You are on page 1of 12

CA~LOS E .

A_mcwo~6~ Maps Between Some Different


Ds165 ~KI_~SON
Kinds of Contraction Function:
The Finite Case

Abstract. In some recent papers, the authors and Peter Ggrdonfors have defined.
and studied two different kinds of formal operation, conceived as possible represen-
tations of the intuitive process of contracting a theory to eliminate a proposition.
These are partial meet contraction (including as limiting cases full meet contraction
and maxiehoice contraction) and safe contraction. It is known, via the representation
theorem for the former, that every safe contraction operation over a theory is a partial
meet contraction over that theory. The purpose of the present paper is to study the
relationship more finely, by seeking an explicit map between the component orderings
involved in each of the two kinds of contraction. It is shown that at least in the finite
case a suitable map exists, with the consequence that the relational, transitively
relational, and antisymmetricaUy relational partial meet contraction functions form
identifiable subclasses of the safe contraction functions, over any theory finite modulo
logical equivalence. In the process of constructing the map, as the composition of
four simple transformations, mediating notions of bottom and top contraction are
introduced. The study of the infinite case remains open.

1. Background
T h e n o t i o n of p a r t i a l m e e t c o n t r a c t i o n was i n t r o d u c e d b y t h e a u t h o r s
a n d P e t e r G ~ r d e n f 0 r s in [2], a n d t h e n o t i o n of safe c o n t r a c t i o n b y t h e
a u t h o r s alone in [3]. T h e special case of p a r t i a l m e e t c o n t r a c t i o n k n o w n
as m a x i c h o i c e c o n t r a c t i o n h a d b e e n s t u d i e d earlier b y t h e a u t h o r s in [1].
T h e r e is also a g e n e r a l s u r v e y of t h e s e n o t i o n s , w i t h o u t proofs, in [r
To a v o i d l e n g t h y r e p e t i t i o n we shall a s s u m e t h a t t h e r e a d e r is f a m i l i a r
w i t h [2] a n d [3] or a t l e a s t w i t h t h e s u r v e y [4] f o r c o n c e p t s a n d m o t i -
v a t i o n , a n d we r e c a l l h e r e o n l y t h e m o s t salient definitions.
L e t A b e a set of p r o p o s i t i o n s , a n d Cn a c o n s e q u e n c e o p e r a t i o n . W e
a s s u m e t h a t Cn i n c l u d e s classical t a u t o l o g i c a l i m p l i c a t i o n , is c o m p a c t ,
a n d satisfies t h e r u l e of " i n t r o d u c t i o n of d i s j u n c t i o n i n t o t h e p r e m i s s e s ' ,
t h a t is, t h a t y e C n ( A w{xlvx~}) w h e n e v e r y e C n ( A w {xl}) a n d y e Cn(A ~J
u{x~}). W e w r i t e A k x as a n a l t e r n a t i v e n o t a t i o n f o r x e Ca(A). A ~ x
is d e f i n e d t o b e t h e set of all m a x i m a l s u b s e t s JP _~ A s u c h t h a t B non ~ x.
A selection function o v e r A is a f u n c t i o n y s u c h t h a t f o r e v e r y p r o p o s i t i o n
x, if A _Lx is n o n - e m p t y t h e n y ( A A_x) is a n o n - e m p t y s u b s e t of A •
~md if A • -----O t h e n 7 (A J_x) = {A}. T h e partia~ meet contraction f u n c t i o n
= d e t e r m i n e d b y y is d e f i n e d b y p u t t i n g A ' - - x = A F ( A .Lx) f o r e v e r y
p r o p o s i t i o n x. As a special case, if r(A ix) is a s i n g l e t o n {B} w h e r e B e A L x
f o r all r a b i e s of: x ~ Ca(O), t h e n t h e selection f u n c t i o n y is called a ~naxi-
choice selection function, a n d t h e p a r t i a l m e e t c o n t r a c t i o n f u n c t i o n -- t h a t
it d e t e r m i n e s is called a maxichoice contraction function.
188 C. E. AZvhourrSn and D. 2dakinson

