Countable Additivity and the Continuity of Integration

Abstract

Philosophers cannot agree on whether the rule of Countable Additivity should be an axiom of probability. Edwin T. Jaynes attacks the problem in a way which is original to him and passed over in the current debate about the principle: he says the debate only arises because of an erroneous use of mathematical infinity. I argue that this solution fails, but I construct a different argument which, I argue, salvages the spirit of the more general point Jaynes makes. I argue that in Jaynes's objective Bayesianism we might have good reasons to adopt Countable Additivity, and some of the major problems this adoption is known to entail need not worry us. In particular, I propose to adopt this new angle on Countable Additivity in Jon Williamson's version of objective Bayesianism.

Access options

Buy single article

Instant access to the full article PDF.

39,95 €

Price includes VAT (Indonesia)

Notes

The closest mentions are the following: Hájek (2011) refers the reader to Jaynes's sections 15.3-5 for a discussion of conglomerability; and in a recent paper Myrvold (2015) cites section 15.7 of Jaynes's book on the Borel–Kolmogorov paradox. The discussion on CA is in section 15.6.
If a measure is countably additive, then it is also finitely additive. Let P be a countably additive measure. We want to show that \( P\left( \bigcup \nolimits _{n=1}^{N} A_n \right) = \sum \nolimits _{n=1}^{N} P(A_n) \), for \(A_1, \dots , A_n\) pairwise disjoint. We can extend this sequence by a countably infinite sequence of empty sets: \(A_{n+1}, A_{n+2}, \dots = \emptyset \). We now can write \( P\left( \bigcup \nolimits _{n=1}^{N} A_n \right) = P\left( \bigcup \nolimits _{n=1}^{\infty } A_n \right) = \sum \nolimits _{n=1}^{\infty } P(A_n) = P(A_1) + \cdots + P(A_N) + P(\emptyset ) + P(\emptyset )+ \cdots = \sum \nolimits _{n=1}^{N} P(A_n)\). For an example of a probability function which is FA but not CA, see function Q below.
see, for example Dunford and Schwartz (1958, chapter III), or de Finetti (1972, Chapter 6)
For classic expositions that adopt the principle of CA see, for example, the following historical progression: Kolmogorov (1933/1956, 15), Halmos (1950/1974, 187) and Cohn (2013, 7).
For other important contributions to this lively philosophical debate, see, amongst others, the following: Levi (1980), Seidenfeld and Schervish (1983), Kadane et al. (1986) Kelly (1996, Chapter 13), Howson (2009).
n here is the number of propositions treated; while Jaynes speaks of propositions, I will use the terms proposition and event interchangeably to mean elements of the domain of the probability functions treated.
I use this terminology to reflect Jaynes's, although 'unbounded' might be more natural.
Note, however, that while this function exists, it is not trivial to define, if we require it to be defined for all subsets of \(\mathbb {N}\): see Kadane and O'Hagan (1995) and my discussion below.
For example, the Vitali sets are members of the power set of \([0,1]\subset \mathbb {R} \) for which the Lebesgue measure is not defined.
A real interval can be partitioned into at most countably many non-empty intervals, as each non-empty sub-interval must contain a rational number because rational numbers are dense in \(\mathbb {R}\), and there are countably many rational numbers.
The property of operators in deductive logic Jaynes refers to is usually called functional completeness: through conjunction and negation all other logical operators can expressed; it is not obvious that Jaynes's "adequacy" property in probability is a natural parallel, but this is unimportant for the present argument.
Kadane and O'Hagan treat probability distributions over the natural numbers: events are represented by the sets \(\{1\},\{2\}, \dots \), instead of propositions \(\{A_1, A_2,\dots \}\), but the argument is identical
We can see this easily as follows: if \( P(1)=P(2)=\cdots =P(n)=0 \), \( P\left( \bigcup \nolimits _{i=1}^n i \right) = P(1) + \cdots + P(n) = 0 \). For the second remark, see the following: \(P\left( \bigcup \nolimits _{i=1}^n i \right) = 0 \), but \( P\left( \bigcup \nolimits _{i=1}^n i \right) + P\left( \bigcup \nolimits _{i=n+1}^{\infty } i \right) = 1 \) by FA, so \( P\left( \bigcup \nolimits _{i=n+1}^{\infty } i \right) = 1 \).
Some paraphrase is needed to give context present in their paper.
I am grateful to an anonymous referee for highlighting this point.
De Finetti's argument, re-translated from the Italian, is as follows (this is a comment on a proof such as the above, where it is claimed that if our betting quotients do not abide by CA, then we are open to a certain loss):

