mathematics, what is mathematics, effectiveness, methodology, philosophy, foundations, formalism, Platonism, intuition, nature, theories, axiomatic, formal, theory, Hilbert, program, twin prime conjecture
Back to title page.
by Karlis Podnieks, Dr.Math.
[email protected]
University of Latvia
Institute of Mathematics and Computer Science
This paper presents the translation of the first chapter of my book "Around Goedel's theorem" published 1992 in Russian. View russian original.
CONTENTS
1. Platonism - the philosophy of working
mathematicians
2. Investigation of stable (self-contained)
models - the nature of the mathematical method
3. Intuition and axioms
4. Formal theories
5. Hilbert's program
6. Some replies to critics
7. References
Charles Hermite has said once he is convinced that numbers and functions are not mere inventions of mathematicians, that they do exist independently of us, as do things in our everyday practice. Some time ago in the former USSR this proposition was quoted as the evidence for "the naive materialism of outstanding scientists".
Still, such propositions stated by mathematicians are evidences not for their naive materialism, but for their naive Platonism. As will be shown below, Platonist attitude of mathematicians to objects of their investigationsis determined by the very nature of the mathematical method.
First let us consider "Platonism" of Plato (427-347 BC) itself (click here for other Plato web-sites, and here, and here - to read about Plato's lifetime in the history of Greece). The particular form of Plato's system of philosophy was determined by Greek mathematics.
In VI-V centuries BC the evolution of Greek mathematics led to mathematical objects in the modern meaning of the word: the ideas of numbers, points, straight lines etc. stabilized, and thus they were detached from their real source - properties and relations of things in the human practice. In geometry, straight lines have zero width, and points have no size at all. Such things actually do not exist in our everyday practice. Here, instead of straight lines we have more or less smooth stripes, instead of points - spots of various forms and sizes. Nevertheless, without this passage to an ideal (partly fantastic, yet simpler, stable, fixed, self-contained) "world" of points, lines etc., the mathematical knowledge would have stopped at the level of art and never would become a science. Idealization allowed to create an extremely effective instrument - Euclidean geometry.
The concept of natural numbers (0, 1, 2, 3, 4, ...) developed from human operations with collections of discrete objects. This development process was completed already in VI century BC, when somebody asked: how many prime numbers do there exist? And the answer was found by means of reasoning - there are infinitely many prime numbers. Clearly, it is impossible to verify such an assertion empirically. Still, by that time the concept of natural number was already stabilized and detached from its real source - the quantitative relations of discrete collections in the human practice, and it began to work as a stable (self-contained) model. The system of natural numbers is an idealization of these quantitative relations. People abstracted it from their experience with small collections (1, 2, 3, 10, 100, 1000 things). They extrapolated their rules onto much greater collections (millions of things) and thus idealized the real situation (and even deformed it - see Rashevsky [1973] and van Dantzig [1955]).
Note. The idea of the Infinite deviates significantly from the situation in the physical Universe - this fact was clearly stated already in the famous June 4, 1925 lecture of David Hilbert "On the Infinite":
D.Hilbert. Ueber das Unendliche. "Math. Annalen", 1925, Vol.95, pp.161-190 (see also van Heijenoort [1967])
My example: let us consider "the number of atoms in this sheet of paper". From the point of common arithmetic this number "must" be either even or odd at any moment of time. In fact, however, the sheet of paper does not possess any precise "number of atoms" (because of, for example, spontaneous nuclear reactions). Moreover, the modern cosmology claims that the "total number" of particles in the Universe is less than 10200. What should be then the real meaning of the statement "10200+1 is an odd number"? Hence, in arithmetic not only practically useful algorithms are discussed, yet also a kind of pure fantastic matter without any direct real meaning. (Rashevsky proposed to develop a new kind of arithmetic allowing a more adequate treatment of large natural numbers.) Of course, Greek mathematicians could not see all that so clearly. Discussing the amount of prime numbers they believed that they are discussing objects as real as collections of things in their everyday practice.
