems. Traditionally the study of these systems have covered only those
semantical systems whose languages are formal and whose models are rela-
tional structures. Here, both languages and models are classes, moreover
there is no dependency on foundational considerations. The concepts of
theory, worlds, semantical consequence, negation and complete-
ness are presented from a strictly semantical point of view. Some of the
results obtained subsume similar results found in the Theory of Models.
Understanding something means to own a system of criteria to evaluate per-
ceived facts. An evaluation system works as a ¯lter through which concepts
may be generated. Di®erent individuals possess di®erent ¯lters and so per-
ception of reality is something of a subjective nature. Not only do sensorial
observations form an individual's knowledge, but concepts may be apprehended
by other kinds of perception and deductive processes performed by the human
mind. Knowledge is, therefore, the mental appropriation of sensorial and
extra-sensorial perceptions that can be deepened by the exercise of thought and
experimentation. Concepts, in their turn, will in°uence the evaluation process
changing the cultural ¯lter and so our perception of reality.
We can say that for a ¯xed instant of an individual's development an evalua-
tion system is acting for the immediately following perceptions and such percep-
tions change the system itself re-elaborating the evaluation criteria. The steady
existence and functioning of this evaluation system enables the knowledge ac-
quisition process to continue, generating con¯gurations or models of reality.
The impossibility of directly communicating di®erent models by the corre-
sponding sensation causes the necessity for the creation of codes or systems of
meaning. The systems of meaning also result from social acculturation and
their most obvious manifestation is the usage of various forms of language -
spoken, gestual, written, etc. Such languages emerge from social conventions
which associate objects to particular signs.
1
We can say that a signal S stands for an object O as if it indicates the
object itself. This relation is indirect, because its interpretation is mediated
by the concept C that we have of the object. The properties of the objects, the
relations among them and the transformations to which they may be subjected
are also coded by signals.
Ogden and Richards proposed the Meaning Triangle as a representation
of the relations between the objects and their corresponding conceptions in our
minds and the associated signals (symbols).
1 Semantical Systems
Knowledge about a given world, actual or arti¯cial, is formed by the continuous
process of perception and evaluation. This knowleldge can be understood as a
class of all the conceivable con¯gurations of the considered world. Each of these
con¯gurations is called a model. These models can be fuzzy or fragmented, as
our compreehension is limited. The class of models, also fuzzy and fragmented
since we cannot perceive all of its elements, is what we usually call system of
beliefs of an individual regarding a given world. The class of models, because
of its conceptual nature, is situated at vertex C of the meaning triangle.
Regarding a given state of a®airs we may perceive and even conceive dif-
ferent con¯gurations. Each of them has a valuation with respect to a possible
world. This valuation comes from a set of values and a process of evaluation.
By evaluation we mean a judgement given to a con¯guration (perceived, com-
municated or conceived) related to the world. We may think that the new or
recent con¯guration is compared to each of the con¯gurations that are already
considered as models or real world con¯gurations. The set of abstract values
can be quite extensive and imprecise, varying with the nature of knowledge or
the state of a®airs considered. Therefore we may have a de¯nite set of values,
such as: ff air; unf airg or fdef f icient; insuf f icient; regular; good; excellentg,
which however not very precise, are well determinate. The most usual ones
ftrue; f alseg are taken for granted, despite the extreme di±culties of their un-
derstanding. Nevertheless other sets may be fuzzy or even indeterminate.
2
We may note that even for de¯nite sets like those mentioned we may still
question the meaning of each element. The degree of indetermination as well as
the complexity of the elements are directly related to the extent to which reality
is understood. Those values are situated at vertex C of the triangle.
their study to further development, when the meaning associated to them will, in most cases,
be relative to their arithmetical interpretation.
3
The interpretation process is car-
ried out in the same way as the
direct perception of reality, nev-
ertheless the generated con¯gura-
tions are acknowledged as possi-
ble relative to the models we pre-
viously acquired by the process
of acculturation. The use of lan-
guages creates, sometimes, a lack
of contact with reality.
In the process of interpretation, the linguistic units, that is, symbols, words
and sentences are interpreted as objects (real or imaginary, concrete or abstract)
as well as relations or activities among these objects.
Evaluation systems together with languages are called Semantical Systems,
and are made up of 4 components:
L- A language
M- A class of models
V- A set of abstract values
We must note that by using the word function for the evaluation process
we are making a simpli¯cation. We are suposing that the interpretation of a
sentence generates only one con¯guration and that the underlying evaluation
process produces a de¯nite abstract value. The function ' might not be com-
putable, i.e. there may not be an algorithmic way to produce a value for it.
4
p
. Models Mpn = } ( f p1 ; p2 ; ¢ ¢ ¢ ; pn g ). In the case of the propositional system
'S`2 Mpn = f ; ; f p1 g ; f p2 g ; f p1 ; p2 g g, which, together with inclusion, form a
lattice.