A selection function $ is called relational iff there is a relation ~ over


[.Jx{A _kx} such t h a t for all x r Cn(O), ~,(A [.Lx) consists of j u s t those
elements of A_l.x t h a t " d o m i n a t e " A j x under ~ , i.e. ? ( A _ L x ) = {B
e A _kx: B ' ~ B for all B ' c A _kx}. E q u i v a l e n t l y a n d more conveniently
when A is a t h e o r y , i.e. when A -= Cn(A): iff there is a relation ~ over
U~ = ( . J x ~ { A _Lx} such t h a t for all x ~ A , x ~ Cn(O), the same identity,
which we call the "marking off" identity, holds.
On the other hand, the concept of safe contraction is defined as follows.
A hierarchy over A is a relation < over A t h a t is non-circular in t h e sense
t h a t for no al, ..., a n c A (n ~ 1), a~ < a~ < ... < a n < a~; a n d the safe
contraction function -- over A determined b y a hierarchy < over A is
defined b y p u t t i n g A - - x ~- A n C n (A/x). t t e r e A / x is the set of all elements
a c A t h a t are safe with respect to x, in the sense t h a t a is not a minimal
element (under < ) of a n y minimal subset (under set inclusion) of A t h a t
implies x. W h e n A is a t h e o r y this reduces to A : x = Cn(A[x).
N o w we k n o w f r o m [3] Observation 3.2 t h a t e v e r y safe contraction
function over ~ set A of propositions satisfies the GiLrdenfors postulates
(--1) to (--6) for contraction; and we Mso k n o w from the first l~epresen-
ration Theorem 2.5 of [2] t h a t when A is a theory, then a n y operation
satisfying those postulates is a partial m e e t contraction function over A.
Thus, as n o t e d in Corollary 3.3 of [3], e v e r y safe contraction function
over a t h e o r y is a partial m e e t contraction function over t h a t theory.
Our present purpose is to s t u d y more directly a n d deeply t h e inter-
connection b e t w e e n the two kinds of contraction, b y constructing an
explicit m a p b e t w e e n the relations involved, in the finite ease. ~u give
constructions in detail, b u t a b b r e v i a t e and sometimes omit the more
routine verifications.

2.1 Dominance and Maximality

The first step in constructing the desired m a p is to reformulate the


definition of the relationality of a selection function in terms of m a x i m a l i t y
rather t h a n dominance, as follows: a selection function ~ over a t h e o r y A
is relational iff there is s o m e relation < over U~ such t h a t for all x e A ,
x ~ C n ( O ) , we have ~ ( A i x ) = { B c A J x : for no B ' c A A _ x , B < B ' } .
The equivalence o f this to the original definition of relationMity is
easily verified using the m a p f l t h a t takes a relation ~ over U~ to the
c o m p l e m e n t of its converse, t h a t is, p u t t i n g B ' < B iff B ~ B'. Clearly
f~ is a one-one transformation of the set of all relations over U~ onto
itself; and clearly we have t h a t { B c A . L x : for all B ' cA_l_x, B ' ~ B }
= {B c A _l_x: for no B ' c A _kx, B < B'}, where < is f l (~)-
Maps between some different kinds... 189

_N-ow suppose f r o m this point onwards to t h e end of the p a p e r t h a t


t h e t h e o r y A is finite modulo Cn, t h a t is, t h a t there are only finitely m a n y
elements a e A m u t u a l l y non-equivalent under t h e relation of logical
equivalence a ~ a ' defined to hold iff Cn(a) -----Cn(a'). Suppose t h a t 7
is a relational selection function over A, with < a relation over U~ such
t h a t for all ~ c A , x r w e h a v e 7 ( ^ _ I x ) = {B e A • for no B '
e A .~x, B < B'}. Then it is n o t difficult to show t h a t < m u s t b e non-
-circular over UA, t h a t is, for no B1, ..., B , e U~ (n~> 1) do we h a v e
BI<B,<...<B.<B1. F o r suppose t h a t B ~ , . . . , B ~ e U ~ . Now by
finiteness it is easy to show t h a t for each i ~ n there is a b~ e A w i t h A •
---- {B~} -- take for example b/ to be the disjunction of all (up to logical
equivalence) the a e A such t h a t B, non ~ a, and use t h e fact t h a t for each
such a, A ~ Gn(B~u{a}). A n d A I_!_b~A ... ^ b n = A • ... u A l b , b y
Observation 4.1 of [2], so A / b ~ A ... Ab, ----{B1, . . . , Bn} , so since 7 is
a selection function, 7(A • ... ^ b,) is a n o n - e m p t y subset of {B~, . . . , Bn}
a n d so contains some B,. H e n c e Be is maximal in {B~, ..., B,} u n d e r < ,
i.e. B, ~: Bj for all j ~ n a n d so in particular B, ~ B~+~ (reading i + 1 to
be 1 if i = n) as desired.
Thus when A is a t h e o r y finite modulo Cn, we k n o w t h a t if a function
7 is a relational selection function over A, then there is a non-circular
relation < over U~ such t h a t for all x r A, x r On(O), 7(A • consists
of precisely those elements of A • t h a t are maximal under < , i.e. 7(A •
={Be_4• for no B ' e A • B<B'}.
Moreover, again under the assumption t h a t A is t h e o r y finite modulo
Ca, and supposing f u r t h e r t h a t 7(A • {A} in the limiting cases
t h a t x e A or x e Cn(J3)~ we also have t h e converse implication. F o r if
7(A • = {A} in the limiting cases, a n d there is such a relation < over
U~, then using t h e non-circularity of < a n d the finiteness of A • we
can easily verify t h a t 7(A • is n o n - e m p t y whenever A • is non-empty,
so t h a t 7 is indeed a selection function, and so, as shown in t h e first para-
graph of this section, is a relational selection function.
P u t t i n g these two implications together we can therefore say: the
relational partial meet contraction functions over a t h e o r y A finite modulo
Cn are just the functions -- such t h a t A - - x = A in the limiting cases
t h a t x r A o r x e Cn(O), and such t h a t there is a non-circular relation <
over U~ with A - - x = N { B e A • for no B ' e A • B<B'}, or
more briefly expressed, A - - x = ^ m a x ( ^ • for all x c A , x r
W e note for later use t h a t there is a systematic correspondence b e t w e e n
properties of ~< and properties of f~ (~<) = < . I n particular: ~< is reflexive
(over U~) iff < is irreflexive; ~< is connected iff < is a s y m m e t r i c ; ~< is
transitive iff < is virtually connected in the sense of [3], i.e. for all B , B',
. B " e Ua, if B < B' t h e n either B < B " or B " < B'; a n d ~< is antisym-
metric (which suffices to render its partial meet contraction function
190 ~. E. Alehourr6n and D. Makinsou