Thus, the process of idealization ended in stable, fixed, self-contained concepts of numbers, points, lines etc. These concepts ceased to change and were commonly acknowledged in the community of mathematicians. And all that was achieved already in V century BC. Since that time our concepts of natural numbers, points, lines etc. have changed very little. The stabilization of concepts testifies their detachment from real objects that have led people to these concepts and that are continuing their independent life and contain an immense variety of changing details. When working in geometry, a mathematician does not investigate the relations of things of the human practice (the "real world" of materialists) directly, he investigates some stable notion of these relations - an idealized, fantastic "world" of points, lines etc. And during the investigation this notion is treated (subjectively) as the "last reality", without any "more fundamental" reality behind it. If during the process of reasoning mathematicians had to remember permanently the peculiarities of real things (their degree of smoothness etc.), then instead of a science (effective geometrical methods) we would have an art - simple, specific algorithms obtained by means of trial and error or on behalf of some elementary intuition. Mathematics of Ancient Orient stopped at this level. Greeks went further.
(See online paper "Babylonian and Egyptian mathematics" in the MacTutor History of Mathematics archive).
Studying mathematics Plato came to his surprising philosophy of two worlds: the "world of ideas" (strong and perfect as the "world" of geometry) and the world of things. According to Plato, each thing is only an imprecise, imperfect implementation of its "idea" (which does exist independently of the thing itself in the world of ideas). Surprising and completely fantastic is Plato's notion of the nature of mathematical investigation: before a man is born, his soul lives in the world of ideas and afterwards, doing mathematics he simply remembers what his soul has learned in the world of ideas. Of course, this is an upside-down notion of the real situation. Plato treats the end product of the evolution of mathematical concepts - a stable, self-contained system of idealized objects as an independent beginning point of the evolution of the "world of things".
Nevertheless, being an outstanding philosopher, Plato tried to explain (in his own manner) those aspects of the human knowledge, which remained inaccessible to other philosophers of his time. To explain the real nature of mathematical objects, Greeks had insufficient knowledge in physics, biology, human physiology and psychology, etc.
Today, any philosophical position treating ideal objects of human thought as a specific "world", should be called Platonism. Particularly, the philosophy of working mathematicians is a Platonist one. Platonist attitude to objects of investigation is inevitable for a mathematician: during his everyday work he is used to treat numbers, points, lines etc. as the "last reality", as a specific "world". This sort of Platonism is an essential aspect of mathematical method, the true source of the surprising efficiency of mathematics in the natural sciences and technology. It explains also the inevitability of Platonism in the philosophical position of mathematicians (having, as a rule, very little experience in philosophy). Habits, obtained in the everyday work, have an immense power. Therefore, when a mathematician, not very strong in philosophy, tries to explain "the nature" of his mathematical results, he unintentionally brings Platonism into his reasoning. The reasoning of mathematicians about the "objective nature" of their results is, as a rule, rather a kind of "objective idealism", and not materialism.
Whether your own philosophy of mathematics is Platonism or not, can be easily determined by using the following test. Let us consider the twin prime number sequence:
(3, 5), (5, 7), (11, 13), (17, 19), (29, 31), (41, 43),...
(two prime numbers are called twins, if their difference is 2). It is believed (at least since 1849 - see http://forum.swarthmore.edu/epigone/math-history-list/thahtwecha) that there are infinitely many twin pairs (the famous twin prime conjecture), yet the problem remains unsolved up to day. Suppose that it will be proved undecidable from the axioms of set theory. Do you believe that, still, the twin prime conjecture possesses an "objective truth value"? Imagine you are moving along the natural number system:
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,...
And you meet twin pairs in it from time to time: (3, 5), (5, 7), (11, 13), (17, 19), (29, 31), (41, 43), ... It seems there are only two possibilities:
a) We achieve the last pair and after that moving forward we do not meet any twin pairs (i.e. the twin prime conjecture is false),
b) Twin pairs appear over and again (i.e. the twin prime conjecture is true).
It seems impossible to imagine a third possibility...
If you think so, you are, in fact, a Platonist. You are used to treat the natural number system as a specific "world", very like the world of your everyday practice. You are used to think that any sufficiently definite assertion about things in this world must be either true or false. And, if you regard natural number system as a specific "world", you cannot imagine a third possibility: maybe, the twin prime conjecture is neither true nor false. Still, such a possibility would not surprise us if we would realize (following Rashevsky [1973]) that natural number system contains not only some information about real things of the human practice, but it also contains many elements of fantasy. Why do you think that fantastic "worlds" (a kind of Disneyland) always are completely perfect?