In the ¯gure:
M0 =;
M1 = f;g
M2 = f f p1 g g
M3 = f f p2 g g
M4 = f f p1 g; f p2 g g
the remaining ones are obtained by the union of the above, for instance M 5 =
M 0 [ M 1 = f ; ; f p1 g g
. Abstract Values V = f 0 ; 1 g
. Evaluation Function 'pn = de¯ned by:
5
the logical connectives. There are various kinds of knowledge that can be rep-
resented by the propositional semantical systems and actually all the digital
electronics depends greatly upon or at least results from the discipline known as
Boolean Algebra, which is but another syntatical representation of the propo-
sitional semantical systems. The propositional semantical systems have ¯nite
sets of models and therefore there are computable evaluation functions 'pn for
them.
The models are certain mathematical structures like E = < D ; R ; © >, where:
D - Is a set, called the domain of the structure.
R - Is a set of relations on D.
© - Is a set of functions de¯ned on D
6
For a particular ¯rst order language, its interpretation with respect to an ade-
quate structure, is a function I which assigns:
- to all constants a 2 C an element aI 2 D,
- to the functions f 2 F a function f I : Dn ¡! D, and
- to the predicates P 2 P a relation P I µ Dn
7
4 Planes of Meaning
Douglas Ho®stadter, in [Ho®], introduced the concept of Implicosphere as re-
lated to the concept of creativity. Implicosphere can be imagined as a kind
of a fuzzy collection of impressions one has about an object, concept or state,
concrete or abstract. They are like clouds of impressions. This cloud becomes
thicker and more complex as we gain familiarity with such entities. The Im-
plicosphere of a concept includes its conections with other Implicospheres, and
those conections are overlaps between implicospheres.
This is a very important idea because sentences do have meanings. Even
though a sentence can be considered as an assemble of units (words), some of
them carrying no semantical values - without independent meaning- , they might
carry implications which depend more on the context in which they occur. We
could say that sentences, even atoms or just symbols, can generate an in¯nity
of inferences. Implicospheres are, in a way, theories.
These theories are formed through interactions inside a social group and they
are conceptual entities which are developed in a very dynamical process. The in-
terplay between the real plane and the conceptual plane is the main trade of the
Natural and Social Sciences. Mathematicians, Logicians and Philosophers deal
mostly with the interplay between the conceptual and symbolic planes. These
planes are not as determined as shown in the presented pictures. As a matter of
fact, knowledge is structured, and concepts varyes in their level of abstraction.
There is a whole in¯nity of planes, and primitive concepts belong to lower levels
of abstraction, following from them the derived concepts. The separation we
did, in our presentation, in conceptual and symbolic planes is methodological,
and the di®erence resides only in the degree of development of the semantical
system. If we consider the primitive concepts of a theoretical discipline in a
certain level, then derived concepts are in higher conceptual planes which are
in some sense more symbolic, because they need of symbolic mechanisms such
as de¯nitions or derivation to be understood besides the understanding of the
more basic or primitives.
8
We will only analyse the interplay between the conceptual plane (CP) and
the symbolic plane (SP). We can say that a certain state of a®airs is observed
and acquires meaning through a cultural ¯lter of concepts which are, in general,
already represented in a semantical system. Then we assume that in a semantical
system
Cn( S ) = T h( M od( S ) )
In this work we de¯ne functions Mod, Th without any dependence on syntatical
L M
considerations, moreover we relativize both to subclasses of T1; M and W1;L .
In order to understand, abstractly, i.e. outside syntatical considerations, the
goals and contributions of this work as a general framework to the discussion
of mathematical and computational systems we refer the reader to Tarski's work
[Tar1]. There Tarski presents four axioms related to two primitives notions: a
set S of sentences and logical consequence Cn as an operator de¯ned on S.
The ¯rst axiom is about the cardinality of the set S of sentences:
This axiom is intuitively reasonable, given that the set of sentences must
ful¯ll certain requirements of communication. We could be more rigorous by
saying:
0
A1 : The set S is recursively enumerable.
which means that concerning the Deductive Sciences, the languages are formally,
ideally, de¯ned through formal grammars.
Even stronger, would be:
9
00
A1 : The set S is recursive.
which implies the existence of recognition algorithms for the languages of Sci-
ences.
The second axiom refers to the operational character of the function Cn:
A2 : If A µ S then A µ Cn( A) µ S
The third axiom states that once the consequences of a set of sentences are
obtained no further knowledge can be gathered by re-applying the operator Cn:
The last axiom states that all consequences obtained from a set of sentences A
must also be obtainable from ¯nite subsets of A:
S
A4 : If A µ S then Cn( A ) = X 2 F in( A ) Cn( X )
Despite the series of important results derived by Tarski, from the four axioms
above, their meaning has been overlooked, at least with respect to their gener-
ality, in further research. Here we started with the notion of semantical systems
[Carn1] and we shall prove A2 and A3 with the aid of very simple mathematical
tools. By starting with simpler assumptions than the above axioms we shall
attempt to clarify the results obtained by Tarski, as well as some others which
make up the necessary theoretical basis for better evaluating some of the recent
achievements in the ¯eld of Applied Logic.
