Gödel's intuitions about the continuum

P₂. Gödel cites both the first edition, (Warsaw: Garasinski), 1934, and the second, (New York: Chelsea), 1956.
¹⁶ The continuum hypothesis implies that there is an ordering of the real numbers in which for each x there are only countably many y less than x. The axiom of choice allows us to pick for each x a function h_x from the natural numbers onto the set of such y. Then we may define functions f_n(x) = h_x(n), and the graphs of these functions, plus their reflections in the diagonal y = x, plus the diagonal itself, give countably many "generalized curves" filling the plane.
¹⁷ Even Mandelbrot's more expansive conception of what is intuitive seems to take in only F_ or G_ or anyhow low-level Borel sets (to which classifications his "fractals" all belong), not arbitrary "generalized curves."
¹⁸ "What is Cantor's continuum problem" p. 271. This passage comes from the supplement added to the second version of the paper.
¹⁹ See Mathematical Thought and Its Objects, p. 8. This book has had a greater influence on the present paper than will be evident from my sporadic citations of it. Inversely, Parsons holds, as a consequence of his structuralism, that we can have an intuition that every natural number has a successor, though we have no intuition of natural numbers. See Mathematical Thought and Its Objects, §37 "Intuition of numbers denied," pp. 222-224.
²⁰ The most plausible account to date of how and in what sense we might be said to perceive sets is that of Penelope Maddy in Realism in Mathematics (Oxford: Oxford University Press), 1990, especially chapter 2 , "Perception and Intuition." But on this account set-theoretic perception is mainly of small sets of medium-sized physical objects, just as sense-perception is mainly of medium-sized physical objects themselves. The theoretical extrapolation to infinite sets then seems to have the same status as the theoretical extrapolation to subvisible physical particles, and this would seem to leave the axiom of infinity with the same status as the atomic hypothesis: historically a daring conjecture, which by now has led to so much successful theorizing that we can hardly imagine doing without it, but still not something that "forces itself upon us."
²¹ The passage comes from §3 of the paper (p. 262) and leads into Gödel's exposition of the cumulative hierarchy or iterative conception of set (which is what the phrase "the way sketched below" in the quotation refer to).
²² "Truth and Proof: The Platonism of Mathematics," Synthese, vol. 69 (1986), pp. 341-370. See note 3, pp. 364-365.
²³ "Gödel's conceptual realism," Bulletin of Symbolic Logic, vol. 2 (2005), pp. 207-224. I will not be doing justice to this study, which would require extended discussion of structuralism. In particular I will not be discussing what real difference, if any, there would be between perceiving the structure of the universe of sets as Martin understand it and perceiving the concept of set as Gödel understands concepts. (Both are clearly different from perceiving the individual sets that occupy positions in the structure and exemplify the concepts.)
²⁴ See "Platonism and mathematical intuition in Kurt Gödel's thought," Bulletin of Symbolic Logic, vol. 1 (1995), pp. 44-74, where he discusses the passage at issue on p. 65. In helpful comments on a preliminary version of the present study, Parsons remarks, "One piece of evidence … is that Gödel frequently talks [elsewhere] of perception of concepts but hardly at all about perception or intuition of sets. It may be that any perception of sets that he would admit is derivative from perception of concepts," here alluding to the suggestion made in footnote 43 of the cited paper that those sets, such as the ordinal  that individually definable may be "perceived" by perceiving the concepts that identify them uniquely — though, of course, what it identifies uniquely is really only the position of the ordinal in the set-theoretic universe.
²⁵ See Parsons, Mathematical Thought and Its Objects, §52 "Reason and 'rational intuition'" for some healthy skepticism about the appropriateness of this traditional term.
²⁶ Diogenes Laertius, with English translation by R. D. Hicks, Lives of Eminent Philosophers, Loeb Classical Library (Cambridge: Harvard University Press), 1925, Book VI, Diogenes, p. 55.
²⁷ In particular, Kai Hauser in a talk at the 2009 NYU conference in philosophy of mathematics cited as evidence of Husserlian influence the following somewhat concessive passage (which has also drawn the attention of earlier commentators):

However, the question of the objective existence of the objects of mathematical intuition … is not decisive for the problem under consideration here. The mere psychological fact of the existence of an intuition which is sufficiently clear to produce the axioms of set theory and an open series of extensions of them suffices to give meaning to the question of the truth or falsity of propositions like Cantor's continuum hypothesis. (penultimate paragraph of the supplement, p. 272)