"Does the Bernay's number 67**275**729 [** stands for exponentiation - K.P.] actually belong to every set which contains 0 and is closed under the successor function? The conventional answer is yes yet we have seen that there is a very large element of fantasy in conventional mathematics which one may accept if one finds it pleasant, but which one could equally sensibly (perhaps more sensibly) reject." (Parikh [1971], p.507).
As another striking example of Platonist approach to nature of mathematics let us consider an expression by N.N.Luzin from 1927 about the continuum problem (quoted after Keldysh [1974]):
"The cardinality of continuum, if it is thought to be a set of points, is some unique reality, and it must be located on the aleph scale there, where it is. It's not essential, whether the determination of the exact place is hard or even impossible (as might have been added Hadamard) for us, men".
The continuum problem was formulated by Georg Cantor in 1878: does there exist a set of points with cardinality greater than the cardinality of natural numbers (the so called countable cardinality) and less than the cardinality of the continuum (i.e. of the set of all points of a line)? In the set theory (using the axiom of choice) one can prove that the cardinality of every infinite set can be measured by means of the so-called aleph scale:
aleph0 |----| aleph1 |----| aleph 2|----| ... |----| alephn |----| alephn+1 |----| ... |----| alephw |----| ...
Here aleph0 is the countable cardinality, aleph1 - the least uncountable cardinality etc., and alephw follows after all alephn with natural number n.
Cantor established that aleph0 < c (c denotes the cardinality of the entire continuum), and then he conjectured that c = aleph1. This conjecture is called continuum hypothesis. Long-drawn efforts by Cantor itself and by many other outstanding people did not lead to any solution of the problem. In 1905 J.Koenig proved that c is not equal to alephw, and that is almost all we know today ...
Now we know that the continuum problem is undecidable, if we are using commonly acknowledged axioms of set theory. Kurt Goedel in 1939 and Paul Cohen in 1963 proved that one could assume safely (i.e. without contradiction) any of the following "axioms":
c = aleph1 , c = aleph2 , c = aleph3 , ...
And even (a joke by N.N.Luzin, see Keldysh [1974]): c = aleph17. Thus, the axioms of set theory do not allow determining the exact place of c on the aleph scale, although we can prove that (Ex) c = alephx, i.e. that c "is located" on this scale.
The Platonist, looking at the picture of the aleph scale, tries to find the exact place of c ... visually! He cannot imagine a situation when a point is situated on a line, yet it is impossible to determine the exact place. This is a normal Platonism of a working mathematician. It stimulates investigation even in the most complicated fields (we never know before whether some problem is solvable or not). Still, if we pass to methodological problems, for example, to the problem of the "meaning" of Cohen's results, we should keep off our Platonist habits. Do we think that, in spite of the undecidability of the continuum problem "for us, men", some "objective", "real" place for the cardinality of the continuum on aleph scale does exist? If yes, then we assume something like Plato's "world of ideas" - some fantastic "world of sets", which exists independently of the axioms used in reasoning of mathematicians. At this moment the mathematical Platonism (a pure psychological phenomenon) has converted into a kind of philosophy. Such people say that the axioms of set theory do not reflect the "real world of sets" adequately, that we must search for more adequate axioms, and even - that no particular axiom system can represent the "world of sets" precisely. They pursue a mirage, of course, no "world of sets" can exist independently of the principles used in its investigation.
The real meaning of Cohen's results is very simple. We have established that (Ex) c = alephx , yet it is impossible to determine the exact value of x. It means that the traditional set theory is not perfect and, therefore, we may try to improve it. And it appears that one can choose between several possibilities.
For example, we can postulate the axiom of constructibility (see Jech [1971], Devlin [1977]). Then we will be able to prove that c = aleph1 , and to solve some other problems that are undecidable in the traditional set theory.
We can postulate also a completely different axiom - the axiom of determinateness (see Kleinberg [1977]). Then we will be forced to reject the axiom of choice (in its most general form) and as a result we will be able to prove that every set of points is Lebesgue-measurable, and that the cardinality of continuum is incompatible with alephs (except of aleph0). In this set theory continuum hypothesis can be proved in its initial form: every infinite set of points is either countable or has the cardinality of the continuum.
