Popper: Philosophy of Probability

There are two central questions concerning probability. First, what are its formal features? That is a mathematical question, to which there is a standard, widely (though not universally) agreed upon answer. This answer is reviewed in the next section. Second, what sorts of things are probabilities---what, that is, is the subject matter of probability theory? This is a philosophical question, and while the mathematical theory of probability certainly bears on it, the answer must come from elsewhere. To see why, observe (...) that there are many things in the world that have the mathematical structure of probabilities---the set of measurable regions on the surface of a table, for example---but that one would never mistake for being probabilities. So probability is distinguished by more than just its formal characteristics. The bulk of this essay will be taken up with the central question of what this “more” might be. (shrink)

How are we to understand the use of probability in corroboration functions? Popper says logically, but does not show we could have access to, or even calculate, probability values in a logical sense. This makes the logical interpretation untenable, as Ramsey and van Fraassen have argued. -/- If corroboration functions only make sense when the probabilities employed therein are subjective, however, then what counts as impressive evidence for a theory might be a matter of convention, or even whim. So isn’t (...) so-called ‘corroboration’ just a matter of psychology? -/- In this paper, I argue that we can go some way towards addressing this objection by adopting an intersubjective interpretation, of the form advocated by Gillies, with respect to corroboration. I show why intersubjective probabilities are preferable to subjective ones when it comes to decision making in science: why group decisions are liable to be superior to individual ones, given a number of plausible conditions. I then argue that intersubjective corroboration is preferable to intersubjective confirmation of a Bayesian variety, because there is greater opportunity for principled agreement concerning the factors involved in the former. (shrink)

This paper is a re-examination of Popper’s propensity interpretation of probability in respect of its potential methodological value in social science. A long standing problem for the frequency interpretation of probability is that whilst it is able to treat both aggregate and individual phenomena as having measurable properties, it cannot explain the ontological relationship between such concrete individual cases and aggregates. Popper’s interpretation treats single cases as both real, but also as realisations of a propensity to occur. The frequency and (...) propensity interpretations are compared and whilst some common objections raised must be upheld, they do not devalue the importance of the propensity interpretation as at least a metaphysical basis for probabilistic claims in social science. However the value of the approach may also lie in its methodological potential. Here I sugest that single case probabilities must be analysed in terms of the anterior probabilities of prior constituent events. In this I move beyond Popper’s own programme, but suggest that such a move is theoretically compatible with recent complexity approaches in social science and goes some way toward meeting anti-naturalist concerns about intentionality. (shrink)

I am indebted to Zwirn and Zwirn [1989] for their extended and careful comments on the arguments of Popper & Miller [1983], [1987], and also for friendly and illuminating conversations. Their judgement seems to be that although Popper and I fail to make a satisfactory case for our conclusion that inductive probability is impossible, that conclusion is nonetheless defensible on quite other grounds. I don’t really agree with this, as I shall explain.

One of the many intriguing features of the axiomatic systems of probability investigated in Popper (1959), appendices _iv, _v, is the different status of the two arguments of the probability functor with regard to the laws of replacement and commutation. The laws for the first argument, (rep1) and (comm1), follow from much simpler axioms, whilst (rep2) and (comm2) are independent of them, and have to be incorporated only when most of the important deductions have been accomplished. It is plain that, (...) in the presence of (comm1), the principle (sub), which says that terms that are intersubstitutable in the first argument are intersubstitutable also in the second argument, implies (comm2), and in Popper’s systems the converse implication obtains. It is naturally asked what is needed in an axiomatic theory of probability in order to enforce this equivalence. Leblanc (1981) offered a rather weak set of axioms, containing (comm1) and (comm2), that suffice for the derivation of (sub). In this paper Leblanc’s result is improved in a number of different ways. Three weaker systems, one of which is incomparable with the other two, are shown to admit the same implication. DOI:10.5007/1808-1711.2011v15n2p271. (shrink)