a maxichoice contraction function) iff < is connected up to i d e n t i t y (i.e.


either B < B" or B ' < B or B = B ' for all B , B ' e y ~ ) .

2.2 Bottom Contraction


W h e n A is finite modulo Cn, t h e elements of U,~ can of course be iden-
tified, up to equivalence modulo Cn, with t h e atoms of A. Clearly then,
there is a n a t u r a l one-one correspondence f2 between the class of a:ll non-
,circular relations < over UA a n d the class of all normal non-circular
relations over t h e set • of all atoms of A. Here a relation < is said to
be norma~ over a subset A ' of A iff it respects logical equivalence over
t h a t set, i.e. iff for all a, a', b e A ' , if C n ( a ) = Cn(a') t h e n a < b iff a' < b
a n d likewise b < a ifI b < a'. I t is thus easy to verify t h a t an operation
-- is a relational partial meet contraeti0n f u n c t i o n over a t h e o r y A finite
modulo Cn iff there is a normal non-circular relution over the atoms of A
such t h a t A - - x = A i n t h e limiting cases t h a t x : ~ A or x e Cn(O), and
such t h a t in the remaining (principal) cases A - - x = C n ( V m a x L ( x ) ) .
Here A_(x) is the set of all atoms k of A such t h a t k non b x, a n d m a x •
is t h u s t h e set of all atoms of A t h a t are m a x i m a l a m o n g the atoms t h a t
do n o t i m p l y x, This set will be infinite, b u t as it is finite modulo Cn,
disjunction o v e r i t m a y b e u n d e r s t o o d as a disjunction o f arbitrarily
chosen representatives of its finitely m a n y cells:
We call such an operation --, for a n o r m a l a n d non-circular relation <
over A_~, the bottom contraction over A determined b y < . Thus, cumu-
lating our results so far, we k n o w t h a t the relational partial m e e t contraction
functions over a t h e o r y A finite modulo Cn are j u s t the b o t t o m contraction
f u n c t i o n s over A.

2.3 Top Contraction

I t appears to be difficult to construct a map directly f r o m b o t t o m


contraction to safe contraction. So we proceed via a dual notion of top
contraction.
Clearly, the function ~ on A = Cn(ao) defined b y p u t t i n g ~-,a ---- a0v -Ta
sets up a one-one correspondence (up to logical equivalence) between
the atoms a n d co-atoms of A. Given a relation < over the atoms, we
define the relation f s ( < ) ---- < ' over the set TA of co-at0ms b y p u t t i n g
s < ' t iif ~ t ~ ,-~s. Clearly f3 establishes a one.one correspondence between
t h e class of all normal relations over I__~ a n d the c!uss of all normal rela-
t i ( n s over T~, a n d moreover, a normal relution < over / ~ is non-cir-
c u l r iff < ' over T.t is likewise so.
Maps between some different kinds... 191

Nor is the m a p fs a merely formal t r a n s f o r m a t i o n ; it has an epistemo-