Both directions (the axiom of constructibility and the axiom of determinateness) have produced a plentiful collection of beautiful and interesting results. These two set theories are at least as "good" as the traditional set theory, yet they contradict each other, therefore we cannot speak here about the convergence to some unique "world of sets".
Our main conclusion is the following: everyday work is permanently moving mathematicians to Platonism (and, as a creative method, this Platonism is extremely efficient), still, passing to methodology we must reject such a philosophy deliberately. Many essays on philosophy of mathematics disregard this problem.
The term "model" will be used below in the sense of applied mathematics, not in the sense used in mathematical logic (i.e. we will discuss models intended to "model" natural processes or technical devices, not sets of formulas).
Following the mathematical approach of solving some (physical, technical etc.) problem, one tries "to escape the reality" as fast as possible, passing to investigation of a definite (stable, self-contained) mathematical model. In the process of formulating a model one asks frequently: Can we assume that this dependency is linear? Can we disregard these deviations? Can we assume that this partition of probabilities is normal? It means that one tends (as fast as possible and using a minimum of postulates) to formulate a mathematical problem, i.e. to model the real situation in some well known mathematical structure (or to create a new structure). Solving the mathematical problem one hopes that, in spite of the simplifications made in the model, he will obtain some solution of the original (physical, technical etc.) problem.
After mathematics has appeared, all scientific theories can be divided into two classes:
a) Theories based on a developing system of principles,
b) Theories based on a stable system of principles.
In the process of development theories of class (a) are enriched with new basic principles that do not follow from the principles acknowledged before. Such principles arise due to fantasy of specialists, supported by more and more perfect experimental data. The progress of such theories is first of all in this enrichment process.
On the other hand, in mathematics, physics and, at times, in other branches of science one can find theories, whose basic principles (postulates) do not change in the process of their development. Every change in the set of principles is regarded here as a passage to a new theory. For example, A.Einsteins's special relativity theory can be regarded as a refinement of the classical mechanics, a further development of I.Newton's theory. Still, since both theories are defined very precisely, the passage "from Newton to Einstein" can be regarded also as a passage to a new theory. The evolution of both theories is going on today: new theorems are proved, new methods and algorithms are developed etc. Nevertheless, both sets of basic principles remain constant (such as they were at the lifetime of their creators).
Stable (i.e. self-contained) system of basic principles is the distinctive character of mathematical theories. A mathematical model of some natural process or technical device is essentially a stable model that can be investigated independently of its "original" (and, thus, the similarity of the model and the "original" is only a limited one). Only such models can be investigated by mathematicians. Any attempt to refine the model (to change its definition in order to obtain more similarity with the "original") leads to a new model that must remain stable again, to enable a mathematical investigation of it.
Hence, mathematical theories are "the part of science you could continue to do if you woke up tomorrow and discovered the universe was gone" (I do not know the author of this elegant definition put on the web by Dave Rusin).
Working with stable (self-contained) models mathematicians have learned to draw maximum of conclusions from a minimum of premises. This is why mathematical modeling is so efficient.
It is very important to note that a mathematical model (because it is stable) is not bound firmly to its "original". It may appear that some model is constructed badly (in the sense of the correspondence to the "original"), yet its mathematical investigation goes on successfully. Since a mathematical model is defined very precisely, it "does not need" its "original". One can change some model (obtaining a new model) not only for the sake of the correspondence to "original", but also for a mere experiment. In this way we easily obtain various models (and entire branches of mathematics) that do not have any real "originals". The stable (self-contained) character of mathematical models makes such deviations possible and even inevitable.
The mathematical method is (by definition) investigation of stable models. What is then mathematics itself? Models can be more or less general (let us compare, for example, arithmetic of natural numbers, the relativity theory and some model of the solar system). Very specific models will be investigated more successfully under the management of specialists who are creating and using them. A combination of special experience with sufficient experience in mathematics (in one person or in a team) will be here the most efficient strategy. Investigation of more general models that can be applied to many different specific models draws up contents of a specific branch of science that is called mathematics.
For example, the Calculus has many applications in various fields and, therefore, it is a striking example of a theory that undoubtedly belongs to mathematics. On the other hand, a model of solar system (used, for example, for exact prediction of eclipses) is too specific to be encountered as a part of mathematics (although it is surely a mathematical model).