Sir Karl Popper is well known for explicating science in falsificationist terms, for which his degree of corroboration formalism, C(h,e,b), has become little more than a symbol. For example, de Queiroz and Poe in this issue argue that C(h,e,b) reduces to a single relative (conditional) probability, p(e,hb), the likelihood of evidence e, given both hypothesis h and background knowledge b, and in reaching that conclusion, without stating or expressing it, they render Popper a verificationist. The contradiction they impose is easily (...) explained--de Queiroz and Poe fail to take account of the fact that Popper derived C(h,e,b) from absolute (logical) probability and severity of test, S(e,h,b), where critical evidence, p(e,b), is fundamental. Thus, de Queiroz and Poe's conjecture that p(e,hb) = C(h,e,b) is refuted. Falsificationism, not verificationism, remains a fair description of the parsimony method of inference used in phylogenetic systematics, not withstanding de Queiroz and Poe's mistaken understanding that "statistical" probability justifies that method. Although de Queiroz and Poe assert that maximum likelihood has the power "to explain data", they do not successfully demonstrate how causal explanation is achieved or what it is that is being explained. This is not surprising, bearing in mind that what is assumed about character evolution in the accompanying likelihood model M cannot then be explained by the results of a maximum likelihood analysis. (shrink)

Popper's paradox of ideal evidence has long been viewed as a telling criticism of Keynes's logical theory of probability and its associated concept of the weight of argument. This paper shows that a simple addition to Keynes's definitions of irrelevance enables his theory to elude the paradox with ease. The modified definition draws on ideas already present in Keynes's Treatise on Probability (1973). As a consequence, relevant evidence and the weight of argument may increase, even when new evidence leaves the (...) probability unaltered. (shrink)

The propensity interpretation of probability, bred by Popper in 1957(K. R. Popper, in Observation and Interpretation in the Philosophy of Physics,S. Körner, ed. (Butterworth, London, 1957, and Dover, New York, 1962), p. 65; reprinted in Popper Selections,D. W. Miller, ed. (Princeton University Press, Princeton, 1985), p. 199) from pure frequency stock, is the only extant objectivist account that provides any proper understanding of single-case probabilities as well as of probabilities in ensembles and in the long run. In Sec. 1 of (...) this paper I recall salient points of the frequency interpretations of von Mises and of Popper himself, and in Sec. 2 I filter out from Popper's numerous expositions of the propensity interpretation its most interesting and fertile strain. I then go on to assess it. First I defend it, in Sec. 3, against recent criticisms(P. Humphreys, Philos. Rev.94,557 (1985); P. Milne, Erkenntnis25,129 (1986)) to the effect that conditional [or relative] probabilities, unlike absolute probabilities, can only rarely be made sense of as propensities. I then challenge its predominance, in Sec. 4, by outlining a rival theory: an irreproachably objectivist theory of probability, fully applicable to the single case, that interprets physical probabilities asinstantaneous frequencies. (shrink)

I survey the difficulties of several probabilistic views of non-deductive argument and of inductive probability and propose to explicate non-deductive reasoning in terms of rational preference. Following a critical examination of Popper’s allegedly deductive theory of rational preference, I draw upon the work of Popper and Rescher to present my view which includes: (i) the conjecturing of a set of alternative answers to or theories or hypotheses about the questions prompting the inquiry and (ii) the “reduction” of this set via (...) plausibilistic principles of rational preference. (shrink)

This paper uses Popper's treatment of probability and an epistemic constraint on probability assignments to conditionals to extend the Bayesian representation of rational belief so that revision of previously accepted evidence is allowed for. Results of this extension include an epistemic semantics for Lewis' theory of counterfactual conditionals and a representation for one kind of conceptual change.