logical rationale. F o r suppose t h a t the relation < is read as indicating
degree of security, as suggested in the discussion in Section 4 of [3]. 3fore
specifically, suppose t h a t we u n d e r s t a n d b < e as indicating t h a t in
the circumstance t h a t A as a whole is regarded as n o t entirely tru% b
is more open to doubt, or more in danger of falsehood, t h a n c, or in other
words t h a t c is more secure t h a n b. :Now of all the situations or "possible
worlds" t h a t fail to render A as a whole true~ t h a t is which render a o
false, those which m a k e an element a e A false are precisely those which
make its relative negation H a = a0v -la true, as is e v i d e n t from the
classical t r u t h conditions for disjunction a n d negation. The danger of
a being false~ given the falsehood of a0~ is thus the same as the suscepti-
bility of H a being true. Consequently, when b < e so t h a t b is more in
danger of being false t h a n c, t h e n --~b is more susceptible of being true
t h a n is w e , which is to say t h a t H e is more in danger of falsehood t h a n
~ b . Similarly for the converse. Thus if < over I__A is understood as indi-
cating degree of security given the falsehood of ao, t h e n the relation
< ' = f ~ ( < ) over T~ also indicates degree of security in the same circum-
stance.
I t is not difficult to verify~ using s t a n d a r d boolean considerations a n d
the n o r m a l i t y of the relations < , f a ( < ) involved, t h a t C n ( V m a x Z ( x ) )
= C n ( T ~ - - m i n T ( x ) ) whenever x ~Cn(12I) a n d x c A , where T(x) is t h e
set of all co-atoms t e T~ such t h a t x F t, a n d where ma~rimality on the
left means m a x i m a l i t y u n d e r < whilst m i n i m a l i t y on t h e right is u n d e r
<' =A(<)-
Hence we can say t h a t an operation -- is a b o t t o m contraction o~.er A
iff there is a normal a n d non-circular relation over T~ such t h a t A--'x = A
in the limiting cases t h a t x ~ A or x E Cn (O), a n d such t h a t A --x = Cn (TA
--minT(x)) in the r e m a i n i n g (principal) cases. We call such an operation
--~ for a n o r m a l a n d non-circular relation < over T~, the top contraction
function over A determined b y < . The b o t t o m contraction functions are
thus j u s t t h e top contraction functions. Cumulating our results so far~
we can say t h a t the relational partial m e e t contraction functions over
a t h e o r y A finite modulo ~ are j u s t the top contraction functions ever A.

2.4 Regular Safe Contraction


The passage f r o m top contraction to safe contraction is r a t h e r different
f r o m the t r a n s f o r m a t i o n s of the p r e v i o u s sections. This is essentially
because a finite t h e o r y A with n m u t u a l l y non-equivalent co-atoms will
have 2 ~ non-equivalent elements, a n d so instead of a one-one correspon-
deuce between relations over the f o r m e r set a n d relations over t h e l a t t e r
set, we shall c o n s t r u c t a slightly more complex functional connection.
192 O. E. Alchour~Sn and D. ~Iakinson

Using the terminology of [3], we say t h a t a relation < over A continues


up ~ over A iff whenever a~ b, c e A a n d a < b whilst b ~ c t h e n a < c.
Analogously we say t h a t < continues down ~ over A iff whenever a, b~ c
e A a n d a F b whilst b < v t h e n a < c. A relation over A is said to be
regular over A iff it continues both up a n d down ~ over A. I t is easy to
check t h a t when < is regular over A t h e n it is n o r m a l over A
I n this section we wish to show t h a t for a t h e o r y A finite modulo Cn,
a n operation - - i s a top contraction f u n c t i o n over A iff it is a safe contrac-
t i o n function d e t e r m i n e d b y some regular hierarchy over A. To do this
we make use of two maps, f4 a n d g. W h e n < is a relation over T ~ , we
define f 4 ( < ) to be the relation < ' over A such t h a t a < ' b iff there is some
s e T(a) such t h a t for all t e T(b), s < t. 'When < is a relation over A ,
we define g ( < ) to be simply the restriction of < to T~.
I t is easy to check t h a t when < is a normal a n d non-circular relation
over T.t, t h e n f~(<) is also normal a n d non-circular over A , a n d also
continues up a n d down b over A, a n d so is a regular hierarchy over A.
1VIoreovcr it is clear t h a t when < is normal over T~ t h e n f 4 ( < ) is a conser-
v a t i v e extension of < , t h a t is, g ( f 4 ( < ) ) = <. I n view of this, to prove
t h e desired equivalence it will suffice to verify more specifically the follow-
ing:
L E p t A . Let A be a theory finite modulo Cn and < a regular hierarchy
ever A . Then an operation -- is the safe contraction, function determined
by < i f f -- is the top contraction function determined by g ( < ) .
To see t h a t this l e m m a will suffice, note first t h a t if -- is a safe contrac-
t i o n f u n c t i o n d e t e r m i n e d b y a regular hierarchy < over A, t h e n b y the
] e m m a -- is t h e top contraction f u n c t i o n d e t e r m i n e d b y t h e (normal
~nd non-circular) relation g ( < ) over A. Conversely, suppose t h a t : is
top contraction f u n c t i o n over A, a n d so d e t e r m i n e d b y a n o r m a l a n d
non-circular relation <----g(f4(<)) over T~. Then f 4 ( < ) is a regular
h i e r a r c h y over A, and so b y the lamina again, -- is the safe contraction
f u n c t i o n d e t e r m i n e d b y f4 (<).
P~ooP oF T ~ . L v . ~ . L e t < be a regular hierarchy over A. Then
its restriction g ( < ) to T~ is normal a n d non-circular, so we need only
show t h a t C n ( A / x ) = C n ( T ~ - - m i n T ( x ) ) for all x c A , x r Cn(O), where
A / x on the left is determined b y < a n d m i n T ( x ) on the right is determined
b y g ( < ) , which for simplicity we henceforth also write as < . B y boolean
considerations it will suffice to shaw t h a t L H S n T A = t t H S n T ~ , a n d
clearly since < is normal over T ~ , I ~ H S r ~ T ~ - ~ T ~ - - m i n T ( x ) . Thus
we need only show t h a t C n ( A / x ) r ~ T ~ -~ T A - - m i n T ( x ) whenever x e A,

