A paper I wrote for a philosophy of logic seminar.
The Radical
Implications of Dialetheism
I. Introduction
Those logics are called paraconsistent which allow for true
contradictions while attempting to prevent collapsing the system into
triviality, that is, without making every statement true. The
development and appeal of paraconsistent logics (henceforth PL) can
be traced to perhaps three major factors: technical considerations,
such as what paraconsistent logic can do as against classical logic;
its handling of paradoxes, particularly the Liar; and philosophical
considerations relating to the possibility of contradiction.
II.
Technical Aspects of Paraconsistent Logic
In PL, sentences are not simply assigned a truth-value, but rather
are related to sets of
values. That is, a
sentence, φ,
may be related to values
(t, f, p), where t is true, f is false, and p is both truth and
false.
It is important that the
value p
be thought of as both
true and false, not an in-between and certainly not as an undefined
gap. In
Priest's
version of PL,
there are no truth-value gaps. This
means that Priest accepts the Law of Excluded Middle (LEM). That is,
for any sentence A, it is true that A v ~A.
Every sentence is mapped onto one of the three aforementioned
value-sets by the evaluation function, v, such that v: P → {t,p,f},
where P is the set of sentences of our propositional language. In
section III the question of sentences related to multiple value-sets
will be raised, but for now the singly-related scheme is sufficient.
PL,
since inconsistent, is semantically closed, containing its own
predicates for truth, falsity, &c. Indeed, one of the primary
motivations of PL is its relationship to natural language, its lack
or even denial of the legitimacy of constructed metalanguages. It
skirts Gödel's proof of incompleteness because that proof relies on
the language's consistency. Hence, the age-old decision for
consistency as against semantic completeness is here turned on its
head.
In
allowing for sentences to be simultaneously both true and false,
questions
of how to deal with such sentences immediately arises. Classical
logic makes use of what is commonly called the Explosion Principle,
which means that everything follows from a contradiction: A & ~A
├
B, for any A and any B. Clearly
explosion must be denied if contradictions are to be true, and if the
logic is not to be trivial.
However,
PL also comes with a number of losses that may at first seem quite
significant. In the first place, it is easy to check that Modus
Ponens (MP), Reductio ad Absurdum (RaA), and the Disjunctive
Syllogism (DS) fail when paradoxical sentences are present. We
will examine each of these in turn.
One
of either MP
or
Absorption
must be rejected in order to avoid Curry's Paradox. This
paradox states that semantically closed theories that have both the
Absorption Principle and MP lead to explosion. In particular, where
T(x) is 'x is true':
- T(1) → A
- T(1) ↔ (T(1) → A) by the T-scheme
- T(1) → A by absorption
- T(1) MP
- A MP
Thus, since any sentence A is
provable given MP and absorption, we must dispense of one of the two.
Priest decides that Absorption is relatively innocent, and therefore
gets rid of MP. While this is not necessary, it is at least a
reasonable decision.
Perhaps
the most fundamental loss, however, is that of DS. By
this I mean that the problems with MP would seem to be reducible to
those with DS: p → q may be written as ~p v q, but then given p we
cannot conclude q, because we may also get ~p. Furthermore, and
perhaps worse, DS and contradiction together lead to explosion:
- A & ~A Assumption
- A 1 S
- A v B 2 Add
- ~A 1 S
- B 3, 4 DS
Finally, RaA is lost for obvious
reasons.
It
would seem at this point that PL
loses too much to be of logical use, particularly in its loss of MP.
Indeed, it would not be too great a stretch to consider the
conditional of prime logical interest. It
is difficult to imagine a conditional without MP, as it is difficult
to imagine deductive reasoning without a properly intuitive
conditional.
There are, however, reasons for taking a different
conditional than that commonly used in classical logic. In other
words, perhaps the equation of p → q with ~p v q is what is causing
the problem. Priest
suggests that there may be extensions of PL that can deal with these
problems, but for now we will consider the concept of quasi-validity
as a solution. In particular, let us assume that MP, RaA, DS, etc.
are not valid but quasi-valid, meaning they are valid in cases where
the sentences in
question are not dialetheias (that is, true contradictions).
It
can easily be checked that the problems with the aforementioned
inference rules are all related to dialetheias.
