Professional Documents
Culture Documents
BF 00763511
BF 00763511
ABSTRACT. We define a system IR of first-order intuitionistic ex falso quodlibet. Intuitionistic relevant proofs
relevant logic. We show that intuitionistic mathematics (on the "accumulate" to yield deductive progress; and when
assumption that it is consistent) can be relevantized, by virtue of the
they fail to do so, we thereby learn something stronger:
following metatheorem: any intuitionistic proof of A from a set X
of premisses can be converted into a proof in IR of either A or that the accumulated premisses are inconsistent (or that
absurdity from some subset of X. Thus IR establishes the same the sought conclusion is a logical truth).
inconsistencies and theorems as intuitionistic logic, and allows one
to prove every intuitionistic consequence of any consistent set of
premisses. Some notions
(i) (i)
A(a)
where there must be an assumption of where there need not be an
the form A to be discharged assumption of the form A(a) to
be discharged; and a does not
_k 3xA(x) B occur in 3xA(x), B or any
- - (i) (i)
ADB B assumption other than A(a) on
which the upper occurrence of
B depends
--A--(i)
Relevant 3Elimination
where there need not be an assumption
of the form A to be discharged
(i)
A(a)
B
- - (i) where there must be an as-
ADB
sumption of the form A(a) to
(N.B.: These rules make D behave differently from the be discharged; 3xA(x) is not
weak and strong conditionals discussed in Tennant 3xA(x) B (i) introductory; and a does not
(1987a, b)) B occur in 3xA(x), B or any
assumption other than A(a) on
Ordinary DElimination
which the upper occurrence of
A ADB B depends
B
A word of explanation would be in order about
Relevant DElimination the special relevant forms of VElimination and
DIntroduction.
A ADB where ADB is not introductory
Example: Relevant VElimination allows one to prove
B
disjunctive syllogism relevantly:
Ordinary and Relevant Vlntroduction
in FI0. But that can be remedied, by supplying an DIntroduction stated separately, if we are to be able to
application of the absurdity rule at the end of the first prove the extraction theorem for the full language with
subordinate deduction for VElimination: D.
Recall our statement of our main result above:
~-- (1) -A For the full language based on -, &, V, D, 3 and V,
A 2_ any intuitionistic relevant natural deductions estab-
AVB --B --B- (1) lishing A0, B 1. . . . . B. ~ A, A 1 ~ B 1. . . . . A. ~ B~
(1) can be converted into an intuitionistic relevant natural
B
deduction of A or of 2- from some subset of UiAi.
so that we now have a non-normal but standard Schematically we have
intuitionistic deduction of B from {A, ~A}. Then we
normalize it by applying VReduction to get Al An
I'I1, . . . , I'I n
A -A
Bl Bn A (__ t_JiAi)
2_
H
B
/~0, B 1 , - . . , Bn ±/A
Finally we extract the application of the absurdity rule 1-10
to obtain as our desired relevant deduction 17 the A
relevant deduction
The proof of the main result proceeds along the
A ~A
_k following lines, already illustrated in the example
above. When the relevant deductions I ' I 1 . . . . , I-In are
of 2_ (rather than B) from the accumulated undischarged grafted onto the respective assumptions B1. . . . . B n of
assumptions A = {A, -A}. the relevant deduction H 0, one obtains a tree that can be
turned into an intuitionistic deduction (albeit possibly
Observation 1. Note that any application of the relevant an abnormal one) by supplying extra applications of the
version of VElimination that has 2_ as one subordinate absurdity rule wherever they are needed in order to turn
conclusion and A as the other can always be turned into applications of relevant VElimination and Dlntroduction
a standard intuitionistic application of VElimination by into applications of the ordinary intuitionistic rules (by
supplying one application of the absurdity rule to get A Observations 1 and 2 above). By the normalization
from 2_ in the case that ends with 2_. theorem one then finds a normal intuitionistic deduc-
tion of A from A. By the extraction theorem one extracts
Observation 2. The same sort of remark applies in the from it all applications of the absurdity rule in such a
case of the extra relevantist portion of DIntroduction. way as to find a relevant deduction of 2- or of A from
In the standard case one does not need to state the first (some subset A of) UiA i. In order to carry the extrac-
half of relevant DIntroduction separately: tion through, the liberalized versions of VElimination
and Dlntroduction are crucial.
~- (i) The proof of extraction is by induction on the length
where there must be an assumption of of intuitionistic deductions in normal form.
the form A to be discharged
Basis. Obvious.
2_
(i) Inductive hypothesis. The result holds for all normal
ADB
intuitionistic deductions less complex than II.
since the presence of the absurdity rule allows one to
turn any application of it into an application of the Inductive step. By cases, according to the last rule
ordinary form o f Dlntroduction. But in the case of applied in the normal intuitionistic deduction Yl. The
intuitionistic relevant deduction, where we cannot use only cases where care is required are VElimination and
the absurdity rule, it is vital to have that first half of Dlntroduction; but I shall also illustrate the ideas in
INTUITIONISTIC MATHEMATICS 131
A,A- (i)
£ VElimination
_L
(i) The complication that may arise when rI ends with an
~A
application of VElimination with (non-introductory)
where A is not in A. By inductive hypothesis applied to major premiss AVB is that the relevant deductions
E there is a relevant intuitionistic deduction ~ of ± guaranteed by inductive hypothesis may not have the
from some subset F o f A U {A}. I f A is not in F, take same conclusions: one may end with ± (and involve the
g~ as the required relevant intuitionistic deduction case assumption A), the other may end with C (and
corresponding to H. If A is in F, then F = • U {A} involve the case assumption B); or the other way round.
where A is not in O, and we can construct In either case, the relevant form of VElimination is
specifically designed to cope: one may apply it to the
sub-deductions guaranteed by inductive hypothesis to
O,A- (i) obtain overall a relevant intuitionistic deduction of C
Y~ from some subset of the assumptions of Yl.
±
~/~ (i)
D lntroduction
as the required relevant intuitionistic deduction corre-
sponding to H.
Suppose the normal intuitionistic deduction II is of the
form
132 NEIL TENNANT