# A Puzzle about Belief and the Limits of Knowledge

Heidi Howkins Lockwood
Yale University
May, 2008

“As we know,
there are known knowns.
There are things we know we know.
We also know there are known unknowns.
That is to say
we know there are some things
we do not know.
But there are also unknown unknowns,
the ones
we don‟t know we don‟t know.”

~ U.S. Defense Secretary Donald Rumsfeld
at a Feb. 12, 2002 Department of Defense news briefing
in response to questions about U.S. intelligence on
weapons of mass destruction in Iraq
Overview

I. Introduction

II. Proof of the Possibility of Believing an Impossibility
Part 1 – A conditional proof
Part 2 – Discharging the “Cartesian” premise

III. The Limits of Knowledge (and other operators):
A Löb-like incompleteness result

IV. Discussion
Diagonalizing on propositions

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Introduction

It is generally accepted that it is impossible to know the
impossible. To hold otherwise would require radical
revisions to the logic of knowability.

What is impossible is always false, what is false is never
true, and knowledge presupposes truth. It is therefore
impossible to know p if p is impossible.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion

But what about operators that don‟t presuppose truth ÷
operators such as imagining, representing, and believing?

Mere belief ÷ as opposed to true belief ÷ is a case in point.

Is it possible to believe that 2+3=6, or that a triangle has 5
sides, or that there is a greatest even number, assuming full
comprehension of concepts such as addition, greatest, even,
and so forth?

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Many philosophers have thought not.

Many theories of belief have presumed, either explicitly or
implicitly, that it is not possible to even merely believe a
logical or conceptual impossibility.

This would be true, for example, of Bob Stalnaker‟s
possible-worlds view of belief propositions, David Lewis‟
map-like representational account of belief, and Ruth
Barcan Marcus‟ traditional dispositionalism ÷ all of which
presuppose externalist theories of belief.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
The thesis that it is impossible to believe an impossibility
also appears at many points in the history of philosophy.

Berkeley links belief to imagination and denies that we can
imagine impossibilities.

Logical positivists such as Schlick use the unthinkability of
impossibilities and logical impossibilities.

And so on.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion

Let‟s dub the generic view that impossibilities or contradictions
cannot possibly be the object of a possible belief possibilism.

In symbols:

~^-p (^Bp . ~^p)

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion

As we will prove in a moment, this view is impossible.

That is, possibilism is provably false.

In symbols:

├ ^-p (^Bp . ~^p)

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion

The proof proceeds in two parts.

The first part is a conditional proof of
the possibility of believing an impossibility
based on the “fixed point” premise
that it is possible to believe that it is
possible that it is possible to believe an impossibility.

In the second part of the proof,
we will discard the reliance on the fixed point.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Proof of the Possibility
of Believing an Impossibility

Part 1

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion

Suppose that I assert, contrary to possibilism, that I believe that it
is in fact possible that it is possible to believe an impossibility.

In symbols:
B(^-p (^Bp . ~^p))

Let‟s call this the “fixed point” premise. As we will see in a
moment, it guarantees the truth of the proposition believed.

To make the proof more readable, let‟s use „c‟ to refer the formula
-p (^Bp . ~^p).
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Apart from calling me cognitively impaired and perhaps in the
grip of a Kripkean referential delusion, the possibilist will no
doubt assert this: it is impossible that it is possible to believe an
impossibility.

And therein lies the catch.

For, if the is correct in asserting that the object of my belief ÷
the claim that it is possible to believe an impossibility ÷ is
impossible, then she is admitting that I have succeeded in
believing an impossibility, and thereby refuting her own claim.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Let‟s take a closer look at this argument in symbols.

Aside from the fixed point premise, the only assumptions are
those of classical logic and a standard S4 modal framework.