5 Basic Functions
This section introduces four functions which represent the relationship between
L and M. These functions transpose some concepts which are traditionally
treated in Model Theory to the wider context of a two-valued semantical system.
The ¯rst two functions represent projections from the symbolic plane to the
conceptual plane, i.e. from language to models or interpretations:
TL WM WM
(a) ¢ ¢ ¢ Mod : 2 1;M £ 2 1;L ¡! 2 1;L
TL WM WM
(b) ¢ ¢ ¢ Cont : 2 1;M £ 2 1;L ¡! 2 1;L
10
² The function Cont applied to
a set S of sentences in L and
to a subfamily N of models in
W M returns a subfamily of mod-
1;L
els in N.
² The function Mod applied to a
set S of sentences in L and to a sub-
family N of models in M returns a
subfamily of models in N.
11
² The function Th applied to a set
N of models in W M and to a sub-
1;L
set of K, a dialect K, returns a set
of sentences in K.
² The function Absd applied to a
set N of models in W M and to a
1;L
subset of T L , a dialect Kreturns
1;M
a set of sentences in K.
12
We assume here that if f Xi g is a family of subclasses of a class X then:
T
( a1 ) ¢ ¢ ¢ x 2 Si 2 A Xi i® for all i 2 A ; x 2 Xi
( a2 ) ¢ ¢ ¢ x
S 2 i 2 A Xi i® for some i 2 A ; x 2 Xi
( a3 ) ¢ ¢ ¢ Ti 2 ; Xi = ;
( a4 ) ¢ ¢ ¢ Ti 2 ; Xi = X T T
( a5 ) ¢ ¢ ¢ i 2 A [ B Xi = i 2 A Xi \ i 2 B Xi
L M
Lemma 5.1 If S µ K µ T1; M and N µ W1;L , then:
(a1 ) Mod(S; N) \ Cont(S;N) = ;
(a2 ) Mod(S; N) [ Cont(S;N) = N
(a3 ) Mod(S; N) = N ¡ Cont(S;N)
(b1 ) Th(N;K) \ Absd ( N ; K ) = ;
(b2 ) Th(N;K) [ Absd ( N ; K ) = K
(b3 ) Th(N;K) = K ¡ Absd ( N ; K ) 2
(a1 ) and (b1 ) are independent from ' being a total function. There are im-
portant connections between this lemma and the notions, to be developed, of
consistency and completeness. In the case of the semantical system of ¯rst order
languages, ' is assumed to be a total function, despite the fact that the class of
models is constituted of structures over very abstract domains.
The following lemmas show the anti-monotonicity of Mod and Th and the
monotonicity of Cont and Absd relative to the class S of sentences.
L M
Lemma 5.2 Let S1 ; S2 µ K µ T1; M and N µ W1;L . If S1 µ S2 then
Mod ( S2 ; N ) µ Mod ( S1 ; N ).
M L
Lemma 5.3 Let N1 ; N2 µ N µ W1; L and K µ T1;M . If N1 µ N2 then
Th(N2 ;K) µ Th(N1 ;K). 2
13
If we increase the amount of concep-
tual objects, i.e. models, we can say
less (smaller number of sentences)
about their common properties.
Monotonicity of the functions with respect to the second parameter comes from:
L M
Lemma 5.5 Let S µ K µ T1; M and W µ N µ W1;L , then:
L
. Th(W;K) = Th(W;T1;M) \ K
L
. Absd ( W ; K ) = Absd ( W ; T1; M) \ K
M
. Mod(S; N) = Mod(S; W1; L) \ N
M
. Cont(S;N) = Mod(S; W1; L) \ N 2
6 Preliminary results
In all that follows we will consider only semantical systems in which the evalu-
ation function is not a constant function, the reason for this is given by:
Lemma 6.1 Let S ' ` = < L ; M ; f 0 ; 1 g ; ' > be a bivalued semantical sys-
tem, and ' a constant function, then:
L M
(a) . If for all < ® ; M > 2 T1; M £ W1;L ' ( ® ; M ) = 0 then:
M
(a.1) For all N µ W1; L Mod(®; N) = ;
L
(a.2) For all S µ T1; M Mod(S; N) = ;
L
(a.3) For all K µ T1;M Th(M ;K) = ;
M L
(a.4) For all N µ W1; L and all K µ T1;M Th(N;K) = ;
14
L M
(b) . If for all < ® ; M > 2 T1;M £ W1;L ' ( ® ; M ) = 1 then:
L
(b.1) For all ® 2 T1;M Mod(®; N) = N
L
(b.2) For all S µ T1; M Mod(S; N) = N
L
(b.3) For all K µ T1; M Th(M ;K) = K
M L
(b.4) For all N µ W1;L µ T1;M and all K Th(N;K) = K
Proof: Imediate. 2
So semantical systems for which ' is a constant function are not to be
considered as proper semantical systems, they are pseudo- semantical systems.