But this is a bit of a vicious circle, because only if I knew complete additivity to be valid could I think of extending the notion of 'fair combination of bets' to combinations of infinite bets, and of basing them on the series of the betting odds (de Finetti 1949, 12) [emphasis as in the original, which follows: Un motivo che tenderebbe ad avvalorare l'additività completa: se le probabilità \(p_n\) hanno somma \(p<1\), stipulando tutte le infinite scommesse posso ricevere in ogni caso 1 pagando p, e quindi avrei un'incongruenza. Ma è un po' un circolo vizioso, perchè solo se sapessi valida l'additività completa potrei pensare di estendere la nozione di 'combinazione di scommesse equa' a combinazioni di infinite scommesse, e di basarle sulla serie delle quote di scommessa.]

Howson (2008) studies this argument but is puzzled by it, as, he writes, are many authors before him. In the English version quoted by Howson and others, the word serie (series) is wrongly translated as 'sequence'
Note also that \(I_{CAJ}\) does not imply \(I_{CAW}\), as the the latter principle is about bets, on which \(I_{CAJ}\) is completely silent. That is, supposing it true that the probabilities of elementary events should determine the probabilities of the compound events they form, we still have no indication on whether bets should be countably additive or not. It is a tenable position to uphold \(I_{CAJ}\) but deny the interpretation of degrees of belief as bets, thus, for example, believing \(I_{CAJ}\), the countable additivity of degrees of belief, but not the countable additivity of bets.

References

Bartha, P. (2004). Countable Additivity and the de Finetti Lottery. The British Journal for the Philosophy of Science, 55(2), 301–321. https://doi.org/10.1093/bjps/55.2.301.

Article Google Scholar
Bingham, N. H. (2010). Finite additivity versus countable additivity. Electronic Journal for History of Probability and Statistics, 6(1), 1–35.

Google Scholar
Boole, G. (1854/1958). An investigation of the laws of thought. New York: Dover Publications.
Cohn, D. L. (2013). Measure theory (2nd ed.). Birkhäuser advanced texts Berlin: Springer.
Cox, R. T. (1961). The algebra of probable inference. Baltimore: The John Hopkins Press.

Google Scholar
de Finetti, B. (1949). Sull'impostazione assiomatica del calcolo delle probabilità, 1942, 1947 Box 5, Folder 26, Bruno de Finetti Papers, 1924–2000, ASP.1992.01, Archives of Scientific Philosophy, Special Collections Department, University of Pittsburgh.
de Finetti, B. (1972). Probability, induction and statistics: The art of guessing. New York: Wiley.

Google Scholar
de Finetti, B. (1974). Theory of probability: A critical introductory treatment (Vol. 1). New York: Wiley.

Google Scholar
Diaconis, P. (2004). Review of Probability theory: The logic of science by E.T. Jaynes. SIAM News: Volume 37, Number 2. https://archive.siam.org/news/news.php?id=81.
Dunford, N., & Schwartz, J. T. (1958). Linear operators, part I. New York: Interscience Publishers.

Google Scholar
Easwaran, K. (2013). Why countable additivity? Thought, 2, 53–61.

Article Google Scholar
Faris, W. G. (2006). Review of Probability theory: The logic of science by E.T. Jaynes. Notices of the AMS, 53(1), 33–42.

Google Scholar
Fraenkel, A. A., Bar-Hillel, Y., & Levy, A. (1973). Foundations of set theory. Amsterdam: Elsevier.

