PHIL
411 Advanced Logic
LARKIN
__________________________
Metalogic
[For:
Richard Jeffrey, Formal Logic: Its Scope and Limits]
I.
Formal
Logic
A.
A
system of formal (deductive) logic aims at developing a method for testing
whether arguments are valid.
B.
As
validity is a property of logical structure (not content), the system should
provide first a language into which one can translate arguments in ordinary
language so as to reveal the relevant level of logical structure. This language is essentially specified by
providing:
1.
An
alphabet—a list of symbols that can be used
2.
Rules
of formation—rules that determine which strings of symbols count as well-formed
formulas (wffs).
3.
Rules
of Interpretation—rules that determine the allowable ways symbols can be
interpreted and under what conditions a wff will be true in an
interpretation. These semantic rules
determine when a conclusion validly follows from some set of premises.
C.
Proof
Procedure—Constituted by Rules of Inference, which essentially determine the
allowable ways wffs can be manipulated.
These syntactic rules determine when some conclusion can be derived
from some set of premises.
D.
Desiderata
1.
In
a good system of formal logic, there should be certain relationships between
the semantic rules of interpretation and the syntactic rules of inference:
a.
Soundness: If some conclusion is derivable from some
set of premises, then that conclusion validly follows from that set of
premises.
b.
Completeness: If some conclusion validly follows from some
set of premises, then that conclusion is derivable from that set of premises.
2.
In
a good system of formal logic the inference rules should provide a method that
will pronounce correctly on the validity of any finite argument.
a.
Decidability:
The method will pronounce on the validity of any finite argument.
b.
Soundness: The pronouncements of ‘valid’ will be
correct.
c.
Completeness: The pronouncements of ‘invalid’ will be
correct.
II.
Decidability
A.
Definitions:
1.
A
system S decides validity =df For any finite argument A: S
will pronounce either that A is or is not valid after a finite number of
finitely long mechanical steps.
2.
The
tree method T decides validity =df For any finite argument A: T
will claim either that A is valid or not after a finite number of mechanical
(and finitely long) steps.
B.
Exhaustive
Rules
An inference rule is exhaustive iff applying
the rule to a line (input) yields a finite number of smaller lines (output).
C.
Basic
Decidability Proof
P1: If a
tree corresponding to any finite argument A is finished, then it either has an
open path or it does not.
P2: If the
tree has an open path, then T pronounces A ‘not valid’.
P3: If the
tree has no open paths, then T pronounces A ‘valid’.
C1: So if
the corresponding tree for A finishes, then T pronounces on the validity of
A.
P4: If every
rule is exhaustive, then every tree will finish.
P5: Every
rule is exhaustive.
C2: So every
tree will finish.
C3: So T
pronounces on the validity of every finite argument A.
D.
Propositional
Logic (Ch. 2)
1.
Every
truth-functional rule of inference is exhaustive.
E.
Predicate
Logic
1.
Generality
(Ch. 3):
a.
We
add QN rules, EI and UI.
b.
UI
is not exhaustive. Applying UI to a
single finite line can yield an infinite number of new lines. So P5 of the basic decidability proof is
false for our system of predicate logic.
c.
But
UI is effectively exhaustive if there are only a finite number of names
that can appear in a tree. For then we
can instantiate any universal claim to all relevant names in a finite number of
applications of UI. All relevant
applications of UI to a line will yield a finite number of smaller lines.
d.
We
can then make the following changes to our basic decidability proof:
P1*: If a
tree for A is finished or effectively finished, then it either has an open path
or it does not.
C1*: If a
tree for A finishes or effectively finishes, then T pronounces on the validity
of A.
P4*: If
every rule is either exhaustive or effectively exhaustive, then every tree will
finish or effectively finish.
P5*: Every
rule is either exhaustive or effectively exhaustive.
C2*: So
every tree will finish or effectively finish.
2.
Multiple
Generality (Ch. 4):
a.
An
existential quantifier within the scope of a universal quantifier can generate
an infinite number of names.
b.
So
when we can have existential quantifiers within the scope of universal
quantifiers, UI ceases to be even effectively exhaustive.
c.