M L
Lemma 6.2 Let N1 ,N2 ,W µ N µ W1; L and S1 ; S2 ; S µ T1;M , then:
(a) Mod(S1 [ S2 ; N) = Mod(S1 ; N) \ Mod(S2 ; N)
(b) Mod(S; N1 [ N2 ) = Mod(S; N1 ) [ Mod(S; N2 )
(c) Th(N1 [ N2 ;K) = Th(N1 ;K) \ Th(N2 ;K)
(d) Th(W;K1 [ K2 ) = Th(W;K1 ) [ Th(W;K2 )
Proof: (a),(b),(c) and (d) are imediate consequences of (a1 ), page 13, and lemma
7.5. 2
M L M
Lemma 6.3 If N µ W1; L S ; T µ T1;M , and N1 , N2 µ W1;L then:
(i) Mod ( S [ T ; N ) = ; , Mod ( S ; N ) µ Cont(T ;N)
(ii) Th(N1 [ N2 ;K) = ; , Th(N1 ;K) µ Absd ( N2 ; K )
Proof: from lemma 8.2 (a) and hypothesis (i) Mod ( S [ T ; N ) = ; ,
Mod ( S ; N ) \ Mod(T ; N) = ; , Mod ( S ; N ) µ Cont(T ;N). The proof of (ii)
is the dual of (i). 2
Note: The last step in the proof of (i) depends on the complementarity of
Mod and Cont and so on the admission that ' is a total function. Also, the
above result is apparently strange in face of hypothesis (ii) which seems to be
unreasonable. It indicates a state of a®airs where not even true sentences are
possible. Nevertheless we could admit that the language L of the semantical
system is inadequate for expressing valid sentences for all states of a®airs.
M L
Lemma 6.4 Let W µ N µ W1; L and K µ T1;M , then ® 2 Th(W;K) ,
W µ Mod(®; N)
T
Proof: ® 2 Th(W;K) , ® 2 M 2 W Th(M ;K) , for all M 2 W; ® 2
Th(M ;K) , for all M 2 W '( M ; ® ) = 1 , for all M 2 W; M 2
Mod(®; N) , W µ Mod(®; N) 2
In the same way we can prove:
15
L M
Lemma 6.5 Let S µ K µ T1;M and M 2 N µ W1;L , then M 2 Mod(S; N)
, S µ Th(M ;K). 2
M L
Lemma 6.6 Let W µ N µ W1; L and ® 2K µ T1;M , then ® 2 Absd ( W ; K )
, W \ Cont(®;N) 6 = ;
S
Proof: ® 2 Absd ( W ; K ) , ® 2 M 2 W Absd ( M ; K )
, exists M 2 W such that ® 2 Absd ( M ; K ) , exists M 2 W
such that '( M ; ® ) = 0 , exists M 2 W; and M 2 Cont(®;N) , W \
Cont(®;N) 6 = ; 2
Similarly we can prove:
L M
Lemma 6.7 Let S µ K µ T1; M and N µ W1;L then M 2 Cont(S;N) if and
only if S \ Absd ( M ; K ) 6
= ; 2
7 Semantical Consequence
The concept of consequence is usually associated to the concepts of cause and
e®ect. To account for the occurence of an event A there exists a sequence of
events B1 ; B2 ; ¢ ¢ ¢ ; Bn ; ¢ ¢ ¢, ending with A such that for each Bi there exists
a sequence B1 ; B2 ; ¢ ¢ ¢ ; Bi¡1 that accounts for it. In this way, the events
Bi are like links of a chain, or generally, like nodes of a tree or of a directed
graph. At an arbitrary point of an explanation this graph could be seen as a
hierarchical structure, that is, a subgraph without loops. This intuitive notion of
an explanation could be a model of what is known as the Hypothetical-Deductive
method.
The causality between consecutive nodes is generally not as simple as a con-
ditional expression such as If < premiss > then < conclusion >, but relies
upon a wider context to which the concepts (model level) and the symbols (lan-
guage level) representing the regarded situation belong. Thus the notion of con-
sequence must depend on representational features, such as the language avail-
able for the description of the events, and on abstractional features such as the
models that convey only certain aspects of the overall state of a®airs. Consider-
ing our notion of two-valued semantical system S ' ` = < L ; M ; f 0 ; 1 g ; ' >,
to de¯ne consequence we must be concerned with the dialect K of L, actually
used to describe a certain state of a®airs3 N as part of the overall state of a®airs
M of 'S`.
In the previous section we have seen that Mod(S; N) is a subclass of
N which satis¯es the class S of sentences of a dialect K of L, which means
that Mod(S; N) is the class of all situations where the events represented by
3 In Mathematical Logic the context does not a®ect the notion of consequence. Instead, the
language must have considerable expressive power (¯rst order languages and its extensions)
and a su±ciently comprehensive class of models (relational structure).