Suppose t e C n ( A / x ) n T ~ . If t diT(x) t h e n t o T , t - - m i n T ( x ) as desi-


r e d , so we m a y suppose t e T(x). Since t is a co-atom of A a n d t e C n ( A / x ) ,
_/]laps between some different I~inds... 193

t h e n b y a boolean p r o p e r t y of co-atoms, there is an a r with a F t.


B u t since < is regular over A a n d so continues up F over A, we can infer
b y L e m m a ~.1 of [3] t h a t t e A / x , so t h a t t is n o t ~ m i n i m a l element of
a n y minimal subset B _ A with B F x. Now t h e relation ___ of logical
equivalence u n d e r Cn partitions T ( x ) into cells; let B be a choice set
containing j u s t one s f r o m each of these cells, with t e T ( x ) chosen as
the representative of its celt. I t is easy to see t h a t B is a minimal subset
of A t h a t implies x, so since t e B, ~ve can say t h a t t is n o t m i n i m a l i n / L
H e n c e since B c Y(x)~ t is n o t minimal in T ( x ) , so t e T A - - m i n T ( x ) as
desired.
F o r the converse, suppose t h a t t e T ~ - - m i n T ( x ) . Then t is a co-atom
of A a n d either x non F t or t h e r e is an s e T(x) w i t h s < t. I n t h e case t h a t
x non F t, it is n o t difficult to show ~hat since t is a co-atom of A, t is n o t
a n e l e m e n t bf a n y m i n i m a l B _~ A with B I-x, so t h a t t e A / x a n d so
t eCn(A./x)r~:T~ as desired. Suppose t h e n x I-t, so t h a t there is an
a ~ T(x) with s < t. Again, we show t h a t t e A / x a n d ~hus t e C n ( A / x ) n T ~ .
L e t B be a n y minimal subset of A with B F x a n d t e B ; it. will suffice to
show t h a t t is n o t m i n i m a l in B. Since B F x F s a n d s is a co-~tom of A,
t h e r e is u b e B with b F s. B u t since < is r e g ~ a r over A, i t continues
d o w n i- over A, so since b 1- s < t we h a v e b < t, a n d so t is n o t minima,]
in B . . a s desired. []

3. Conclusions

P u t t i n g t o g e t h e r the chain of equivalences established in Section 2


we t h u s h a v e :

Tm~O~n~ 1. I'or any theory A fi~..ite modulo Cn, the following condi-
tions are equivalent:
(1) -- is a relational partial meet contraction function over A ;
(2) -- is a bottom contraction function over A ;
(3) "-- is a top contraction function over A ;
(4) "-- is a regular safe contraction function over A ;

Whilst the principal interest of conditions (2) a n d (3) lies in the finite
case, conditions (1) a n d (4) are of interest whether or n o t A is finite modulo
Cn. H o w e v e r it remains a n open question w h e t h e r (1) implies (4), or
conversely, in t h e infinite case.
We recall in this connection t h a t several conditions on a selection
f u n c t i o n ~ t h a t are equivalent to t h e relationality of ~, a n d so imply con-
dition (1) above for t h e partial m e e t contraction -- d e t e r m i n e d b y ~ are
given in [2] Observation 4.10 a n d its final note a d d e d in proof.
194 O. E. Alchourr6n and D. Makinson