The question of truth is a broadly broached subject in Philosophy as it features along the entire historical and polemical growth of the discipline right from the time of the Ancients down to our Post-Modern era. Yet, the delimiting realization of being unable to register general success in our dogged attempts at truth and knowledge, mostly stares us blankly in the face, for matters on which philosophy endeavours to speculate on, are beyond the reach of definite knowledge.1 Our theories of (...) the universe open up to modifications, refutations, and further propositions, evidencing a historical development in philosophical inquiry. This generally is the growth of our science, of our knowledge. This paper critically seeks to examine Popper’s notion of verisimilitude. It takes us through the scientist’s journey from ignorance to truth, and the difference between probability and verisimilitude. It addresses the relevance of the theory of content in understanding verisimilitude, under its distinctions as quantitative and qualitative. Finally, it discusses corroboration and the criteria for theory-choice. (shrink)

There are (at least) three approaches to quantifying information. The first, algorithmic information or Kolmogorov complexity, takes events as strings and, given a universal Turing machine, quantifies the information content of a string as the length of the shortest program producing it [1]. The second, Shannon information, takes events as belonging to ensembles and quantifies the information resulting from observing the given event in terms of the number of alternate events that have been ruled out [2]. The third, statistical learning (...) theory, has introduced measures of capacity that control (in part) the expected risk of classifiers [3]. These capacities quantify the expectations regarding future data that learning algorithms embed into classifiers. Solomonoff and Hutter have applied algorithmic information to prove remarkable results on universal induction. Shannon information provides the mathematical foundation for communication and coding theory. However, both approaches have shortcomings. Algorithmic information is not computable, severely limiting its practical usefulness. Shannon information refers to ensembles rather than actual events: it makes no sense to compute the Shannon information of a single string – or rather, there are many answers to this question depending on how a related ensemble is constructed. Although there are asymptotic results linking algorithmic and Shannon information, it is unsatisfying that there is such a large gap – a difference in kind – between the two measures. This note describes a new method of quantifying information, effective information, that links algorithmic information to Shannon information, and also links both to capacities arising in statistical learning theory [4, 5]. After introducing the measure, we show that it provides a non-universal analog of Kolmogorov complexity. We then apply it to derive basic capacities in statistical learning theory: empirical VC-entropy and empirical Rademacher complexity. A nice byproduct of our approach is an interpretation of the explanatory power of a learning algorithm in terms of the number of hypotheses it falsifies [6], counted in two different ways for the two capacities. We also discuss how effective information relates to information gain, Shannon and mutual information. (shrink)

Following the pioneer work of Bruno De Finetti [12], conditional probability spaces (allowing for conditioning with events of measure zero) have been studied since (at least) the 1950's. Perhaps the most salient axiomatizations are Karl Popper's in [31], and Alfred Renyi's in [33]. Nonstandard probability spaces [34] are a well know alternative to this approach. Vann McGee proposed in [30] a result relating both approaches by showing that the standard values of infinitesimal probability functions are representable as Popper functions, and (...) that every Popper function is representable in terms of the standard real values of some infinitesimal measure. Our main goal in this article is to study the constraints on (qualitative and probabilistic) change imposed by an extended version of McGee's result. We focus on an extension capable of allowing for iterated changes of view. Such extension, we argue, seems to be needed in almost all considered applications. Since most of the available axiomatizations stipulate (definitionally) important constraints on iterated change, we propose a non-questionbegging framework, Iterative Probability Systems (IPS) and we show that every Popper function can be regarded as a Bayesian IPS. A generalized version of McGee's result is then proved and several of its consequences considered. In particular we note that our proof requires the imposition of Cumulativity, i.e. the principle that a proposition that is accepted at any stage of an iterative process of acceptance will continue to be accepted at any later stage. The plausibility and range of applicability of Cumulativity is then studied. In particular we appeal to a method for defining belief from conditional probability (first proposed in [42] and then slightly modified in [6] and [3]) in order to characterize the notion of qualitative change induced by Cumulative models of probability kinematics. The resulting cumulative notion is then compared with existing axiomatizations of belief change and probabilistic supposition. We also consider applications in the probabilistic accounts of conditionals [1] and [30]. (shrink)