16
S occur. Intuitively, a sentence ¯ of K is a consequence of S if whenever all
sentences of S occur then ¯ occurs.
We shall later discuss the variation of CnS regarding the language K and N .
17
Lemma 7.1 S µ CnS(S;N,K)
proof: By the corollary 7.1 we have only to show that for all ®, if ® 2 S then
Mod(S; N) µ Mod(®; N), and this is a direct consequence of de¯nition 5.2. 2
The dual concept of CnS is obtained by:
De¯nition 7.2 CnM(W;K,N) = Mod(Th(W;K); N) 2
Corollary 7.2 M 2 CnM(W;K,N) , Th(W;K) µ Th(M ;K)
proof: Similar to corollary 7.1. 2
18
M
T
Lemma 8.3 CnS(S;W1; L ,K) = W
M
µ W1; L
CnS(S;W,K) 2
proof: Immediate. 2
proof: Immediate. 2
9 Closure Properties
The results to be presented in this section are corollaries of the previous de¯-
nitions and results. Nevertheless, they are essential for the reader to acquire a
geometrical view of the relationships between languages and models
19
Lemma 9.2 Th(W;K) = Th(Mod(Th(W;K); N);K)
(A) says that CnS(S;N,K) is the largest class of sentences K that contain
S and preserves its models, this suggests the following:
CnM(W;K,N) is the largest class of models that contain W and preservs its
theory. This, again suggests the following:
M L M
Lemma 9.5 W1; L is a in T1;M and W1;L .
proof: trivial. 2
M
Corollary 9.3 T0;LM = Th(W1; L
L ;T1;M ) is the smallest theory
20
T
proof: follows immediately from T0;LM = M
W µ W1;
L
Th(W;T1; M) 2
L
L M
Corollary 9.4 T1;M is the largest theory and W1;L is the largest world.
L M L M L M
proof: T1; M = Th(W0;L ;T1;M ) and W1;L = Mod(T0;M ; W1;L ). 2
TL = Th(W M ;T L )
0;M 1;L 1;M
and between the classes:
- W M and T L , given by
0;L 1;M
W M = Mod(T L ; WM )
0;L 1;M 1;L
Despite the facts that, neither every set of sentences S is a theory, nor ev-
ery class of models W is a world, there exists as many theories as there are
M L
subclasses of W1; L and as many worlds as subsets of T1;M . Thus, if in
a semantical system 'S` the cardinality of T1; L
M is º there will be 2
º
M ´
worlds, and if the cardinality of W1;L is ´ there will be 2 theories.
L
Lemma 10.1 For all ° 2 T1; L M
M ; ° 2 F0 i® Mod(°; W1 ) = ;.
21
T
proof: ° 2 F0L i® ° 2 M
M 2 W1
L
Absd ( M ; T1; M
M ) i® for all M 2 W1 M 2
L M M
Absd ( M ; T1; M ) i® for all M 2 W1 '( ° ; M ) = 0 i® for all M 2 W1 M 2
M M M M
Cont(°;W1 ) i® Cont(°;W1 ) = W1 i® Mod(°; W1 ) = ; 2
- W0; N M N M
L = W0;L \ N, W 1 K = W1 \ N
K L
- T0;KN = T0;LM \ K, T 1 M = T1;M \K
11 Negation
In this section we will present several forms of negation. The relevance of
this study is that through abstract de¯nitions we can distinguish each form of
negation from the others. By doing so, we might be able to discuss certain
mathematical concepts regardless of syntatical considerations.
22
As for the propositional semantical systems the above class (a set in this case)
of sentences coincides with the class of sentences logically equivalent to the
p
negation of ®. For instance, in the propositional semantical system 'S`n we have
Mod(®; M p p p p p
n ) = Cont(: ®;M n ), and so NegC ( ® ; M n ; L n ) = f ¯ 2 L n j : ® ´
¯ g. Note that our de¯nition of negation (NegC) is relative to a certain state
of a®airs N, and to a particular dialect K. Therefore, by modifying either
N or K we change the value of NegC, that is, di®erent classes of sentences
are obtained. Our de¯nition satis¯es the following criteria [Ebb] for classical
negation:
For all ¿ and all ' 2 L[ ¿ ] there is a sentence à 2 L[ ¿ ] such that
M od¿L ( Ã ) = Str[¿ ] ¡ M od¿L ( ' )
At ¯rst the relativization of negation seems strange, but we can consider the
sentence p1 ¢ `John is married to Ann' as being the negation of the sentence
p2 ¢ `John is married to Mary' in a world in which Ann and Mary are di®erent
women and monogomy is a certain fact, only one of the two sentences is true.