²⁸ "Russell's mathematical logic," in Benacerraf & Putnam, pp. 221-232, with the quoted passage on pp. 215-216. Gödel's "Platonism" or "realism" is nearly as evident in this work as in the continuum problem paper. Parsons, in correspondence, while agreeing that Gödel acknowledged the fallibility of rational intuition, and emphasizing that in so acknoweldging Gödel was departing from the earlier rationalist tradition, nonetheless warns against reading too much into the quoted passage, on the grounds that Gödel's usage of "intuition" may have been looser than at the time of the Russell paper than it later became.
²⁹ The documents (two notes and an unsent letter by Gödel), and an informative discussion of the unedifying episode by Robert Solovay, can be found in Collected Works, vol. III, pp. 405-425. Another example of the fallibility of intuition may perhaps be provided by the fact mention by Solovay, that the pioneering descriptive set theorist Nikolai Luzin, who disbelieved CH, connected his disbelief with "certainty" that every subset of the reals of size ₁ is coanalytic. We now know, however, that assuming a measurable cardinal, if CH fails then no set is of size ₁ is coanalytic (since assuming a measurable cardinal, every coanalytic set is either countable or of the power of the continuum).
³⁰ His formulations, however, in "What is Cantor's continuum problem?" p. 264, footnote 20 and the text to which it is attached, are rather cautious, and he mentions on the next page that "there may exist … other (hitherto unknown) axioms."
³¹ In §3 "Restatement of the problem…" or in other expositions of the same kind, several of which can be found in §IV "The concept of set" of the second edition of Benacerraf & Putnam. Note, however, that two of the contributors there, George Boolos ("The iterative conception of set," pp. 486-502) and Charles Parsons ("What is the iterative conception of set?" pp. 503-529) in effect deny the reality of Gödelian experiences, deny that the axioms do "force themselves upon us." They do so also in other works (Boolos in "Must we believe in set theory?" in Logic, Logic, and Logic (Cambridge: Harvard University Press), 1998, pp. 120-132. Parsons in Mathematical Thought and Its Objects, §55 "Set theory," pp.338-342). In this paper I will not debate this point, but will simply grant for the sake of argument that Gödel is right and in fact there occurs such a phenomenon as the axioms "forcing themselves upon one." The issue I wish to discuss is, granting that in fact such experiences occur, whether we need to posit rational intuition to explain their occurrence.
³² The kind of view I am attributing to Gödel resembles the kind of view Tyler Burge attributes to Frege. See "Frege on sense and linguistic meaning," in Truth, Thought, Reason (Oxford: Clarendon Press), 2005, pp. 242-269. Frege sometimes says that everyone has a grasp of the concept of number and sometimes says that even very eminent mathematicians before him lacked a sharp grasp of the concept of number. Burge proposes to explain Frege's speaking now one way, now the other, by suggesting that Frege distinguishes the kind of minimal grasp of the associated concept possessed by anyone who knows the fixed, conventional linguistic meaning of an expression, with the ever sharper and sharper grasp to which not every competent speaker of the language, by any means, can hope to achieve.
³³ Something like the contrast I have been trying to describe was, I suspect, ultimately the issue between Gödel and Carnap, but examination of that relationship in any detail is out of the question here. A complication is that Gödel sometimes uses "meaning" related terms in idiosyncratic senses, so that he ends up saying that mathematics is "analytic" and thus sounding like Carnap, though he doesn't at all mean by "analytic" what Carnap would. Martin and Parsons both discuss examples of this usage.
³⁴ It would be very difficult to formulate any such new axiom about extreme rarity, since nothing is more common in point-set theory than to find that sets small in one sense are large in another. Right at the beginning of the subject comes the discovery of the Cantor set, which is small topologically (first category) and metrically (measure zero), but large in cardinality (having the power of the continuum). Another classic result is that the unit interval can be written as the union of a first category set and a measure zero set. See John C. Oxtoby, Measure and Category (Berlin: Springer), 1971, for more information (The particular result just cited appears as Corollary 1.7, p. 5.) The difficulty of finding a rigorous formulation, however, is only to be expected with dim and misty rational intuitions.
³⁵ Here "something of the sort" may be taken to cover the suggestion of looking for some sort of maximal principle, made in footnote 23, p. 266. Gödel also mentions (p. 265) the possibility of justifying a new axiom not by rational intuitions in its favor, but by verification of striking consequences. Gödel cites no candidate example and even today it is not easy to think of one, if one insists that the striking consequences be not just æsthetically pleasing, like the pattern of structural and regularity properties for projective sets that follow from the assumption of projective determinacy, but verified. The one case I can think of is Martin's proof of Borel determinacy (as a corollary of analytic determinacy) assuming a measurable cardinal before he found a more difficult proof without that assumption. And in this example the candidate new axiom supported is still a large cardinal axiom.
³⁶ To be sure, in the wake of Cohen's work, Azriel Levy and Solovay showed that no solution to the continuum problem is to be expected from large cardinal axioms of a straightforward kind. (See their "Measurable cardinals and the continuum hypothesis," Israel Journal of Mathematics, vol. 5 (1967), pp. 233-248.) But the present-day Woodin program can nonetheless be considered as in a sense still pursuing the direction to which Gödel pointed. According to Woodin's talk at the 2009 NYU conference in philosophy of mathematics, one of the possible outcomes of that program would be the adoption of a new axiom implying (1) that power of the continuum is ₂ and (2) that Martin's Axiom (MA) holds. (1) is something Gödel came, at least for a time, to believe (in connection with the unedifying square axioms incident alluded to earlier). (2) is shown by Martin and Solovay, in the paper in which MA was first introduced ("Internal Cohen extensions," Annals of Mathematical Logic, vol. 2 (1970), pp. 143-178; see especially §5.3 "Is A true?" pp. 176-177), to imply many of the same consequences as CH. In particular, MA implies several of the consequences about extreme rarity that Gödel judges implausible, plus a modified version of another that Gödel might well have judged nearly equally implausible.
³⁷ The implausibility judgments are at least indirectly classified as "intuitions" by commentators. Martin and Solovay contrast Gödel's opinion with their own "intuitions," thus:

If one agrees with Gödel that [the extreme rareness results] are implausible, then one must consider [MA] an unlikely proposition. The authors, however, have virtually no intuitions at all about [the extreme rareness results]… (p. 176)

Martin ("Hilbert's First Problem: The Continuum Hypothesis," in F. Browder, ed., Mathematical Developments Arising from Hilbert Problems, Proceedings of Symposia in Pure Mathematics, vol. 28 (Providence: American Mathematical Society), pp. 81-92) refers to Gödel's judgments as "intuitions" as he expresses dissent from them, thus:

While Gödel's intuitions should never be taken lightly, it is very hard to see that the situation is different from that of Peano curves, and it is even hard for some of us to see why the examples Gödel cites are implausible at all.

The usage of the commentators here is in conformity with the kind of usage of "intuition" in mathematics to be discussed in the next section; but it seems Gödel's usage is more restricted than that.
³⁸ Jacques Hadamard, The Psychology of Invention in the Mathematical Field (New York: Dover), 1945. George Polya, Mathematics and Plausible Reasoning (Princeton: Princeton University Press), 1954, vol. I Induction and Analogy in Mathematics, vol. II Patterns of Plausible Inference. The resemblance between mathematical and scientific methodology is most conspicuous in Polya's second volume, where the patterns of plausible inference Polya detects in mathematical thought closely resemble the rules of Bayesian probabilistic inference often cited in work on the epistemology of science. It is, however, difficult to view them as literal instances, since the Bayesians often require that all logicomathematical truths be assigned probability one.
³⁹ There are as well principles for which we do not even have a rigorous statement, let alone a rigorous proof. Such is the case with the Lefschetz principle, or Littlewood's three principles, for instance. Rigorous formulations of parts of such principles are possible, but always fall short of their full content. The "rules of thumb" in set theory identified by Maddy ("Believing the axioms," Journal of Symbolic Logic, vol. 53 (1988), part I pp. 481-511, part II pp. 736-764) may also be considered to be of this type.
⁴⁰ "Believing the Axioms, " §II.3 "Informed opinion," pp. 494-500. To give an example not in Maddy's collection, one might argue heuristically against the continuum hypothesis as follows. CH implies not only that all uncountable subsets of the line have the same number of elements, but also that all partitions of the line into uncountably many pieces have the same number of pieces. But even looking at very simple partitions (those for which the associated equivalence relation, considered as a subset of the plane, is analytic) with uncountably many pieces, we find what seem two quite different kinds. For it can be proved that the number of pieces is exactly ₁ and that there is no perfect set of pairwise inequivalent elements, while for others it can proved that there is such a perfect set and (hence) that the number of pieces is the power of the continuum. (Compare Sashi Mohan Srivastava, A Course on Borel Sets (Berlin: Springer), 1988, chapter 5.)
⁴¹ The suspicion was confirmed by Cohen just a little too late for any more discussion than a very short note at the end to be incorporated into the paper.
⁴² Especially the one Maddy calls "Maximize." This looks closely related to Gödel's thinking in footnote 23, p.266, already cited.
⁴³ This formulation may need a slight qualification. Suppose you are walking through a city you have never visited before, and are approaching a large public building, but are still a considerable distance away, and that the air is full of dust. Despite distance and dust, you are able to form some visual impression of the building. You are equally able to make conjectures about the appearance of the building by induction and analogy, taking into account the features of the lesser buildings you are passing, which you can see much better, and of large public buildings in other cities in the same country that you have recently visited under more favorable viewing conditions. Owing to the influence of expectation on perception, it is just barely possible, if the building is distant enough and the air dusty enough, to mistake such a conjecture for a visible impression, and think one is seeing what one is in fact only imagining must be there. But these are marginal cases.
⁴⁴ David Hume, Enquiry Concerning the Principles of Morals, §III, part II, ¶ 10. (In version edited by J. Schneewind (Indianapolis: Hackett), 1983, the passage appears on p. 29.) There is, of course, this difference from the situation described by Hume, that it isn't so clear that the interests of society or even of mathematics demand a ruling on the status of the continuum hypothesis.
⁴⁵ Parsons, in correspondence, suggests that Gödel might emphasize that potential new axioms force themselves upon us as flowing from the very concept of set, something that is rather obviously not the case with his implausibility judgments, though it is equally obviously not the case with the "square axioms" Gödel was later to propose. The danger I see with emphasizing this feature, in order to distinguish rational from heuristic intuition, is that it may make it more difficult to distinguish rational from linguistic intuition.