Karl Popper (1902-1994) was one of the most influential philosophers of science of the 20th century. He made significant contributions to debates concerning general scientific methodology and theory choice, the demarcation of science from non-science, the nature of probability and quantum mechanics, and the methodology of the social sciences. His work is notable for its wide influence both within the philosophy of science, within science itself, and within a broader social context. Popper’s early work attempts to solve the problem of (...) demarcation and offer a clear criterion that distinguishes scientific theories from metaphysical or mythological claims. Popper’s falsificationist methodology holds that scientific theories are characterized by entailing predictions that future observations might reveal to be false. When theories are falsified by such observations, scientists can respond by revising the theory, or by rejecting the theory in favor of a rival or by maintaining the theory as is and changing an auxiliary hypothesis. In either case, however, this process must aim at the production of new, falsifiable predictions. While Popper recognizes that scientists can and do hold onto theories in the face of failed predictions when there are no predictively superior rivals to turn to. He holds that scientific practice is characterized by its continual effort to test theories against experience and make revisions based on the outcomes of these tests. By contrast, theories that are permanently immunized from falsification by the introduction of untestable ad hoc hypotheses can no longer be classified as scientific. Among other things, Popper argues that his falsificationist proposal allows for a solution of the problem of induction, since inductive reasoning plays no role in his account of theory choice. Along with his general proposals regarding falsification and scientific methodology, Popper is notable for his work on probability and quantum mechanics and on the methodology of the social sciences. Popper defends a propensity theory of probability, according to which probabilities are interpreted as objective, mind-independent properties of experimental setups. Popper then uses this theory to provide a realist interpretation of quantum mechanics, though its applicability goes beyond this specific case. With respect to the social sciences, Popper argued against the historicist attempt to formulate universal laws covering the whole of human history and instead argued in favor of methodological individualism and situational logic. Table of Contents 1. Background 2. Falsification and the Criterion of Demarcation a. Popper on Physics and Psychoanalysis b. Auxiliary and Ad Hoc Hypotheses c. Basic Sentences and the Role of Convention d. Induction, Corroboration, and Verisimilitude 3. Criticisms of Falsificationism 4. Realism, Quantum Mechanics, and Probability 5. Methodology in the Social Sciences 6. Popper’s Legacy 7. References and Further Reading a. Primary Sources b. Secondary Sources -/- . (shrink)

In 1983, in an open letter to the journal Nature, Karl Popper and David Miller set forth a particularly strong critical argument which sought to demonstrate the impossibility of inductive probability. Since its publication the argument has faced many criticisms and we argue in this article that they do not reach their objectives. We will first reconstruct the demonstration made by Popper and Miller in their initial article and then try to evaluate the main arguments against it. Although it is (...) possible to conceptualize logically the idea of induction, it is shown that it is not possible on traditional Bayesian grounds. (shrink)

The form of argument used by Popper and Miller to attack the concept of probabilistic induction is applied to the slightly different situation in which some evidence undermines a hypothesis. The result is seemingly absurd, thus bringing the form of argument under suspicion.

Popper and Miller (1983) have presented an argument purporting to establish the impossibility of inductive probability. Here I discuss critically their characterization of a deductive part of nondeductive support, a point that has not figured centrally in previous responses.

Kolmogorov's account in his [1933] of an absolute probability space presupposes given a Boolean algebra, and so does Rényi's account in his [1955] and [1964] of a relative probability space. Anxious to prove probability theory ‘autonomous’. Popper supplied in his [1955] and [1957] accounts of probability spaces of which Boolean algebras are not and [1957] accounts of probability spaces of which fields are not prerequisites but byproducts instead.1 I review the accounts in question, showing how Popper's issue from and how (...) they differ from Kolmogorov's and Rényi's, and I examine on closing Popper's notion of ‘autonomous independence’. So as not to interrupt the exposition, I allow myself in the main text but a few proofs, relegating others to the Appendix and indicating as I go along where in the literature the rest can be found. (shrink)