An important property of the classical negation is that if ¯ is the negation of
® then the class containing both ¯ and ® has no models:
23
Note that if we consider families of models N 1 and N 2 , then: If N 1 µ N 2 then
Mod(®; N 1 ) µ Mod(®; N 2 ) and If N 1 µ N 2 then Cont(®;N 2 ) µ Mod(®; N 1 ) then
we cannot derive uniform variation of NegC either increasing or decreasing.
On the other hand the variation regarding dialects is more predictable as the
following lemma shows.
L
Lemma 11.3 If K 1 ; K 2 µ T1; M then NegC(®; N; K 1 ) µ NegC( ® ; N ; K 2 )
Proof: It follows immediately from the de¯nition of NegC. 2
We can thus conclude that NegC increases monotonically with the dialect
L
K of T1; M . The next result shows that in the semantical systems where
M L
W0; L6
= ; no sentence of the language T1; M admits the classical negation.
L M M L
Lemma 11.4 If ® 2 T1; M and W0;L 6= ; then NegC(®; W1 ; T1; M) = ;
M
T M
Proof: We know that W0; L = SµT L Mod(S; W1 ) then (1) ¢ ¢ ¢ for all S µ
1;M
M
L
T1; M
M
Mod(S; W1 = ;, if we assume that NegC ( ® ; W1
) 6 ; TL 1;M
) 6
= ;,
L M L
then there is a ¯ 2 T1; M such that ¯ 2 NegC ( ® ; W1 ; T1; M ), so by
M
lemma 11.1, Mod(f® ; ¯g; W1 )= ; , which contradicts (1). 2
L M M
Corollary 11.1 If ® 2 T1; M and W0;L 6
= ; then NegC(® ; W1 ; K) = ;
There are reasons related to the concepts of completeness and consistency for
considering these extreme types of semantical systems: such systems have typ-
ical metamathematical behavior. From lemma 11.1 we can see that semantical
M
systems in which W0; L6 = ; are positive semantical systems.
24
The relation between classical negation, false sentences and true sentences in
symmetrical semantical systems is given by:
25
Lemma 12.2 If ¯ 2 NegD ( ® ; N ; K ) then ® 2 NegD ( ¯ ; N ; K )
Proof: ¯ 2 NegD ( ® ; N ; K ) if and only if Cont(®;N) µ Mod(¯; N) if and only
if Cont(¯;N) µ Mod(®; N) if and only if ® 2 NegD ( ¯ ; N ; K ). 2
The following lemma shows that NegD is the theory of the class of models M
that falsi¯es ®, i.e. '( ® ; M ) = 0.
Lemma 12.3 NegD ( ® ; N ; K ) = Th(Cont(®;N);K)
Proof: ¯ 2 NegD ( ® ; N ; K ) if and only if Cont(®;N) µ Mod(¯; N) if and only if
¯ 2 Th(Cont(®;N);K) (according to lemma 8.4). 2
The next lemma shows that the only sentences commom to CnS(®;N,K) and
NegD ( ® ; N ; K ) are the ones in T0;LN .
26
N
Lemma 12.6 If W0; K 6
= ; then for all ® 2 K NegD ( ® ; N ; K ) = ;
Proof: Immediate. 2
27
Lemma 12.13 (1) ¢ ¢ ¢ NegE ( ® ; N ; K ) µ NegED ( ® ; N ; K )
(2) ¢ ¢ ¢ NegD ( ® ; N ; K ) µ NegED ( ® ; N ; K )
Proof: Immediate. 2
The last kind of negation we will consider is very important. It is the semantical
counterpart of Post's negation by non demonstrability.
Proof: Obvious. 2
28
13 Complete Theories
Intuitively a theory T is complete when it fully describes a certain state of
a®airs ( Mod(T; N)). This means that its sentences are su±cient to determine
both the situations which satis¯es it - its models - and the complementary
situations - its empirical content. Therefore if a sentence ® does not belong to
T then if T is a complete theory the models which satisfy ®, Mod(®; N), do
not satisfy T , and so these models are among those which do not satisfy T , i.e.
Mod(®; N) µ Cont(T;N), or equivalentely Mod(T; N) µ Cont(®;N). Formally we
have the following (tentative) de¯nition:
In the case of the ¯rst order semantical system the de¯nition above coincides
with the usual one:
T is complete if f f or every ®; ® 2 T or : ® 2 T
29
M
If W0; L = ; the two de¯nitions
coincide.
30
De¯nition 13.2 W is an elementary class in N relative to K ( EC N
K
) i® for
all M 2 W, T h(M; K) = T h(W; K). 2
N
Lemma 13.4 If W0; N
then W0; N
K 6
=; K is EC K .