T~:EO~gE~ 2. Por any theory A finite modulo Cn, the following condi-
tions are equivalent:
(1) -- is a transitively relational partial meet contraction function over A ;
(2) -- is a virtually connected bottom contraction function over A ;
(3) -- is a virtually connected top contraction function over A ;
(4) -- is a safe contraction function determined by a regular hierarchy
that is virtually connected over A.
P~ooP. The meaning of (2) should be clear: -- is a b o t t o m contrac-
tion function determined a (normal and non-circular) relation over _Lx
t h a t is also virtually connected over _kx. Similarly for (3).
W e need only make some minor additions to t h e verification of Theorem
1. Specifically, as a l r e a d y n o t e d in Section 2.1~ ~ relation ~< over Ux is
transitive iff f~(~<) over Ux a n d thus also f 2 ( f l ( < ) ) over -1-4 is virtually
connected; this establishes the equivalence of (1) and (2). I t is easy to
check t h a t a n y normal relation < over I__x is virtually connected iff
f a ( < ) over Ta is virtually connected, which establishes ( 2 ) ~ ( 3 ) . Clea~rly,
if < over A is virtually connected over all of A t h e n it is virtually connected
over T x , sO t h a t g ( < ) is virtually connected over /~x, which establishes
(~) ~ (3). Finally, to show t h a t ( 3 ) ~ ( 4 ) we need only check t h a t w h e n
a normal relation < = g(fd(<)) over T~ is v i r t u a l l y c o n n e c t e d over T~
t h e n < ' = f d ( < ) is virtually connected over all of A. F o r this, suppose
a < ' b b u t a ~:' e. F r o m t h e former, there is an s e T(a) such t h a t s < t
for all t e T(b). F r o m the latter, there is a u e T(c) with s ~: u. Since
< is v i r t u a l l y connected over T~, s < t and s ~: u gives us u < t for all t
e T(b), so t h a t c < ' b as desired. []
~ o t e t h a t p a r t of Theorem 2, n a m e l y t h e implication ( 4 ) ~ ( 1 ) , is
also k n o w n to hold quite generally for theories A finite or infinite. F o r if
(4) holds, t h e n as shown in Corollary 6.3 of [3], -- satisfies the G~rdenfors
" s u p p l e m e n t a r y p o s t u l a t e s " (--7) and (--8), and so b y the second represen-
tation T h e o r e m 4.4 of [2], -- is a transitively relational partial m e e t
contraction function. H o w e v e r it remains an open question whether
t h e converse implication ( 1 ) ~ ( 4 ) also holds in the infinite case.
Several f u r t h e r conditions on a partial meet contraction function
t h a t are equivalent to (1), in b o t h the finite and the infinite cases, ~re
given in [2], n o t a b l y in Corollary 4.5 a n d Observation 6.5.
Following the terminology~of [3], we say t h a t a relation < over A is
quotient connected over a set B _~ A of propositions iff for all b, c e B either
b<cor e < b or Cn(b) = C n ( e ) . A s n o t e d in l~emark 6.1 ( a ) of [3], if
< is a normal hierarchy over B and is quotient connected over B, t h e n
it is also transitive over B. This concept is useful for our third theorem.
T~EO~E~ 3. ~or any theory A finite modulo Cn, the following condi-
tions are equivalent:
Ma2~s between some different kinds... 195

(la) "-- is a relational maxichoice contraction function over A ;


( lb ) -- is an antisymmetrically relational partial meet contraction function
over A ;
(2) -- is a quotient connected bottom contraction function over A ;
(3) "-- is a quotient connected top contraction function over A ;
(4) -- is a safe contraction function determined by a regular hierarchy
that is quotient connected over A .

P n o o F . W e need to m a k e minor additions to the proof of T h e o r e m 1~


a.nd also replace the m a p f4 b y a more refined m a p f~ in order to verify
the implication (3) ~ (4).
Clearly (lb) implies (la) even in the infinite ease. F o r suppose B , B'
e y(A A_x). Then b y the marking off identity we have b o t h B ' ~ B and
B ~ B', so t h a t a n t i s y m m e t r y implies B ----B', so t h a t ? ( A Ax) is a singleton
s~s desired. The converse implication holds in the finite ease. 1%r let
be a selection function over a t h e o r y A finite modulo Cn, such t h a t
Js relational under ~ , and let -- b e the partial m e e t contraction function
t h a t it determines. ]3y Observation 5.2 of [2], ~ is connected and so
also reflexive over U~. L e t B , B ' e U~ a n d suppose b o t h B ' ~ B and
B ~ B ' ; we need to show B-----B'. Since A is finite modulo Cn, there
is an m c A , x 6 Cn(O), with A_km = {B, B'}. Since B ' ~ B and B ~ B '
a.nd ~ is reflexive over U~, we have b y the marking off i d e n t i t y t h a t
? ( A . . k x ) = {B, B'}. H e n c e if y is a maxiehoiee selection function, so
t h a t y ( A A x ) is a singleton, we have B = B ' as desired.
As a l r e a d y n o t e d in Section 2.1, it is easy to check t h a t ~ over U~
is a n t i s y m m e t r i c over U~ iff f~(~) is connected up to i d e n t i t y over U,~,
a n d clearly this in t u r n holds iff f~(fx(<)) is quotient c o n n e c t e d over
• I t is also easy to check t h a t a normal relation < over _k~ is quotient
connected over _k~ iff fs ( < ) is quotient connected over T~. This establishes
(]b)r Clearly using the lemma of Section 2.4, (4) ~ (3). I t remains
ttlerefore to verify (3) ~ (4).
Suppose (3) holds~ i.e. -- is a t o p contraction function determined
b y a normal and non-circular relation < over T A t h a t is quotient connected
over T~. Clearly this condition implies t h e transitivity of < over T~,
as a l r e a d y noted. Define f ~ ( < ) = < ' b y p u t t i n g a < ' b, for a n y a, b e A,
iff there is an s e T ( a ) - - T ( b ) such t h a t for all ~ e T ( b ) - - T ( a ) , s < t.
An equivalent formulation of t h e definition, useful in some of the
verifications, is as follows. • t h a t from the conditions imposed on
< it follows t h a t e v e r y n o n : e m p t y subset S of Y~ ha~s a least element up
to logical equivalence, i,e. an s e S such t h a t for all s' e S either s < s'
or Cn(s) -----Cn(s'), and moreover this least element of S is nniqne up to
logical e q u i w l e n e e . I t is then easy t o verify t h a t the s h o v e definition
o f f * ( < ) is equivalent to the following: a <' b iff T ( a ) + T ( b ) is n o n - e m p t y
196 C. E. Alchourrdn and D. ~1abinso~