Hence, if we have no reason to believe there is a dialetheia
involved in our reasoning, we may help ourselves to quasi-valid
rules. When we discover that a dialetheia
was indeed present that we did not previously realize, we must simply
correct our previous arguments to dispense with the quasi-valid
rules. Hence,
quasi-valid rules are truth- and validity-preserving in most cases
(we assume that paradoxical sentences are a relatively small part of
all true sentences).
III.
The Liar Paradox
One of the main attractions of PL is its ability to deal with
paradoxes, the Liar chief among them. In this section, I will first
present the pure Liar and show how PL might deal with it. Then, I
will examine attempted constructions of a revenge Liar, and show how
they raise difficult but not insoluble problems.
The pure Liar may be given as follows:
(L) (L) is false
Thus, if we evaluate (L) as true, that means it must be false and
thus we have a contradiction. Alternatively, evaluating (L) as false
leads to its being true, because it said of itself that it was false.
Furthermore, it does not help if we admit of truth-value gaps,
because then a revenge Liar can be constructed:
(L*) (L*) is not true
This version, if taken as gappy, would seem to imply that it was
true, because it says of itself that it is gappy. So then it's true
and therefore not gappy, and
yet again we have a contradiction.
PL attempts to skirt the structural problems that give rise to the
paradox by taking (L) to be simultaneously true and false. Here it
will help to translate values into value-sets, a more explicit
conception of truth-values in PL. A value-set is here and for the
moment defined as the set of all truth-values that a sentence relates
to. Thus, t can be written as {1}, f as {0}, and p as {0,1}. That is,
since (L) is a dialetheia, it relates to {0,1}. So in this case there
is no problem with the contradictory nature of (L), in the sense that
the paradox can be contained by denial of both explosion and the LNC.
Furthermore, this conclusion would seem quite intuitive, since we can
say that (L) is false or that (L*) is not true without the nasty
consequences associated with doing so in classical logic, or within
an otherwise classical theory allowing for truth-value gaps. In fact,
when confronted with (L), we are immediately tempted either to say it
is both true and false or that it is neither true nor false—the
second option landing us in the aforementioned problem of (L*).
But perhaps a revenge liar can still be constructed—in fact, I
believe it can, given a broad enough conception of revenge. Consider
the following:
(X) (X) is false only
Of course, the paraconsistent theorist may reply that (X) is false
only... and also true. Philosophically, this is a view of evaluation
that relies not simply on simultaneity, not on a “once and for
all”, but on the ever-present possibility of extending the
evaluation without necessarily rewriting what has gone before. Thus,
(X) may be both false only
and yet
also true, where the “true” determination is added information, a
revision that does not take back that which is revised. We might
question whether the intuitive meaning of “only” has here been
captured, but
I believe there are philosophical reasons for why this
is not a problem
(see section IV). However,
we get into a more troublesome situation if we translate (X) into the
language of value-sets:
(X*) (X*) is related to value-set {0}
If (X*) is evaluated with value-set
{1}, which we translate as
“true”, that would mean it has value-set {0}, and therefore it
has both {0} and {1}, but since each sentence can be related only to
one value-set, we have the unfortunate result that 0=1. If (X*) is
evaluated as {0}, that means it's false, and therefore what it was
saying is true, which means it has {1}. Again, we have that 0=1. It
seems likely that a paraconsistent theorist might
evaluate (X*) as having value-set {0,1}, in which case it's true, in
which case it's value-set is {0}, and therefore again we have the
same problem. So, structurally, we may say that the problem arises
when the Liar is made to explicitly define it's own evaluation, and
it appears that PL does not wholly avoid the problem, ending up not
just with contradiction but with wholesale triviality.
However,
perhaps there is another, more radical way out of
the problem, namely the
possibility of a sentence relating to multiple value-sets. There
are several ways to denote this, but I shall use the following
method: when a sentence relates to sets {0} and {0,1} I shall say it
relates to {{0},{0,1}}. In other words, the value-sets related to by
a sentence I will put into another set. The implications of this,
while not wholly worked out, might be as follows. First, why not just
let the {0} in the above example be absorbed into the set {0,1}, so
the sentence only relates to that one latter set as defined by
Priest? It would appear to me
that we cannot do this,
because to accurately capture the difference between “false only”
and “false (and true)” there ought to be a difference between the
two sets. Thus, a sentence relating to {0,1} has a different
truth-profile (or some such construction) than
a sentence relating to {{0},{1}}. This
implies a different
conception of sets than the traditional set defined purely by its
members. While working this
out would be an absolutely enormous undertaking, I think that's the
direction in which PL must go. Intuitively, a contradictory theory
ought to go with some kind of contradictory or otherwise non-standard
set theory.