N K K
Proof: For any M 2 W0; K T1;N µ Th(M ;K) = T1;N . 2
The models of an elementary class regarding a dialect K are undistiguishable
by this dialect. We say that they are equivalent, or more precisely:
M
De¯nition 13.3 Let M1 ; M2 2 W1; L , we say that M1 is equivalent to M2 in
K, ( M1 ´K M2 ) , we say that M1 is equivalent to M2 in K, ( M1 ´K M2 ) if and
only if Th(M1 ;K) = Th(M2 ;K). 2
Proof: Trivial. 2
The following is a very useful little result:
Lemma 13.6 Let S ' ` be a semantical system, NNµ W1;L and K µ T1;M . Then
M L
N
for every W µ W1;K Th(W;K) = Th(W ¡ W0;K ;K)
'S` is
N
Proof: If symmetrical, as W0; K is empty there is nothing to prove.
µ ¶
T T
In any case Th(W;K) = M 2W
Th(M;K) =
M 2W ¡W
N Th(M;K) \
T 0;K
N
N Th(M;K) = Th(W ¡ W0; K ;K) 2
M 2W
0;K
The next result tells us that a complete theory T has a nice property with
respect to its models Mod(T; N), which is, we can identify for such theories,
N
besides W0; K , other elementary classes relative to K.
M L
Theorem 13.1 Let N µ W1; L and T 2 K µ T1 . T is complete in K and N
N
if and only if Mod(T; N) ¡ W0; N
K is EC K .
31
N N
Mod(T; N) ¡ W0; K Th(M ;K) = Th(Mod(T; N) ¡ W0;K ;K), but then from
N
lemma 13.6 we have to show that for every M 2 Mod(T; N) ¡ W0; K Th(M ;K) =
T . As T µ Th(M ;K), it remains to be shown that Th(M ;K) µ T . Suppose that
N
for some N 2 Mod(T; N) ¡ W0; K Th(M ;K) 6 µ T , then there is ® 2 K such that
(4) ¢ ® 2 Th(M ;K) and (5) ¢ ® 6 2 T . From (5) and from the fact that T is a
N
complete theory in K and N we have that Mod(T; N) \ Mod(®; N) = W0; K.
N
As M 2 Mod(T; N) ¡ W0;K , we have that M 6 2 Mod(®; N), i.e. ® 6 2 Th(M ;K),
which contradicts (1) above. 2
Theorem 15.1 is very important, because it shows that for complete theories T
in a language K , all the properties expressed in K , of a particular model of T
are properties of all models in Mod(T; N). This is the meaning of Mod(T; K) ¡
N N
W0; K being EC K .
The results presented so far are valid for any semantical system S ' `. In the next
section we will study the case of symmetrical semantical systems.
32
Theorem 13.2 Let S ' ` be a symmetrical semantical system, M 2 N µ W1;L ,
M
Lattice < Mp
3 ; v>
33
We can see in the lattice < Mp p
3 ; v > that for each model M 2 M3 the class of models
of the theory of M are all the models above M, including M itself.
So,
(1) ¢ ¢ ¢ Mod(Th(f p1 g;Lp;+ p
3 ); M3 ) = f f p1 g ; f p1 ; p2 g; f p1 ; p3 g ; f p1 ; p2 ; p3 g g
p p;+
(2) ¢ ¢ ¢ Mod(Th(f p2 g;M3 ); L3 ) = f f p2 g ; f p1 ; p2 g; f p2 ; p3 g ; f p1 ; p2 ; p3 g g
Moreover Mod(p2 ; Mp p;+
3 ) = Mod(Th(f p2 g;L3 ); M3 )
p
p;+ p;+
As p2 62 Th(f p1 g;L3 ), we can see that Th(f p1 g;L3 ) is not complete, for
Mod(Th(f p1 g;Lp;+ p p
3 ); M3 ) \ Mod(p2 ; M3 ) = f f p1 ; p2 g ; f p1 ; p2 ; p3 g g
So, the question is: which models in Mp3 have complete theories relative to the language
p;+
Lp;+
3 ? A partial answer to this question for the semantical system 'S`3 is that the
only complete theories are those of the models
f p1 ; p2 g ; f p1 ; p3 g ; f p2 ; p3 g
because above these models there is only the model f p1 ; p2 ; p3 g which is the only
M
one in W0; L for the considered semantical system.
The models in Mp3 are ¯nite. So, one would expect their theories to be complete
because as such they can be completly described by enumerating all of its elements.
p;+
It is surprising that only few of them are complete. The semantical system
p
' `3
S was
obtained from ' `3
S by restriction of the language and consequently of the evaluation
p;+
function
power.
'p3 . The system 'S`3 is very poor as a declarative system as it lacks expressive
p;+
Let us suppose we want to describe the model f p1 g in 'S`3 . A strong, natural
and intuitive candidate would be the sentence p1 of Lp;+
3 ,
but then as we just saw this
sentence has other models besides f p1 g which we intended to describe (completly).