We will discard the fixed point premise in Part 2 of the proof.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
First, using „c‟ to refer to -p (^Bp . ~^p),
we have the fixed point premise:
(1) B^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
First, using „c‟ to refer to -p (^Bp . ~^p),
we have the fixed point premise:
(1) B^c

From this and the principle that whatever is actual is possible
(p ÷ ^p), we get:
(2) ^B^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c

(3) ~^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c

(3) ~^c

From (3) and the principle that whatever is impossible is not
possibly possible (~^p ÷ ~^^p), we have:
(4) ~^^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c

The conjunction of (2) and (4) gives us:
(5) ^B^c . ~^^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c

The conjunction of (2) and (4) gives us:
(5) ^B^c . ~^^c

And existential generalization on (5) produces:
(6) -p(^Bp . ~^p)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)

But (6), of course, is just:
(7) c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c

Another application of the principle that the actual is possible
gives us:
(8) ^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c
(8) ^c

And (8) contradicts (3). So we have:
(9) ~~^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c
(8) ^c
(9) ~~^c
Applying double negation to (9) of course gives us:
(10) ^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c
(8) ^c
(9) ~~^c
(10) ^c
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c
(8) ^c
(9) ~~^c
(10) ^c
This argument proves ^c assuming
the truth of the fixed point
premise, B^c.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c
(8) ^c
(9) ~~^c
(10) ^c
This argument proves ^c assuming
the truth of the fixed point
premise, B^c.

In other words, the affirmation of
belief in the possibility of
believing the impossible is
infallible because it is self-fulfilling.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(1) B^c
(2) ^B^c
(3) ~^c
(4) ~^^c
(5) ^B^c . ~^^c
(6) -p(^Bp . ~^p)
(7) c
(8) ^c
(9) ~~^c
(10) ^c
This argument proves ^c assuming
the truth of the fixed point
premise, B^c.

In other words, the affirmation of
belief in the possibility of
believing the impossible is
infallible because it is self-fulfilling.

If B^-p(^Bp . ~^p),
then in fact ^-p(^Bp . ~^p).

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
But the truth of B^c is not incontrovertible.

One might, for example, worry about the semantic stability
of the terms in question.

We turn now to the project of strengthening the proof by
eliminating the dependence on the fixed point premise.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Proof of the Possibility
of Believing an Impossibility

Part 2

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Recall that „c‟ refers to -p (^Bp . ~^p). In Part 1 of the proof we
have shown that ├ (B^c ÷ ^c).

To improve the readability of the second half of the proof, let‟s
embed the redundant „^‟ in a new constant. In other words, let „c*‟
refer to

^-p (^Bp . ~^p)

So we have now shown that ├ (Bc* ÷ c*), or, alternatively
(weakening it slightly by disregarding the first line of the proof),
that ^Bc* ÷ c* is provable. This gives us:

(11) ^Bc* ÷ c*
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*

For the remainder of the proof, it will be convenient to note
that believability, which we have been representing as „^B‟,
intuitively satisfies what are known as the Hilbert-Bernays-Löb
(HBL) conditions in provability logic.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
The Hilbert-Bernays-Löb (HBL) conditions:

(i) if ├ p , then ├ ^Bp
(ii) ├ ^B(p ÷ q) ÷ (^Bp ÷ ^Bq)
(iii) ├ ^Bp ÷ ^B^Bp
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(i) if ├ p , then ├ ^Bp

Condition (i) says that if p is a theorem (is provably true), then the
fact that it is possible to believe that p is also a theorem. Note that
this condition does not encounter the well-known troubles
associated with its counterpart for knowability, the problematic
knowability thesis or strong verificationist thesis. In particular, (i) is
not susceptible to the woes associated with the family of Fitch-
Church knowability paradoxes because belief and believability are
not factive.

