Remarks on purity of methods (Q1982004)

Starting from the work of the philosophers Michael Detlefsen and Andrew Arana, Pillay reflects on the notion of purity of proofs. He also discusses the notions of mathematical progress and what, following Michael Atiyah, he calls the unity of mathematics. To understand Pillay's paper, a few words on the work of Michael Detlefsen are in order. Detlefsen worked to clarify epistemic ideals that shape the practice of mathematics. He devoted several articles to distinguishing between different notions of rigor, intuition, construction, purity of proofs, and efficiency in mathematical reasoning. Concerning purity, Detlefsen wrote two articles: [\textit{M. Detlefsen}, ``Purity as an ideal of proof'', in: The philosophy of mathematical practice. Oxford: Oxford University Press. 179--197 (2008)] and [\textit{M. Detlefsen} and \textit{A. Arana}, ``Purity of methods'', Philosopher's imprint 11, No. 2, 1--22 (2011)] (this last one co-authored with Andrew Arana). In the sense that Detlefsen studied the most, \textit{topical purity}, purity indicates a relationship between the conceptual resources that one needs to understand a theorem and the conceptual resources needed to prove that theorem. A pure proof uses only resources that, in a sense at least partially clarified in Detlefsen's articles, are ``intrinsic'' to the theorem that it establishes or are ``necessary'' to understand that theorem. Detlefsen shows that concern with purity of this type is pervasive (it influences the practice of many different mathematical fields) and persistent (it appears in many different periods of the history of mathematics). To give an example, according to Detlefsen and Arana, Furstenberg's proof that there are infinitely many primes is impure because an agent could understand the statement of that theorem without being rationally committed to accepting the notion of topological space (which they take Furstenberg's proof to use). Pillay finds the concern with purity stimulating. However, \begin{itemize} \item[(i)] he wants to partially redefine the notion of purity to make it more relevant to mathematicians, and \item[(ii)] he wants to connect the notion of purity with what he calls the ``unity of mathematics''. \end{itemize} Concerning (i), Pillay finds purity important as a stimulus to determine which limited resources (collection of assumptions or techniques) are sufficient (or, given a context, needed) to prove a specific statement. However, \begin{itemize} \item[(i.a)] Pillay finds the specific conceptual apparatus that Detlefsen develops either misapplied or non-explanatory and \item[(i.b)] he finds Detlefsen's focus on specific theorems and proofs and the apparent idea that such items have an immutable meaning misleading. \end{itemize} On pp. 196--197, addressing the first part of (i.a), Pillay says convincingly that Furstenberg's proof only \textit{appears} to refer to topological notions extraneous to the statement that there are infinitely many primes. However, Pillay explains that, looking more closely, according to Detlefsen's and Arana's notion of topical purity, Furstenberg's proof is pure because it refers only to sets definable in the first-order structure $(\mathbb{Z},\ +,\ \times)$. Concerning the second part of (i.a), i.e. the non-explanatoriness of Detlefsen's and Arana's notions, Pillay's argument is less clear (p. 196). He considers the statement $S$ that a finite number of points in the real plane lie on a straight line if each line that joins two of them intersects a third line. Kelly's proof uses the notion of shortest Euclidean distance. Detlefsen and Arana argue that the notion of distance is foreign to the initial statement because one could understand or determine $S$ without referring or being committed to the notion of shortest distance. Pillay considers another proof of $S$: the proof provided by Steinberg. Pillay seems to think that it would be implausible to take Steinberg's proof to be impure although this proof uses the notion of betweenness of points and one does not use this notion to determine $S$ either. Pillay seems to reason as follows: Detlefsen and Arana are committed to saying that Steinberg's proof is impure; however, since the proof is pure, their conceptual analysis does not adequately describe (and therefore does not explain) the distinction between pure and impure proofs. Pillay does not argue for the critical step that one ought to consider Steinberg's proof to be pure (or, more precisely, topically pure). The tone of his discussion gives (at least to me) the impression he takes betweenness to be a basic notion concerning points on a line and, therefore always available to provide a pure proof of a statement about points and line (the general, albeit unexpressed, principle seems to be that the use of a basic notion of a domain does not alter the topic of a statement concerning entities of that domain). In sum, Pillay's observations concerning Steinberg's proof deserve a fuller consideration than the one that the author offers in the paper. Concerning (i.b), Pillay writes: ``To understand a definition or object or statement $\dots$ requires work, and such work may itself change the commitments involved $\dots$ The process of doing mathematics, [sic] changes the subject'' (p. 198). Though Pillay does not provide a specific example, the case of Desargues' theorem may be appropriate. Its statement refers explicitly only to the intersection of lines in a plane. Therefore, it seems that a topically pure proof of Desargues' theorem ought to refer only to notions belonging to plane geometry. However, Hilbert showed that the statement is sufficient and necessary to develop solid geometry. So, the ``intuitive'' or informal assessment of the complexity of concepts involved in that statement is not adequate and a sophisticated mathematical analysis shows the real conceptual complexity. It follows that a proof of Desargues' theorem that uses notions belonging to solid geometry may be pure (for a detailed discussion, see [\textit{M. Hallett}, ``Reflections on the purity of methods in Hilbert's \textit{Grundlagen der Geometrie}'', in: Philosophy of mathematical practice. Oxford University Press. 198--255 (2008)]. Pillay's remark seems to be a criticism of the claim that the determination of what a statement is about (or what concepts an agent ought to employ or accept to understand that statement) is possible without considering new information about specific concepts and topics that mathematical progress makes available. It is true that Detlefsen and Arana do not consider the possibility that what appears to be an impure proof today may be revealed to be pure when we discover new mathematical results or vice versa. However, it is not clear that Detlefsen and Arana ought to accept the static view of meaning that Pillay criticizes (for a discussion of different notions of content of a statement and their connection to the study of purity see [\textit{A. Arana} and \textit{P. Mancosu}, ``On the relationship between plane and solid geometry'', The Review of Symbolic Logic 5, No. 2, 294--353 (2012)]; such a discussion provides some elements that one may use to defend Detlefsen's and Arana's account of purity against Pillay's charge). In any case, Pillay's remark bring to the fore a delicate issue that deserves further reflections. First, new results may modify our understanding of some concepts. However, it is not clear that these results change the meaning of statements that use these concepts. Second, it may also happen that new results lead (perhaps slowly) to use the same name to refer to what are different concepts (like shifts in the meaning of ``space'' from, let's say, Euclid to our days) or to aggregate concepts that one may plausibly take to be distinct like using ``numbers'' to refer to objects used for things as different as counting (where one may use the natural numbers), fixing a position in a sequence (where the integers, for example, may be appropriate), and measuring (where one may use, for example, the reals). Concerning (ii), Pillay says that the search for pure proofs is stimulating but should not be an exclusive ideal. A complementary expansion of resources and the resulting impure proofs that allow one to ``look'' at the same theorem from different points of view and that uncover connections or analogies between different parts of mathematics are also important contributions to mathematical progress (p. 195). On pp. 197--200, Pillay urges philosophers to pay more attention to what, following, Michael Atiyah, he calls the ``unity of mathematics'': \begin{itemize} \item[(U1)] the fact that the ``same'' definitions, concepts, or theorems appear, often independently, in different areas of mathematics and \item[(U2)] the fact that techniques and results from one area of mathematics are often useful or, at least practically speaking, indispensable to prove non-trivial results in other areas of mathematics. \end{itemize} In conclusion, although sometimes Pillay does not develop his remarks in detail, his article is a careful engagement with Detlefsen's thought and full of stimulating material for further philosophical reflections.