Thus
when we expand our tree system to allow for multiple generality, the system
ceases to be able to decide validity.
Some finite arguments will be such that the corresponding tree never
finishes or effectively finishes. And
so some finite arguments will be such that T never pronounces on their
validity.
III.
Soundness
A.
Definition:
1.
A
system S is sound =df For any finite argument A: if S
pronounces A ‘valid’, then A is valid.
2.
The
tree method T is sound =df For any finite argument A
(P1…Pn/\C): if T pronounces a
corresponding initial list of claims L (P1…Pn, ~C) ‘inconsistent’ by there
being no open paths through a finished tree constructed from L, then L really
is inconsistent.
B.
Sound
Rules
1.
An
inference rule is sound =df If the input line of the rule
is true in some interpretation C, then all of the lines along some output
branch are true in C as well.
2.
An
inference rule is constructively sound =df If the input line of the
rule is true in some interpretation C, then we can construct a new
interpretation C* such that everything true-in-C is still true-in-C* and all of
the lines along some output branch are true-in-C* as well.
C.
Basic
Soundness Proof
P1: Assume
that L is consistent.
P2: Then
there is an interpretation C in which all the member of L are true.
P3: If all
of our rules are sound, then truth-in-C will ‘trickle down’ along at least one
branch every time a rule is applied and so there will be a path P along which
all lines are true-in-C.
P4: All of
our inference rules are sound.
C1: So,
there is a path P along which all lines are true-in-C.
P5: If all
lines along P are true in some interpretation C, then P is open (does not
contain and formula and its denial).
C2: So, if L
is consistent, then there is a path P through a tree constructed from L that is
open.
C3: So, if
there is no open path P through a tree constructed from L, then L is
inconsistent.
D.
Propositional
Logic: All of our propositional rules
of inference are sound.
E.
Predicate
Logic
1.
Our
QN rules and UI are sound.
2.
EI
is not sound.
3.
EI
is constructively sound.
4.
All
of the rules that are sound are also constructively sound.
IV.
Completeness
A.
Definitions
1.
A
system S is complete =df For any finite argument A: if A
is valid, then S will pronounce A ‘valid’.
2.
A
system S is complete =df For any finite argument A: if S
pronounces A ‘not valid’, then A is not valid.
3.
The
tree method T is complete =df For any finite argument A
(P1…Pn/\C): if T pronounces a
corresponding initial list of claims L (P1…Pn, ~C) ‘consistent’ by there being
some open path P through a finished (or effectively finished) tree constructed
from L, then L is consistent.
B.
Complete
Rules
1.
An
inference rule is complete =df If all of the lines along
some output branch are true in some interpretation C, then the input line is
true-in-C as well.
2.
An
inference rule is path complete =df If all of the lines along
some output branch on an open (or infinite) path P are true in some
interpretation C determined by P, then the input line on P is true in C
as well.
3.
C.
Basic
Completeness Proof
P1: Assume
that there is some open path P through a finished tree constructed from L.
P2: Since P
does not contain both a formula and its denial, there is some interpretation C
where all atomic formulas and denials of atomic formulas along P are true. (This is an interpretation C determined by
P.)
P3: If all
of our rules are complete, then truth-in-C will be ‘drawn up’ through all of
the lines along P.
P4: All of
our rules are complete.
C1: So there
is some interpretation C that makes all of the lines along P true.
C2: So if
there is some open path P through a tree constructed from L, then there will be
an interpretation C that makes all lines along P true.
P5: All the
claims of L will lie along P.
C3: So if
there is some open path P through a tree constructed from L, then there will be
an interpretation C that makes all the lines in L true.
P6: If there
is an interpretation that makes all the lines in L true, then L is consistent.
C4: So if
there is some open path P through a finished tree constructed from L, then L is
consistent.
D.
Propositional
Logic: All of our propositional rules of inference are complete.
E.
Predicate
Logic
1.
Our
QN rules and EI are complete.
2.
UI
is not complete.
3.
UI
is path complete.
4.
Any
rule that is complete is also path complete.
5.
V.