IV.
Philosophical Considerations
In the previous section I alluded to the problem of capturing the
word “only”, as in the case “This sentence is false only”.
With the above considerations, we have that a sentence may be false
only and also true at one and
the same time. The obvious
question to ask is whether the “also”
in fact undermines the meaning of the previous “only”. The
dialetheist might say that it does not, because the subsequent “also”
simply adds more
information rather than somehow canceling out previous information.
Put this way, the problem is structurally analogous to the
status of contradictions—when we posit something that seems to
directly contradict a previous statement, do we not in fact overwrite
the older statement in favor of the new one? Can we (rationally or
not) accept or even hold in our mind A & ~A, or in this case hold
that something is only
false and also that it is true as well, which would seem to be
explicitly contradictory?
First,
it is an evidential claim
that people can or cannot hold a contradiction. While
I believe it is clear and obvious that people believe many
self-contradictory things, a better question
would be whether that behavior is
rational or not. Following
Priest, I am inclined to say that non-contradiction may be one
criterion for rationality among many others, and that in some cases
it is rational to believe a contradiction, for example where other
rational criteria are better fulfilled than by an alternative,
non-contradictory belief. However,
this is certainly not necessary for a dialetheist to put forward, and
in fact Hegel himself believed in the contradiction of everything,
propositional or otherwise. In
fact, I think Hegel is a good example of a non-trivial dialetheism
wherein everything is contradictory (these words have here changed
from their previous technical meanings, while still, I hope,
capturing the intuitions), though this is far
outside the confines of the
philosophy of logic.
Second,
we may return to the “only” problem and see if we can make any
progress. We would not say
that dialetheism does not take contradiction seriously because of the
“added information” argument, and structurally there is nothing
different about the “only” problem, and indeed it is simply a
specific linguistic example of a contradiction. What should therefore
be said, rather than that dialetheism
cannot understand “only”, is that non-dialetheist
theories do not treat contradiction seriously, in that it is only
dialetheism that takes contradiction as an existing fact not to be
rejected. In most cases,
“false only” means what the non-dialetheist means by it, but
there are also dialetheias, such as the above Liar sentence (X),
which must be taken in a
contradictory way.
Alternatively,
we might say, perhaps even against PL as a whole, that true and false
might not in fact be defined purely negatively, that is, against each
other such that true is simply not-false and false is simply
not-true. Rather, a sentence relating to {0} might in
fact have different effects or conceptual implications associated
with that 0 than a sentence related to {0,1}. This
would make sense epistemologically, because surely
the effects or implications (these concepts unfortunately cannot be
fully worked out here) are not, at least in some cases, the exact
same. While this certainly
takes us away from PL as a system, it is a necessary consideration
since it does not take us away from contradiction, and attempts to
deal with the same problems as does PL, only from a different angle,
from a different translation
of what would seem to be very similar insights. In
any case, this raises many more problems the research of which might
be fruitful.
V.
Conclusion
PL is certainly not taking an easy way out of the problem of
paradoxes. It is certainly not a cure-all for the diseases of
traditional approaches to paradox, and providing a set theory
appropriate for it is a massive undertaking. However, PL does
exemplify the attempt to be comfortable with the universal inability
to make everything consistent. In this sense, dialetheism opens the
way towards reasoning with what are in the last analysis simply facts
of life. In good dialetheist fashion, the rest might follow as so
much more “added information”.
Works Consulted
Priest,
Graham. In Contradiction: A
Study of the Transconsistent. Dordrecht;
Boston: Distributors for the U.S. and Canada, Kluwer Academic, 1987.
Print.
---. “The Logic of Paradox.” The Journal of Philosophy.
Vol 8. No 1 (1979): 219-241. Web. 16 Feb 2014.
---. “What Is so Bad About Contradictions?” The Journal of
Philosophy. Vol 95. No 8 (1998): 410-426. Web. 16 Feb 2014.