and all of its least (up to logical equivalence) elements are in T(a). Equi-
v a l e n t l y again~ given t h e n o r m a l i t y of < over T~, iff T ( a ) + Y ( b ) is non-
e m p t y and at least one of its least Cup to logical equivalence) elements
is in T(a). H e r e -~ is s y m m e t r i c differenee~ i.e. T ( a ) W T ( b ) = ( T ( a ) - -
- - T ( b ) ) w ( T ( b ) - - T ( a ) ) ; note t h a t r ( a ) - ~ r ( b ) = 0 iff C n ( a ) = C n ( b ) .
Clearly f* ( < ) is b o t h normal and a s y m m e t r i c over A. Using t h e transiti-
v i t y of < over T~, we can verify t h a t f~ ( < ) = < ' is transitive over A. F o r
suppose t h a t a < ' b , b < ' c . Then t h e r e is an s e T ( a ) - - T ( b ) such t h a t
for all u e T ( b ) - - T ( a ) , s < u, and likewise there is a t e T ( b ) - - T ( c ) such
t h a t for all u e T(c) - - T ( b ) , t < u. To show t h a t a < ' c~ it will suffice to
find an r e T ( a ) - - T ( c ) such t h a t for all u e T ( c ) - - T ( a ) ~ r < u. Define r
b y p u t t i n g r = s in t h e case t h a t b o t h s ~ T(c) and either t ~ T ( a ) or
there is a w e ( T ( b ) n T ( c ) ) - - T ( a ) with w < t ~ a n d p u t t i n g r = t other-
wise. Then it is easy to verify tha~ r has the desired properties.
Transitivity and a s y m m e t r y of ] * ( < ) give us non-circularity, and
so we can say t h a t ]~* (< ) is indeed a hierarchy over A. Clearly from t h e
last of the equivalent definitions~ f ]*(< ) is quotient connected over A.
Moreover it is straightforward to verify t h a t it continues b o t h up and
down ~ over A, and so is regular over A. Finally~ it is clear t h a t f * ( < )
is a conservative extension of t h e normal < over T~, i.e. g(f~( * < )) = < .
Consequently b y the lemma of Section 2.47 the t o p contraction f u n c t i o n
-- over A determined b y < is identical with t h e safe contraction function
determined b y t h e regular quotient connected hierarchy 9e~(<); so t h a t
condition (4) holds as desired. []
The function f~ defined in the proof of Theorem 3 has a n u m b e r of
interesting features. I n t e r m s of the epistemclogical reading of < , t h e
definition of ]4( * < ) is a v e r y n a t u r a l one. F o r it a m o u n t s to saying t h a t
if a, b are elements of A~ and s is t h e w o r s t co-utom (under t h e given
ordering < of the co-atoms) under which t h e y differ, t h e n a < b or b < a
according as this worst co-atom is implied b y a or b y b. E l e m e n t s of A
are thus c o m p a r e d b y the w o r s t co-atoms on which t h e y differ.
Formally, it is clear t h a t the relation f~ ( < ) includes the relation f4(<).
F o r if a stands in the latter relation to b, then there is an s e T ( a ) with
s < t for all t e T(b), so b y the irreflexivity of < , s ~ r ( b ) so t h a t s e T ( a ) --
- - T ( b ) and s < t for all t e T ( b ) - - T ( a ) ~ T(b). The relation f ~ ( < ) also
has the p r o p e r t y of being strictly order preserving, i.e. when a F b b u t
Cn(a) r Cn(b) t h e n a < ' b. F o r if a b b b u t Cn(a) V: Cn(b) then T(b)
is a proper subset of T(a)~ so T ( a ) - - T ( b ) is n o n - e m p t y whilst T ( b ) - - T ( a )
is empty~ and so v a c u o u s l y a < 'b. Finally, J ~ ( < ) = < ' has the i~verti-
bility p r o p e r t y t h a t for all a, b e A, a < ' b iff ~ b < ' ~ a , where ~-~ is t h e
relative negation operation defined in Section 2.3 b y the equation No
~ - a o v No. F o r s u p p o s e t h a t a < ' b ; then t h e r e is an s e T ( a ) - - T ( b )
such t h a t for all t e T ( b ) - - T ( a ) ~ s < t. Since a ~ s a n d b non ~ s~ we h a v e
Maps between some different kinds... 197