The stable character of mathematical models and theories is simultaneously the force and the weakness of mathematics. The ability of mathematicians to obtain maximum of information from minimum of premises has demonstrated its efficiency in science and technique many times. Still, the other side of this force is weakness: no particular stable (self-contained) model (theory) can solve all problems arising in science (or even in mathematics itself). An excellent confirmation of this thesis was given in K.Goedel's famous incompleteness theorem.
And there is one more weakness. Mathematics, being detached from real problems of other fields, controlled only by its "internal needs", is getting more and more uncontrollable. Theories and whole branches of mathematics are developed that do not have and cannot have any applications to real problems. Polish writer Stanislav Lem joked in his book "Summa Technologiae": a mathematician is like a mad tailor: he is making "all possible clothes" and hopes to make also something suitable for dressing.... As we have seen this problem is due to the very nature of the mathematical method. No other branch of science knows such problems.
Mathematicians have learned the ability "to live" (literally!) in the world of mathematical concepts and even (while working on some particular problem) - in a very specific "world" of a particular model. Investigation of models is mathematician's goal for goal's sake, during their work they disregard the existence of the reality behind the model. Here we have the main source of the creative power of mathematics: in this way, "living" (sometimes, for many years) in the "world" of their concepts and models, mathematicians have learned to draw maximum of conclusions from a minimum of premises.
After one has formulated some model, it usually appears that in mathematics some work has already been done on the problem, and some methods or even algorithms have been created. This allows drawing many important conclusions about the model in real time. Clearly, if the model looks so specific that no ready mathematical means can be found to investigate it, the situation becomes more complicated. Either the model is not good enough to represent a really interesting fragment of the "reality" (then we must look for another model), or it is so important that we may initiate investigations to obtain the necessary new mathematical methods.
The key to all these possibilities is mathematical Platonism - the ability of mathematicians "to live" in the "worlds" of the models they do investigate, the ability to forget all things around them during their work. In this way some of them have got the ill fame of being "rusks", queer customers, etc. Thus we can say that Platonism is in fact the psychology of working mathematicians (and that it is a philosophy only from their subjective point of view).
The above stated picture of the nature of mathematics is not yet commonly acknowledged. Where is the problem, why it is so hard to regard mathematical theories as stable (self-contained) models? A personal communication by S. Lavrov from 1988: " ... Theorems of any theory consist, as a rule, of two parts - the premise and the conclusion. Therefore, the conclusion of a theorem is derived not only from a fixed set of axioms, but also from a premise that is specific to this particular theorem. And this premise - is it not an extension of the fixed system of principles? ... Mathematical theories are open for new notions. Thus, in the Calculus after the notion of continuity the following connected notions were introduced: break points, uniform continuity, Lipschitz's conditions, etc. ... All this does not contradict the thesis about fixed character of principles (axioms and rules of inference), yet it does not allow "working mathematicians" to regard mathematical theories as fixed ones."
Barwise J. [1977]
Handbook of Mathematical Logic, Elsevier Science Ltd., 1977, 1166 pp. (Russian translation available)
Burris S.N. [1998]
Logic for Mathematics and Computer Science, Prentice Hall, 1998, 425 pp. (see also the exciting online Supplementary Text to this book)
Is 10**10**10 a finite number? "Dialectica", 1955, vol.9, N 3/4, pp.273-278
Davis M. [1965]
The Undecidable. Basic papers on undecidable propositions, unsolvable problems and computable functions. - Raven Press, New York, 1965, 440 pp.
Devlin K.J. [1977]
The axiom of constructibility. A guide for the mathematician. "Lecture notes in mathematics", vol. 617, Springer-Verlag, Berlin - Heidelberg - New York, 1977, 96 pp.
Edwards H.M. [1977]
Fermat's Last Theorem. A Genetic Introduction to Algebraic Number Theory. Springer-Verlag, 1977 (Russian translation available)
Feferman S. [1960]
Arithmetization of metamathematics in a general setting. "Fund. Math.", 1960, vol. 49, pp.35-92
Fraenkel A.A., Bar-Hillel Y., Levy A. [1973]
Foundations of Set Theory. Studies in Logic, Vol. 67, Elsevier Science Ltd, 1973, 404 pp. (Russian translation available)
Graham R.L. [1981]
Rudiments of Ramsey theory. AMS, Providence, 1981 (Russian translation available)
Hadamard J. [1945]
An essay on the psychology of invention in the mathematical field. Princeton, 1945, 143 pp. (Russian translation available)
Metamathematics of First-order Arithmetic. Springer Verlag Perspectives in Mathematical Logic, 1993, 460 pp.
