[Date Prev][Date Next][Thread Prev][Thread Next]   [Date Index] [Thread Index] [Author Index

The limits of formalization



During the recent arguments over what pure mathematical logic and formalisms can or can't achieve, I found the following passage of my father's to be useful. I might as well post it.
 From page 5 of Life, Itself, Robert Rosen wrote:
 
"Mathematics over the past century has given little evidence that it is concerned with eternal, timeless, and hence unarguable truth. On the contrary, contemporary mathematics is filled with (no pun intended) chaos and turbulence, bespeaking a profound internal instability. Historically, we are still witnessing what (it is hoped) are the transients arising from two profound shocks: the overthrow of Euclid and the discovery of inconsistencies (paradoxes) in Set Theory.
 
To be sure, most practicing mathematicians, like most practical (empirical) scientists, go on about their business, indifferent to such matters, convinced to the depths of their soul about the reliability of what they do; like corks, they believe they will float on calm and troubled waters alike. But I am speaking of foundations, and it is here we shall look.
 
The two great shocks of which I spoke above have coalesced, beginning in the early years of [the 20th century], into a frantic concern with consistency, with a demand that a system of inferential entailments (e.g., set of axioms or production rules, operating on a set of given propositions or postulates) be free of internal or logical contradictions. How can we be sure that a system of entailment, e.g., a mathematical system in the broadest sense, is consistent? I shall be concerned with one particular kind of answer given to this question, an answer championed by David Hilbert, which can be summed up in a single word: formalization.
 
Hilbert and others thought they had traced down the ultimate source of all the difficulties in mathematics. They pointed out that propositions in mathematics are nominally about something; i.e., they have meanings that involve referents outside themselves. Thus, for instance, in Euclid, the word "triangle" is not just an array of letters to be manipulated in a certain way; it refers to a rich and vivid kind of geometric object. And even beyond that, it even refers to things in the phenomenal world. In that sense, any Euclidean proposition containing the word "triangle" can be thought of as describing a percept or quality manifested by this external referent.
 
Thus, according to this analysis, mathematical truth had come to involve two distinct aspects, one pertaining to how we are allowed to manipulate the word "triangle" from one proposition to another, and another pertaining to the actual referents of that word. I will call the former syntactic truth, and the latter, semantic truth. Hilbert and his colleagues argued that it was precisely by allowing semantic truth into mathematics at all (i.e., in the admissibility of regarding a mathematical proposition as the description of a percept or quality of something, allowing a mathematical proposition to refer to something) that all the difficulty arises.
 
Hilbert and his formalistic school actually asserted much more than this. They argued that what we have called semantic truth could always be effectively replaced by more syntactic rules. In other words, any external referent, and any quality thereof, could be pulled inside a purely syntactic system. By a purely syntactic system, they understood: (1) a finite set of meaningless symbols, an alphabet; (2) a finite set of rules for combining these symbols into strings or formulas; (3) a finite set of production rules for turning given formulas into new ones. In such a purely syntactic system, consistency is guaranteed.
 
[ Statement by Kleene on formal axiomatics, which I will not bother to type in]
 
The idea that all truth can be expressed as pure syntactic truth, which is the essence of the formalist position in mathematics, I claim to be the formal analog of "hardness" and quantitation in science The formalist position is, first of all, an _expression_ of a belief that all mathematical truth can be reduced to, or expressed in terms of, word processing or symbol manipulation. Hence the close association of formalization with the idea of "machines" (Turing machines) and with the idea of algorithms. These embody purely automatic procedures, which require no thought, no perception, indeed, no external agency at all.
 
Second, the formalist position, that the universe of discourse needs to consist of nothing more than meaningless symbols pushed around by definite rules of manipulation, is exactly parallel to the mechanical picture of the phenomenal world as consisting of nothing more than configurations of structureless particles, pushed around by impressed forces.
 
The formalist position seems, on the face of it, very attractive. For, by asserting that all truth is syntactic truth, it tells us that (1) we lost no shred of mathematical truth in the process of formalization, and (2) we are automatically guaranteed that mathematics is consistent. We pay for these benefits by giving up the idea that mathematics is "about" anything, i.e., that its propositions express percepts or qualities, but on the other hand, we are informally free to interpret these propositions in any way we want. These are, of course, exactly the same attractions that the "hard" or quantitative sciences offer in the phenomenal world.
 
Gödel's Theorem
 
The celebrated Incompleteness Theorem of Gödel effectively demolished the formalist program. Basically, he showed that, no matter how one tries to formalize a particular part of mathematics (Number Theory, perhaps the inmost heart of mathematics itself), syntactic truth in the formalization does not coincide with (is narrower than) the set of truths about numbers.
 
There are many ways to look at Gödel's Theorem. Indeed, the Theorem itself has provoked an enormous literature, as might be expected. For our purposes, we may regard it as follows: one cannot forget that Number Theory is about numbers. The fact that Number Theory is about numbers is essential, because there are percepts or qualities (theorems) pertaining to numbers that cannot be expressed in terms of a given, preassigned set of purely syntactic entailments. Stated contrapositively: no finite set of numerical qualities, taken as syntactical basis for Number Theory, exhausts the set of all numerical qualities. There is always a purely semantic residue, that cannot be accommodated by that syntactical scheme.
 
Gödel's Theorem shows that formalizations are part of mathematics, but not all of mathematics. Mathematics, like language itself, cannot be freed of all referents and remain mathematics. Any attempt to do this (i.e., any attempt to capture every percept through a formalization of any finite set of percepts) must already fail in the Theory of Numbers.
 
On the other hand, Number Theory is still mathematics, still a system of inferential entailment in itself. It is only that it is not a purely syntactic system, not entirely a matter of word processing or symbol manipulation, independent of any external referent. In other words, Number Theory is not a closable, finite system of inferential entailment. These facts, as embodied in Gödel's Theorem, do not make us give up Number Theory as part of mathematics nor even give up formalization as a strategy for studying certain kinds of mathematical systems. They express rather the limitations of formalization; it is not, as Hilbert thought, a universal strategy."
End of excerpt.