In other words, classically, probabilities add; quantum mechanically, the probability amplitudes add, leading to the presence of the extra product terms in the quantum case. What this means is that in quantum theory, even though always only one of the various outcomes is obtained in any given observation, some aspect of the non -occurring events, represented by the corresponding complex-valued quantum amplitudes, plays a role in determining the overall probabilities. Indeed, the observed quantum interference effects are correctly captured by the (...) quantum statistical description only because of the presence of these product terms. Therefore, in a realistic construal of quantum theory, if we treat the superposed Ψ function as representing the real state of an individual quantum system, these quantum amplitudes need to be given an ontological status. (shrink)

The propensity interpretation of probability was introduced by Popper ([1957]), but has subsequently been developed in different ways by quite a number of philosophers of science. This paper does not attempt a complete survey, but discusses a number of different versions of the theory, thereby giving some idea of the varieties of propensity. Propensity theories are classified into (i) long-run and (ii) single-case. The paper argues for a long-run version of the propensity theory, but this is contrasted with two single-case (...) propensity theories, one due to Miller and the later Popper, and the other to Fetzer. The three approaches are compared by examining how they deal with a key problem for the propensity approach, namely the relationship between propensity and causality and Humphreys' paradox. (shrink)

In these prefatory remarks, which are designed to locate my topic within the complex and wide-stretching field of Popper's thought and writings, I shall not say anything that those familiar with his work will not already know. Moreover, what I do say will take as understood many of the problems and theories, not to mention the terminology, that I shall later be doing my best to make understandable. My apologies are therefore due equally to those who know something about Popper's (...) discussions of indeterminism and of the propensity interpretation of probability, and to those who know nothing. (shrink)

Shown here is that a constraint used by Popper in The Logic of Scientific Discovery (1959) for calculating the absolute probability of a universal quantification, and one introduced by Stalnaker in "Probability and Conditionals" (1970, 70) for calculating the relative probability of a negation, are too weak for the job. The constraint wanted in the first case is in Bendall (1979) and that wanted in the second case is in Popper (1959).

In the first part of this work(1) we have explicated the spacetime structure of the probabilistic organization of quantum mechanics. We have shown that each quantum mechanical state, in consequence of the spacetime characteristics of the epistemic operations by which the observer produces the state to be studied and the processes of qualification of these, brings in a tree-like spacetime structure, a “quantum mechanical probability tree,” thattransgresses the theory of probabilities as it now stands. In this second part we develop (...) the general implications of these results.Starting from the lowest level of cognitive action and creating an appropriate symbolism, we construct a “relativizing epistemic syntax,” a “general method of relativized conceptualization” where—systematically—each description is explicitly referred to the epistemic operations by which the observer produces the entity to be described and obtains qualifications of it. The method generates a typology of increasingly complex relativized descriptions where the question of realism admits of a particularly clear pronouncement. Inside this typology the epistemic processes that lie—UNIVERSALLY—at thebasis ofany conceptualization, reveal a tree-like spacetime structure. It appears in particular that the spacetime structure of the relativized representation of aprobabilistic description, which transgresses the nowadays theory of probabilities,is the general mould of which the quantum mechanical probability trees are only particular realizations. This entails a clear definition of the descriptional status of quantum mechanics. While the recognition of theuniversal cognitive content of the quantum mechanical formalism opens up vistas towardmathematical developments of the relativizing epistemic syntax.The relativized representation of a probabilistic description leads with inner necessity to a “morphic” interpretation of probabilities thatcan be regarded as a formalized and deepening elaboration of Sir Karl Popper's “propensity” interpretation. A functional is then constructed, the “opacity functional,” that associatesa mathematical expression to the Popperian “propensities”. Furthermore the opacity functional produces adeductive definition of Shannon's “informational entropy.” Thereby there appears an explicitly unified relativized probabilistic-informational approach. This sketches out a second branch of a future mathematical epistemic syntax, to be connected with the branch stemming from quantum mechanics.The problem of the objectivity of probabilistic descriptions acquires certain precise rephrasings and—in a sense—solution. (shrink)