1 Knowing the Numbers 1. Introduction Physicist Richard Feynman was once asked why he thought he was better at mathematics than most people. He thought about it and said, ‘Other people know about numbers, I know the numbers’. Feynman was claiming knowledge by acquaintance of the numbers. In fact we do not think that Feynman explained his expert knowledge correctly. Feynman was undoubtedly familiar with all kinds of facts about mathematics. However, it is a mistake characteristic of naive, Platonist realism to model knowledge of the numbers on perceptual knowledge of objects. We are not opposed to realism. However, we argue that an adequate realist theory of mathematics must conform with the following constraint: (NKA) We cannot know all of the numbers by acquaintance. Some numbers could perhaps be known by acquaintance. It depends on one’s view of numbers. But it seems undeniable that there are some numbers—typically very large numbers—that we cannot know by acquaintance. More precisely we will be happy to demonstrate the existential claim: (NKA’) There are some numbers that we cannot know by acquaintance. Quine’s empiricist philosophy of mathematics has great strength in maintaining realism alongside with (NKA). Maddy’s realism (Maddy (1990)) denies NKA for small numbers and sets, and essentially adopts a Quinean approach to the higher reaches of mathematics such as transfinite set theory. Naive Platonists deny (NKA), where ‘acquaintance’ is understood as a form of ‘intellectual intuition’. Who is right? What follows below are three arguments against NKA. 2. Arguments against knowledge by acquaintance Let’s illustrate our claim that (NKA) is true with reference to the real numbers. 2.1 Indefinables One way of demonstrating the impossibility of such numbers is to consider all the indefinable real numbers. There are vast swathes of the real number system that are comprised by numbers lacking any discernible pattern that cannot be calculated according to a rule. We must believe in the existence of such random reals on pain of the real number system not behaving as we think it should. In particular, the uncountability of the real numbers depends on there being indefinable real numbers. This follows because the number of definable real numbers is merely countably infinite. 1 Equally it’s obvious that indefinable reals are not (and cannot be) known by 1 The argument for the countability of definable reals: you could theoretically draw up a list of such definables by giving them an ordering (eg. a lexicographical ordering). 2 acquaintance. As the indefinable reals lack any pattern, surveying a finite portion of the decimal expansion of such a real is insufficient to extrapolate to the whole decimal expansion. Such numbers are therefore unsurveyable.2 Moreover, by definition, such numbers do not have names or formulas by which we can refer to them. Indeed, singular reference to indefinables, in the sense of direct reference, is otiose. Such features of indefinables make it extremely unlikely that we have acquaintance with such numbers. 2.2 Part of the structure One of the key insights of structuralism is that mathematicians are largely not concerned with the intrinsic properties of numbers; they are concerned rather with more general structures in which the numbers may figure as positions rather than as complete objects in their own right. 3 Anything that occupies the appropriate role or position in a certain structure may assume the identity of a particular number. Acquaintance is a mode of singular knowledge of an object. Acquaintance with an object is what makes it appropriate to refer to it demonstratively rather than descriptively. In the term ‘this F’, the function of ‘this’ is to an item of experience (an F item of experience given the complex term, ‘this F’). In the case of descriptions such as ‘the F’, reference is not necessarily singular: should anything else be F, the description will fail to refer to one particular item. If the structuralist insight is correct, then mathematicians do not engage in singular reference to the numbers. If the mathematician only cares about the structure of the natural numbers up to the point of isomorphism, then none of the intrinsic properties of the numbers (if they had any) matter. Acquaintance with an object, however, requires presentation of the object to the perceiver. It seems reasonable to assume that acquaintance would provide some awareness of an object’s intrinsic properties. Therefore we can conclude that the mathematician does not aim at acquaintance with the numbers. 2.3 Abstractness and Causality Whatever the nature of the ‘acquaintance’ relation, it seems clear that, like perception, acquaintance requires a causal interaction with an object. If x never interacts causally with object o, then we can infer (if x is the right sort of thing, a conscious being that can be aware of objects) that x is not acquainted with o. So causal interaction is a necessary condition on acquaintance with an object. On one traditional conception of abstract objects, an abstract object stands outside spacetime and is causally inert. Many philosophers have assumed, following Plato, that mathematical objects will be abstract in this traditional sense. Hence, mathematical objects will be causally inert. Consequently, it will be impossible to be acquainted with such objects. One might take this line of reasoning as a reductio ad absurdum of the whole traditional Platonic line of reasoning on which mathematical objects are abstract in the Definitions of ‘knowledge by acquaintance’: cf Russell. Definition of ‘intrinsic property’: F is an intrinsic property of x if x would have F without considering x’s relation to any other object. Intrinsic properties are non-relational properties. 2 3 3 traditional sense.4 It is plain that we do have knowledge (if not knowledge by acquaintance) of mathematical objects. However, if we are willing to allow knowledge by means other than acquaintance (knowledge by inference to the best explanation, for example), then the complete rejection of Platonism could be avoided. The Platonist can simply agree that we are not acquainted with mathematical objects while disagreeing that acquaintance is necessary for knowledge. Conclusion (NKA) holds 4 The line of argument is familiar from Benacerraf (1973), ‘What is mathematical truth?’