Google Scholar
Hájek, A. (2011). Conditional probability. In P. S. Bandyopadhyay & M. R. Forster (Eds.), Handbook of philosophy of science (Vol. 7, pp. 99–135). Philosophy of statistics Amsterdam: Elsevier.
Halmos, P. R. (1974). Measure theory. Berlin: Springer.

Google Scholar
Howson, C. (2008). De Finetti, countable additivity, consistency and coherence. British Journal for the Philosophy of Science, 59(2008), 1–23.

Article Google Scholar
Howson, C. (2009). Can logic be combined with probability? Probably. Journal of Applied Logic, 7, 177–187.

Article Google Scholar
Jaynes, E. T. (2003). Probability theory: The logic of science. Cambridge: Cambridge University Press.

Book Google Scholar
Kadane, J. B., & O'Hagan, A. (1995). Using finitely additive probability: Uniform distributions on the natural numbers. Journal of the American Statistical Association, 90(430), 626–631.

Article Google Scholar
Kadane, J. B., Schervish, M. J., & Seidenfeld, T. (1986). Statistical implications of finitely additive probability. In J. B. Kadane, M. J. Schervish, & T. Seidenfeld (Eds.), Rethinking the foundations of statistics (pp. 211–232). Cambridge: Cambridge University Press.

Google Scholar
Kelly, K. T. (1996). The logic of reliable inquiry. Oxford: Oxford University Press.

Google Scholar
Kolmogorov, A. N. (1956). Foundations of the theory of probability. New York: Chelsea Publishing Company.

Google Scholar
Lauwers, L. (2007). The uniform distributions puzzle. Working paper (unpublished).
Lauwers, L. (2010). Purely finitely additive measures are non-constructible objects. Working paper (unpublished). Retrieved December 16, 2016 from https://lirias.kuleuven.be/bitstream/123456789/267264/1/DPS1010.pdf.
Levi, I. (1980). The enterprise of knowledge: An essay on knowledge, credal probability and chance. Cambridge: The MIT Press.

Google Scholar
Myrvold, W. C. (2015). You can't always get what you want some considerations regarding conditional probabilities. Erkenntnis, 80(3), 573–603.

Article Google Scholar
Paris, J. B. (1994). The uncertain reasoner's companion: A mathematical perspective. Cambridge tracts in theoretical computer science 39 Cambridge: Cambridge University Press.
Schurz, G., & Leitgeb, H. (2008). Finitistic and frequentistic approximation of probability measures with or without \(\sigma \)-additivity. Studia Logica, 89(2), 257–283.

Article Google Scholar
Seidenfeld, T., & Schervish, M. J. (1983). A conflict between finite additivity and avoiding Dutch book. Philosophy of Science, 50(3), 398–412.

Article Google Scholar
Szabó, Z. G. (2017). Compositionality. In E. N. Zalta (ed.), The stanford encyclopedia of philosophy (Summer 2017 Edition). https://plato.stanford.edu/archives/sum2017/entries/compositionality/.
Wenmackers, S., & Horsten, L. (2013). Fair infinite lottery. Synthese, 190, 37–61.

Article Google Scholar
Williamson, J. (1999). Countable additivity and subjective probability. British Journal for the Philosophy of Science, 50, 410–416.

Article Google Scholar
Williamson, J. (2010). In defence of objective Bayesianism. Oxford: Oxford University Press.

Book Google Scholar
Yosida, K., & Hewitt, E. (1952). Finitely additive measures. Transactions of the American Mathematical Society, 72(1), 46–66.

Article Google Scholar

Download references

Author information

Authors and Affiliations

University of Tilburg, Warandelaan 2, 5037 AB, Tilburg, The Netherlands

Colin Elliot

Corresponding author

Correspondence to Colin Elliot.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Elliot, C. E.T. Jaynes's Solution to the Problem of Countable Additivity. Erkenn 87, 287–308 (2022). https://doi.org/10.1007/s10670-019-00195-2

Download citation

Received: 19 December 2017
Accepted: 25 November 2019
Published: 03 January 2020
Issue Date: February 2022
DOI : https://doi.org/10.1007/s10670-019-00195-2

smithhicip1967.blogspot.com

Source: https://link.springer.com/article/10.1007/s10670-019-00195-2