What seemed intuitive and natural is a mistake generated by reasoning in the
language of proposicional semantical systems and strongly restricting its language.
p;+
The sentence p1 ^ p2 ^ p3 , in
3 'S`
, describes completely the top model of the
lattice because this is the only model for that sentence.
p;+
One way to strengthen the expressive power of the semantical system
3
is '`
S
modifying its evaluation function so that it may capture this notion about minimum
model. There are some di±culties in this project because this notion is quite elusive
34
as it depends on a complete description of the universe of discourse. For instance, in
p;+
the present case, we know completely all the models of 'S`3
and the order relation µ,
as presented in the last page. Intentional considerations, such as of minimum model
is out the scope of this work.
N
It is very useful to postulate a representative of the world W0; K , even when it is empty.
Proof: Immediate. 2
The need to postulate a bottom model is partially justi¯ed by the existence of empty
N
models in propositional semantical systems. The world W1; K is usually not a class
of equivalence, so we can not take one of its models as representative of it.
Postulate 2 - ?N , or simply ? is a model which corresponds to W1; N
K , in the sense
K
K N
that Th(?;K) = T0;N Th(W1;K ;K). 2
p
It is important to observe that if we consider the whole language of ' `2 then
S
Th(f p1 g;Lp2 ) 6
µ Th(f p1 ; p2 g;Lp2 )
Thus such models are not comparable regarding the notion of subjacency.
N
The following result shows that < W1; K ; v > is a °at lattice:
4 In the case that W N is empty (symmetrical systems), > is the model for contradictions
0;K
and is used analagously to F in Classical Logic.
35
p+
In the above lattice as well as in the one we presented for the positive system
3 '`
S
we can see that the models that possess complete theories with respect to v are the
biggest models immediately below >. The reason for this is lemma 15.3 that is valid
p+
for any semantical systems. The example '`
S
3
serves only as a motivation for the
analysis of the topology of the positive semantical systems relative to v.
p p+
Note: Notice that the evaluation function of
3 '`
S
is the same of 'S`
3
but the topology
with respect to v changed. What remained the same in the two topologies is that
above all the models sits >, corresponding to the whole language as theory. Also, the
level immediately below the top contains complete theories, both in the symmetrical
p+
case and in the analysed case of 'S`
3
. As for the general case, it is a matter to be
analysed in the study of positive semantical systems.
14 Conclusions
Applications of the results of this work are already under way and we have been
working in two direction:
15 Acknowledgements
I am in debt with several people who dedicated some of their precious time to listening
to my intuitions, sometimes quite mistaken, in the last three years. Among these
people are: Maur¶³cio Kritz, Nizam Omar, and my graduate students at IME who
took my courses on Knowledge Representation during the Academic years of 1988 and
1989, specially Claudia Garcia who also helped me with the English version. Thanks
to them all.
16 References
Carn1 Carnap, R. { Introduction to Symbolic Logic and its Applications { Dover
Publications { 1958
Carn2 Carnap, R. Introduction to Semantics { Havard University Press { 1975
36
Carv1 CARVALHO, R. L. at all - Engenharia do Conhecimento, 2a. EBAI, Tandil-
Argentina, 1987.
Chan1 Chang, C. C. and Keisler, H.J. - Model Theory - Studies in Logic and the
Foundations of Mathematics - North - Holland Publishing Company - 1977.
Co1 Cocchiarella, Nino B. -Philosophical Perspectives on Quanti¯cation in Tense
and Modal Logic - Handbook of Philosophical Logic - vol. II - Gabbay D. and
Guenthner, F. - D. Reidel Publishing Company - 1984.
Curr1 Curry, H.B. - Foudations of Mathematical Logic { Dover Publications, Inc {
1977
Ebb1 Ebbinghaus, H.D. -Extended Logics: The General Framework - Model-Theoretic
Logics -Edited by Barwise/Feferman - Springer-Verlag -1985
Fiad1 Fiadero, J. ,Sernadas, A. and Fiadero, C. - Knowledgebases as Structered
Theories - (To appear in Proc. 8th Conf. on Foundations of Software Technology
and Theoretical Computer Science(Springer Verlag) - 1989.
Haj1 H¶ajek, P. and Havr¶
anek, T. - Mechanizing Hypothesis Formation - Mathemat-
ical Foundation for a General Theory - Springer-Verlag - 1978.
Kunh Kunh, T. { The Structure of Scienti¯c Revolutions { The University of Chicago
Press { 1970
Ho® Ho®stadter, D. - Variations on a Theme as the Crux of Creativity { Metamagical
Themas...{ Bantam Books { 1985
Rus1 RUSSEL, B. - Logical Atomism, in Logic and Knowledge, Capricorn Books,
USA, 1971.
Tar1 Tarski, A. { Fundamental concepts of the methodology of the deductive sciences
{ Logic, semantics, metamathematics { Oxford Press { 1956
Vau1 Vaught, R.L. { Some Aspects of the Theory of Models { The Thirteenth Herbert
Ellsworth Slaught Memorial Paper { Papers in the Foundations of Mathematics
{ Published as a supplement to the American Mathematical Monthly - Volume
80, June-July 1973, Number 6.
37