It merely asserts that we can derive the believability of p from the
derivability of p.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(ii) ├ ^B(p ÷ q) ÷ (^Bp ÷ ^Bq)

Condition (ii) says that if it is possible to believe that p entails
q, then if it is possible to believe that p, it is possible to believe
that q. Worries can certainly be raised about the counterpart of
(ii) for belief (as opposed to believability). We might, for example,
argue that the belief that p entails q does not entail that the
belief that p entails the belief that q, on the grounds that a
certain sort of simultaneity of the belief that p entails q and the
belief that p is required for the formation of the belief that q.

But condition (ii) is a claim about the metaphysical or
conceptual landscape, not an assertion about epistemic
entailment. It merely asserts that if it is believable that p entails q,
then if it is believable that p, it is believable that q.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(iii) ├ ^Bp ÷ ^B^Bp

Condition (iii) says that if it is possible to believe
that p, then it is possible to believe that it is
possible to believe that p. Again, the modality takes
the bite out of this claim.

Exhibiting the derivability of the believability of p
suffices to show that the believability of p is
believable.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Now, given that we have proved:

(11) ^Bc* ÷ c*

The technique of diagonalization, introduced by Gödel [1931],
gives us a formula „a’ such that ├ a ÷ (^Ba ÷ c):

(12) a ÷ (^Ba ÷ c*)

Taking just the left-right direction of this biconditional gives us:

(13) a ÷ (^Ba ÷ c*)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)

Clause (i) of the HBL conditions together with (13) gives us:
(14) ^B(a ÷ (^Ba ÷ c*))

By clause (ii) of the HBL conditions, we know:
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))

And from (14) and (15) we get:
(16) ^Ba ÷ ^B(^Ba ÷ c*)

By clause (ii) again we know:
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)

So by (16) and (17) we can infer:
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)

Clause (iii) of the HBL conditions gives us:
(19) ^Ba ÷ ^B^Ba
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba

From (18) and (19) we know:
(20) ^Ba ÷ ^Bc*
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*

And (20), together with (11), gives us:
(21) ^Ba ÷ c*
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
By (12) and (21) we can now derive:
(22) a
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
By virtue of (22) and clause (i), we get:
(23) ^Ba
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
(23) ^Ba
Finally, (23) together with (21) gives us:
(24) c*
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
(23) ^Ba
(24) c*
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
(23) ^Ba
(24) c*
Recall that „c*‟ is ^-p(^Bp . ~^p).
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
(23) ^Ba
(24) c*
Recall that „c*‟ is ^-p(^Bp . ~^p).

We have now eliminated the initial
dependence on the fixed point
premise, and proven that it is
possible to believe an impossibility.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc*)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
(23) ^Ba
(24) c*
Recall that „c*‟ is ^-p(^Bp . ~^p).

Note that this argument is not restricted
to belief.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
(11) ^Bc* ÷ c*
(12) a ÷ (^Ba ÷ c*)
(13) a ÷ (^Ba ÷ c*)
(14) ^B(a ÷ (^Ba ÷ c*))
(15) ^B(a ÷ (^Ba ÷ c*)) ÷ (^Ba ÷ ^B(^Ba ÷ c*))
(16) ^Ba ÷ ^B(^Ba ÷ c*)
(17) ^B(^Ba ÷ c*) ÷ (^B^Ba ÷ ^Bc*)
(18) ^Ba ÷ (^B^Ba ÷ ^Bc)
(19) ^Ba ÷ ^B^Ba
(20) ^Ba ÷ ^Bc*
(21) ^Ba ÷ c*
(22) a
(23) ^Ba
(24) c*
Recall that „c*‟ is ^-p(^Bp . ~^p).

It will work for any operator ¢ for which
the HBL conditions hold and it is
provable that ¢c ÷ c.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Is there an analogous result for knowability?

For provability?
For ^T?
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
The first half of the proof relied only on an S4 modal framework
and classical logic.

So we can extend the results of the first half to:

├ (^Kc
K
÷ c
K
)
├ (Bew(c
Bew
) ÷ c
Bew
)
├ (^Tc
T
÷ c
T
)

where „c
¢
‟ is -p(^¢p . ~^p).