~..a n o n ~ s a n d ~ b k s, so s e T ( ~ b ) --T(.--a). Likewise t e T ( H a ) - - : T ( ~ b )


implies t e T ( b ) - - T ( a ) , so ~ b < -~a. The converse is similar.
Given these features of the m a p f : , it would be pleasant to be able
to use it in t h e proof of T h e o r e m 1, taking it to be defined in the first
of t h e three ways mentioned, since in the more general c o n t e x t where
quotient c o n n e c t i v i t y is no longer assumed, this is no longer equivalent
to t h e o t h e r two. H o w e v e r , it is n o t clear whether this is possible, for
in this more general c o n t e x t there is no longer a n y a p p a r e n t w a y of verifying
the non-eirculaxity of f ~ ( < ) . I t remains an open question whether f * ( < )
is non-circular, for an a r b i t r a r y normal non-circular relation < over T~.
W e r e m a r k t h a t t h e implication ( 4 ) ~ ( l a ) of T h e o r e m 3 can also
be established u n d e r the same hypotheses b y a v a r i a n t a r g u m e n t t h a t
uses no maps. F o r suppose condition (4) holds. Then -- satisfies the Giirden-
fors s u p p l e m e n t a r y postulates (--7) and (--8), as shown b y [3J Obser-
vation 6.2 and l~emark 6.1 (a), and so -- is a transitively relationM partial
m e e t contraction function, b y t h e representation Theorem 4.4 of [2].
B u t since A is finite modulo Cn and condition (~) holds, we also k n o w
from Observation 6.4 of [3] t h a t -- is a maxichoice contraction function.
So p u t t i n g these together (as we may, for if -- is determined b y a selection
function t h a t is a maxichoice selection function, then it is determined
b y a unique selection function) we k n o w theft --" is a relational maxi-
choice contraction function a n d (la) holds as desired.
H o w e v e r it remains an open question w h e t h e r either ( 4 ) ~ (la) or its
converse holds in the infinite case.
Several f u r t h e r conditions which, for a r b i t r a r y partial m e e t contraction
functions~ are equivalent to (la) in b o t h the finite a n d infinite cases are
given in Observation 6.3 of [2]. These include the "decomposition condi-
tion" t h a t A - - ( x ^ y ) = A - - x or A - - ( x ^ y ) = A - - y for all x, y. W h e n
the domain under consideration is restricted to a r b i t r a r y maxichoice
contraction functions, then even more conditions b e c o m e equivalent to
(la), i.e. in this c o n t e x t to relationality, as is shown in [1] Section 8.

Added in Proof
The authors have shown t h a t the m a p f* does preserve non-circularity,
so t h a t it can indeed be used in place of f4 in the proof of Theorem 1.
This answers t h e question raised at t h e top of page 197.

References
[1] C. E. A.LC~O~6~ and D. :M:AKINSO~,On the logic of theory eha~ge: contraction
functions and their associated revision functions, Theorla 48 (1982), pp. I~P-37.
[2] C. E. ALCEOV~R6~, P. G~D:E~FORS and D. MAKI~SO~, On the logic of theory
198 C. E. Alchouer6n a~d 1). ~akinsa~

change: partial meet vontravtlon and revislo~ fu~tions, T h e Journal o[ Sy~n.


bolic L o g i c 50 (1985), pp. 510-530.
[3] C. E. ALCHO~YRR6~an(t D. ]I~'AKINSON,O~ the logic of theory ehamge: safe contrac-
tion, S t u d i a Logica 44: (1985), pp, 4:05-4:22.
[4] D. MA~:INSOI%How to give it up: a survey of some formal aspects of the logic of
theory change, S y n t h e s e 62 (1985), pp. 347-363.

FACULDAD DE DE2EOHO DIVISION OF PHILOSOPHY


I.INIVERSIDAD DE ~UENOS AIEES AND HUMAN SCIENCES
~/~G]~NTINA ~NESC0, PARIS
FRANCE

l~eeeived October 18, 1984:

Studia Logiea XLV, 2

You might also like