The Axiomatization of Arithmetic. "J. Symb. Logic", 1957, Vol. 22, N2, pp.145-158
From Frege to G�del: A Source Book in Mathematical Logic, 1879-1931. Harvard University Press, 1967
Hilbert D., Bernays P. [1934]
Grundlagen der Mathematik. Vol. I, Berlin, 1934, 471 pp. (Russian translation available)
Jech T.J. [1971]
Lectures in set theory with particular emphasis on the method of forcing. "Lecture Notes in Mathematics", vol. 217, Springer-Verlag, Berlin - Heidelberg - New York, 1971 (Russian translation available)
Suslin's hypothesis is incompatible with V=L. "Notices Amer. Math. Soc.", 1968, vol.15, p.935
Kanovei V. [1984]
Axiom of choice and axiom of determinateness. "Nauka Publishers", Moscow, 1984, 64 pp. (in Russian)
The ideas of N.N.Luzin in the descriptive set theory. "Uspekhi matematicheskih nauk", 1974, vol.29, n5, pp. 183-196 (in Russian)
Kleene S.C. [1952]
Introduction to Metamathematics. Van Nostrand, 1952 (Russian translation available)
Kleinberg E.M. [1977]
Infinitary combinatorics and the axiom of determinateness. "Lecture notes in mathematics", vol. 612, Springer-Verlag, Berlin - Heidelberg - New York, 1977, 150 pp.
L�vy P. [1926]
Sur le principe du tiers exclu et sur les th�oremes non susceptibles de d�monstration. "Revue de M�taphysique et de Morale", 1926, vol. 33, N2, pp. 253-258.
Development of Set Theory in the XIX Century. Nauka Publishers, Moscow, 1965, 350 pp. (in Russian)
The Early History of the Axiom of Choice. Nauka Publishers, Moscow, 1982, 304 pp. (in Russian)
Introduction to Mathematical Logic. Fourth Edition. International Thomson Publishing, 1997, 440 pp. (Russian translation available)
Mostowski A. [1969]
Constructibles sets with applications. North-Holland, 1969 (Russian translation available)
Existence and Feasibility in Arithmetic. JSL, 1971, Vol.36, N.3, pp. 494-508
The double incompleteness theorem. Proc. Latvian State University, 1975, Vol. 233, pp. 191-200 (in Russian).
The double incompleteness theorem. Abstracts IV All-union conference on mathematical logic, Kishinev, 1976, p. 118 (in Russian)
Platonism, Intuition and the Nature of Mathematics. "Heyting'88. Summer School & Conference on Mathematical Logic. Chaika, Bulgaria, September 1988. Abstracts.", Sofia, Bulgarian Academy of Sciences, 1988, pp. 50-51.
Platonism, Intuition and the Nature of Mathematics. Riga, Latvian State University, 1988, 23 pp. (in Russian).
Around Goedel's theorem. Latvian State University Press, Riga, 1981, 105 pp. (in Russian). 2nd edition: "Zinatne", Riga, 1992, 191 pp. (in Russian).
Poincare H. [1908]
Science et methode. Paris, 1908, 311 pp. (Russian translation available)
On the dogma of the natural number system. "Uspekhi matematicheskih nauk", 1973, vol.28, n4, pp. 243-246 (in Russian)
Hilbert. Springer-Verlag, 1996 (Russian translation available)
Solovay R., Tennenbaum S. [1971]
Iterated Cohen extensions and Suslin's problem. "Annals of Math.", 1971, vol.94, pp.201-245
Formation of the Mathematical Logic. Nauka Publishers, Moscow, 1967, 400 pp. (in Russian, see also the English translation: Styazhkin, N. I. History of Mathematical Logic from Leibniz to Peano. MIT Press, Cambridge, MA, 1969)
mathematics, what is mathematics, effectiveness, methodology, philosophy, foundations, formalism, Platonism, intuition, nature, theories, axiomatic, formal, theory, Hilbert, program, twin prime conjecture
Back to title page.