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
For the second half of the proof, we need the HBL conditions:

(i) if ├ p , then ├ ¢p
(ii) ├ ¢(p ÷ q) ÷ (¢p ÷ ¢q)
(iii) ├ ¢p ÷ ¢(¢p)

These are standardly assumed to hold in the case in which the modal
operator „¢‟ is taken to be provability (Bew).

They also seem to hold for possible truth (^T) and knowability (^K),
given an appropriate interpretation of „÷‟.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
In other words, the proof we have just seen
can be used to show:

├ ^-p(^Kp . ~^p)

├ ^-p(Bew(p) . ~^p)

├ ^-p(^Tp . ~^p)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
The Structural Limits of Knowledge
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Let‟s suppose for a moment that the HBL conditions hold for
knowability – i.e., let‟s suppose, for a factive knowledge operator
„K‟, that ^K satisfies:

(i) if ├ p , then ├ ^Kp
(ii) ├ ^K(p ÷ q) ÷ (^Kp ÷ ^Kq)
(iii) ├ ^Kp ÷ ^K^Kp

(Notice that (i) is just the knowability principle.)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Given this assumption (that the HBL conditions
hold), then by Löb‟s Theorem [1955], we know:

├ ^K(^Kp ÷ p) ÷ ^Kp

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
It is worth pausing at this point to note that Gödel‟s second
incompleteness theorem is an immediate consequence of Löb‟s
Theorem. This is easiest to see when the theorem is expressed as:

├ (p ÷ p) ÷ p

for if we express the inconsistency of a theory using „±‟
then consistency is representable as:

~ ±

which is equivalent to:

± ÷ ±
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
In other words, with p = ±, Löb‟s Theorem says that the consistency
of the theory is provable only if the theory is inconsistent:

├ (± ÷ ±) ÷ ±

― which is of course just the second incompleteness theorem.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Similarly, in the case of knowability, if we use the symbol „±‟ to
represent a contradiction or impossibility – perhaps the proposition
expressed by ^-p(^Kp . ~^p) – then the knowability of the
possibility of knowing an impossibility is representable as ^K± and
the unknowability of the possibility of knowing an impossibility is
representable as ~^K±, which is equivalent to ^K± ÷ ±.

So, by Löb‟s Theorem, we have:

├ ^K(^K± ÷ ±) ÷ ^K±

And this tells us that either it is possible to know an impossibility, or
it is not possible to know that it is not possible to know an
impossibility.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
So we now know that if the HBL conditions hold for knowability –
in particular, if the knowability thesis expressed by condition (i)
holds – then either it is possible to know a contradiction (knowledge
is inconsistent), or knowledge is incomplete.

In other words, either the knowability thesis doesn‟t hold, in which
case there are unknowable truths (knowledge is incomplete), or the
knowability thesis does hold, in which case knowledge is either
inconsistent or incomplete.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Those familiar with the family of Fitch-Church knowability paradox
may detect a whiff of the knowability paradox in this result.

The Fitch-Church result is a surprising refutation of the thesis that
all truths are knowable based on three assumptions: (1) knowledge is
factive, (2) knowledge distributes across conjunction, and (3) there is
at least one unknown truth.

It is easy to misinterpret the knowability paradox as a generalization
of Gödel‟s first incompleteness theorem, which demonstrates that,
for any consistent, sufficiently strong theory T in the language of
arithmetic, there are truths unprovable in T. The problem with doing
this is that the Fitch-Church result rests on the existence of an
unknown truth, which is arguably a contingent fact.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
The proofs we have just completed
free the Fitch-Church result from the
arguably contingent assumption of the
existence of an unknown truth.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Discussion
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
We now know, for any modal operator ¢ which satisfies the
HBL conditions, that:

├ ^-p(¢p . ~^p)

and

├ ¢(¢± ÷ ±) ÷ ¢±

(i.e., either contradictions are ¢-able, or the fact that
contradictions are not ¢-able is not ¢-able)
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
One way to understand the underlying tension between
consistency and completeness is as a symptom of a
cardinality mismatch.

