You are on page 1of 3

# C1.

## Kant, Cantor, and the Burali-Fortis Paradox

Jari Palomki
Tampere University of Technology/Pori
Pohjoisranta 11, B.O.Pox 300, FI-28101 Pori, Finland
jari.palomaki@tut.fi
The Burali-Fortis paradox demonstrates that the class of all ordinals is not a set. Cantor, however, had no
concern over the paradox, since his conception of a set as any plurality that can be thought of as unity
did not create any contradiction. It will be argued that the Burali-Fortis paradox is a logical paradox
(para+doxa), which is solvable, whereas, following Saarnio, it is an epistemic antinomy
(anti+nomos) la Kant, which remains unsolvable in principle. Thus, Cantor proposed e.g. in 1899
that the system of all ordinals is an inconsistent, absolutely infinite multiplicity, which is impossible to

Cesare Burali-Forti (1861-1931) was the first mathematician, who made public the paradoxes of
transfinite set theory in his paper Una questione sui numeri transfiniti, (1897). He observed that the
entire succession of all ordinal numbers as a well-ordered set has to have a corresponding ordinal
number , which is greater than the totality of all ordinals represented by . Since was supposed to
contain all ordinal numbers, it could not leave out, and consequently < , which is impossible.

Burali-Forti rejected the comparability of ordinal numbers. He suggested that given two ordinal
numbers and , then it is not always true that at least one of the relations < , = , > must
hold. Bertrand Russell (1872-1970), in turn, noted in his Principles of Mathematics, (1903, 323), that
Burali-Fortis suggestion contradicts a theorem, which Georg Cantor (1845-1918) has established in his
article Beitrge zur Begrndung der transfiniten Mengenlehre,1 where Cantor had proved that for any
two well-ordered sets, corresponding ordinals numbers were necessarily and strictly comparable.
Cantor had also proved that given the whole series of ordinal numbers , every segment of was
well-ordered. Russell preferred to interpret Burali-Fortis paradox of the largest ordinal by rejecting the

## 13, Theorem N, as well as 14, Theorem A, (Cantor 1897, 215, 216),

supposition that the entire set was not well-ordered, although all its segments are well-ordered,
(Russell 1903, 323).

If Russell were correct, then Cantorian set theory would weaken a lot. For example, comparability of
all cardinals and the theorem that every set could be well-ordered were both impossible. One could no
longer assert that every cardinal number was indeed an aleph. However, when producing Principia
Mathematica Russell himself hoped to show by an axiomatic method with increasingly advanced and
complicated details that the major results of Cantorian set theory and especially transfinite numbers
could be established within a consistent framework. The axiomatic approach to set theory was initiated
by Ernst Zermelo (1871-1953). A number of variant systems of axioms have been proposed but that
which has become standard is basically Zermelos original system with modifications introduced by
Abraham Fraenkel (1891-1965) and which is known as Zermelo-Fraenkel set theory.

Uuno Saarnio (1896-1977), in What we know about infinity?, (1969), suggested that the Burali-Fortis
paradox is a logical paradox (para+doxa), which is solvable, whereas it is an epistemic antinomy
(anti+nomos) like Immanuel Kants (1724-1804) first, the cosmological antinomy in Kritik der
reinen Vernunft, (1781/1787, A 426-433/B 454-461), which remains unsolvable in principle by human
intellect. Thus, following Saarnios suggestion, we have to make distinctions between contradiction,
paradox, and antinomy. In contradiction two propositions cannot both be true at the same time, and
neither can they both be false at the same time. In paradox two propositions at first appear to be
contradictory but upon further investigation can be logically or rationally resolved. For example,
according to Saarnio, in logic the Burali-Fortis paradox is solvable by Russells theory of types. In
antinomy two propositions appear to be contradictory and which further investigation cannot rationally
or logically resolve - and yet both are held to be true or both are held to be false. For example in
Kants four antinomies, the first two mathematical antinomies Kant held both to be false, whereas the
third and fourth, the dynamical antinomies he held both to be true.

The reason for interpreting the Burali-Fortis paradox as an epistemic antinomy la Kant is the
following:

According to Kant the understanding can be represented as a faculty for judging, (A 69/B 94). When
we abstract from all content of a judgment in general, and attend only to the mere form of the
understanding it, we find that function of thinking quantity of judgment contains under itself three
moments; universal, particular, and singular, (A 70/B 95). Categories of quantity, in turn, has three
moments; unity, plurality, and totality, (A 80/B 106). Now, the cosmological antinomy follows because
the world is supposed to be given as a bounded whole, a totality, (A 522-527/B 550-555). Similarly, the
Burali-Fortis antinomy follows because all ordinal numbers are supposed to form a bounded whole, a
totality, of which we could then form a set. Totality, following Kantian terminology, is a transcendental
idea, without which we cannot live. It is given to our reason a priori, and by means of it we are able to
form totalities even there, where there isnt one, i.e. an idea, of which no set corresponds.

Cantor, however, had no concern over the paradox, since his conception of a set as any plurality that
can be thought of as unity, defined already as early as in his Grundlagen einer allgemeinen
Mannigfaltikeitslehre, (1883), did not create any contradiction, (see Menzel, 1984). Thus, Cantor, in
(1899), i.e., in his letter to Richard Dedekind (1831-1916), suggested that the system of all order
types2 was an inconsistent, absolutely infinite multiplicity, which is impossible to think as whole
without contradiction. If, on the other hand, the totality of the elements of a multiplicity can be thought
of without contradiction as a bounded whole, i.e. unity, so that they can be gathered together into
one thing, Cantor call it a consistent multiplicity or a set. Accordingly, from Cantors epistemic
criteria for determining whether multiplicities form sets or not, it follows that the Burali-Fortis
paradox is a logical paradox, which is solvable, i.e. there simply isnt the greatest ordinal number,
whereas we have an epistemic antinomy by using the idea of a totality without having the
corresponding set. Thus, Cantor was able to suggest the proposition: Omnia seu finite seu infinita
definita sunt et except Deo ab intellectu determinari possunt. (All things, whether finite or infinite, are
definite and, except for God, can be determined by the intellect).3

An ordinal number was defined by Cantor as the order type of a well-ordered set, (Dauben 1990, 199).
When we are comparing Kant and Cantor, there is some irony, since Cantor didnt like Kant, i.e. that yonder sophistical
Philistine, who was so bad a mathematician as Cantor wrote in his letter to Bertrand Russell (19.9.1911).
3