A cardinality mismatch, that is, between the expressible and
therefore at most denumerably infinite number of ¢-ables and
the non-denumerable number of potenial ¢-ables that is
generated through the iteration of ¢.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
These thoughts raise a natural question:

What are we taking the
objects of knowledge (or belief) to be?

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
These thoughts raise a natural question:

What are we taking the
objects of knowledge (or belief) to be?

Do the results presuppose any limits on the
objects of knowledge or belief ?

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
These thoughts raise a natural question:

What are we taking the
objects of knowledge (or belief) to be?

Do the results presuppose any limits on the
objects of knowledge or belief ?

The quick answer to this is:
yes, the results presuppose that
the objects of knowledge or belief are expressible.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
This presupposition enters the proof at
the point at which we diagonalized on the
objects of belief or knowledge.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Propositions are (on most accounts) mind-independent extra-
linguistic abstract entities. So how does diagonalization on
propositions work?

Well, briefly and without entering into the fray on propositions, let‟s
suppose that the objects of knowledge are propositions, and that a
proposition is knowable only if it is expressible. (Propositions which
are ineffable, inexpressible, or otherwise non-assertable and non-
communicable are not knowable, and therefore would not be
suitable candidates for objects of knowledge.)

We know how to diagonalize on sentences and formulae. So in
order to diagonalize on propositions, it suffices to come up with a
systematic method for mapping (via an injective function)
expressible propositions onto sentences.

Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Each proposition may in theory be expressed by many different –
in fact, possibly infinitely many – different sentences.

In other words, for sentences S
1
and S
2
such that S
1
= S
2 ,
given a
meaning relation M which takes sentences to propositions, it may
be the case that M(S
1
) = M(S
2
).

So we can‟t rely on an expression function to take propositions to
sentences.

We can, however, rely on an expression relation, Ex, which takes
propositions to equivalence classes of sentences:

Ex(p) = {S| S expresses p}
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
Here‟s the key: since each equivalence class is a set of sentences,
each of which has finite length, each equivalence class has a well-
ordering and can be put into one-to-one correspondence with the
natural numbers.

In particular, since each member of each equivalence class is a
sentence of finite length, it can be assigned a unique Gödel
number, call it g(S). We can then well-order the members of Ex(p)
by simply using the standard „<‟ relation on {g(S) | S e Ex(p)}.

And, once we have the well-ordering of Ex(p), we can uniquely
encode p by using the least element of the well-ordering.
Diagonalization then proceeds as usual (in this particular proof
using Curry‟s paradox rather than the Liar).
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
One final note:

The result we have discussed does not assume that propositions are
the objects of knowledge, but it does assume that it makes sense
to assert that something is known or is knowable – so at some level
it does assume that there are “objects” of knowledge. These
objects, however, are not restricted to potential truth-bearers such
as sentences or propositions. They could, for example, be some
entity or feature of a possible world that is either a truth maker or a
constituent of a truth maker.

In the case of objects viewed as truth makers in the actual world,
the knowledge could be knowledge by direct acquaintance, i.e.,
knowledge that involves an unmediated relation between the
subject and the truth maker.
Introduction - Proof of Possibility of Believing an Impossibility: Part 1, Part 2 - Structural Limits of Knowledge - Discussion
In other words, the result is not restricted to
knowledge that involves characterizing, representing,
or conceptualizing the objects of knowledge. It does,
however, assume that those objects are what I will call
“discretely graspable”, i.e., either finite or effectively finite
(expressible through a recursively or effectively
enumerable string of symbols, or in principle
observable via deterministic effects).
I am grateful to the participants in the
Yale spring 2008 works in progress seminar for helpful