The Minimum Description Length principle is the modernformalisation of Occam's razor. It has been extensively and successfullyused in machine learning, especially for noisy and long sources ofdata. However, the MDL principle presents some paradoxes andinconveniences. After discussing all these, we address two of the mostrelevant: lack of explanation and lack of creativity. We present newalternatives to address these problems. The first one, intensionalcomplexity, avoids extensional parts in a description, so distributingcompression ratio in a more even way than the MDL principle. (...) The secondone, information gain, forces that the hypothesis is informative wrt. the evidence, so giving a formaldefinition of what is to discover. (shrink)

En este trabajo hago una comparación sistemática entre las dos teorías de la confirmación más populares en la actualidad: el método hipotético-deductivo y el bayesianismo. En primer lugar, enumero los cinco problemas fundamentales de la teoría hipotético-deductivista. Estos son el problema de las hipótesis estadísticas, el del grado de confirmación, el de la conjunción irrelevante, el del holismo epistemológico y el de las hipótesis alternativas. Luego, hago una presentación general de la epistemología bayesiana y muestro de qué manera estos problemas (...) resultan tratables en el marco bayesiano. Finalmente, concluyo que los tres primeros admiten una solución satisfactoria, sobre la base de algunos supuestos razonables. Los dos últimos, en cambio, todavía son problemas abiertos para la teoría bayesiana de la confirmación. Bayesians often affirm they have produced the best available theory of confirmation. In particular, they claim to have superseded the venerable hypothetico-deductive method. In this paper I intend to assess that claim by making a systematic comparison between both theories of confirmation. I begin by listing the main problems of hypothetico-deductive confirmation. I then show that this conception is incompatible with Hempel's conditions of adequacy for qualitative confirmation. I expound the Bayesian theory of confirmation with some detail in order to show that this theory is able to solve three out of the five fundamental problems of hypothetico-deductive confirmation. largue that the remaining two problems are still open questions for the Bayesians. (shrink)

This paper defines in a formal and computational way the notion of ‘consilience’, a term introduced by Whewell in 1847 for the evaluation of scientific theories. Informally, as has been used to date, a model or theory is ‘consilient’ if it is predictive, explanatory and unifies the evide-nce. Centred in a constructive framework, where new terms can be intro-duced, we essay a formalisation of the idea of unification based on the avoidance of ‘sepa-ration’. However, it is soon manifest that this (...) classical approach is vulnerable to the introduction of fantastic concepts to unify disparate sub-theories. Our second approach is constructed by using a detailed evaluation of the relationship between the theory and the evid-ence by means of reinforcement propagation. With the use of reinfor-cement, fantastic concepts can be better detected and the role of con-silience for theory construction and revision can be specialised for dif-ferent inference mechanisms like explanatory induction, abduction, deduc-tion and analogy. (shrink)

In this three-part paper, my concern is to expound and defend a conception of science, close to Einstein's, which I call aim-oriented empiricism. I argue that aim-oriented empiricsim has the following virtues. (i) It solve the problem of induction; (ii) it provides decisive reasons for rejecting van Fraassen's brilliantly defended but intuitively implausible constructive empiricism; (iii) it solves the problem of verisimilitude, the problem of explicating what it can mean to speak of scientific progress given that science advances from one (...) false theory to another; (iv) it enables us to hold that appropriate scientific theories, even though false, can nevertheless legitimately be interpreted realistically, as providing us with genuine , even if only approximate, knowledge of unobservable physical entities; (v) it provies science with a rational, even though fallible and non-mechanical, method for the discovery of fundamental new theories in physics. In the third part of the paper I show that Einstein made essential use of aim-oriented empiricism in scientific practice in developing special and general relativity. I conclude by considering to what extent Einstein came explicitly to advocate aim-oriented empiricism in his later years. (shrink)

While I do not accept any current analysis of theoretical terms I also reject certain criticisms of them. Specifically, I reject the criticism that the paradoxes of material implication and the counterfactual problem eliminate the explicit definition view; and I also reject the criticism that explicitly defined theoretical terms do not refer to anything which "really exists" or do not have "excess meaning." I do argue, however, that the explicit definition view confuses and conflates the concepts of criterion and meaning (...) analysis. I also defend reduction sentences against the counterfactual difficulty, but show, too, how this view is already logically committed to the network or postulational view of meaning. Finally, I show how the concept of reduction sentences confuses in several ways the concepts of criterion and meaning analysis--although not in quite the same way as explicit definitions do. (shrink)

Inductive methods can be used to estimate the accuracies of inductive methods. Call a method immodest if it estimates that it is at least as accurate as any of its rivals. It would be unreasonable to adopt any but an immodest method. Under certain assumptions, exactly one of Carnap's lambda-methods is immodest. This may seem to solve the problem of choosing among the lambda-methods; but sometimes the immodest lambda-method is λ =0, which it would not be reasonable to adopt. We (...) should therefore reconsider the assumptions that led to this conclusion: for instance, the measure of accuracy. (shrink)

From the time of its birth, Popper’s theory of falsifiability has been fiercely criticized from various viewpoints. In the author’s view, however, those various criticisms all have the same root in their assumption that a falsification must be certain and conclusive. As the theory of falsifiability has never had such an assumption, it is the source of misunderstanding. By discarding it, we can reply to every criticism and thereby clarify the role of falsifiability in our search for better knowledge; that (...) is, it makes our attitude open to a yet unknown world. Key Words: Popper • falsifiability • openness • inconclusive falsification • second Popper legend. (shrink)

The following questions are discussed here. Is induction a reasonable procedure in the context of a denial of physically necessary connections? What is physical necessity? If induction does presuppose physical necessity, what amount of it is presupposed? It is argued that with logic as the only restriction on what is to count as a possible world, it is unreasonable to claim that observed connections, whether universal or statistical, will continue to hold. The concept of physical necessity is no more problematic (...) than that of logical necessity, once it is recognized that the necessity of physical and logical necessity is the same. A variant of Keynes' principle of limited independent variety answers the question of the amount of physical necessity presupposed. (shrink)

A consideration of some basic problems that arise in the attempt to provide an adequate characterization of statistical explanation is taken to show that an understanding of the nature of scientific explanation requires us to deal with the philosophical problems connected with the nature of prior probabilities.

The place of induction in the framing and test of scientific hypotheses is investigated. The meaning of 'induction' is first equated with generalization on the basis of case examination. Two kinds of induction are then distinguished: the inference of generals from particulars (first degree induction), and the generalization of generalizations (second degree induction). Induction is claimed to play a role in the framing of modest empirical generalizations and in the extension of every sort of generalizations--not however in the invention of (...) high-level hypotheses containing theoretical predicates. It is maintained, on the other hand, that induction by enumeration is essential in the empirical test of the lowest-level consequences of scientific theories, since it occurs in the drawing of "conclusions" from the examination of empirical evidence. But it is also held that the empirical test is insufficient, and must be supplemented with theorification, or the expansion of isolated hypotheses into theories. Refutation is not viewed as a substitute for confirmation but as its complement, since the very notion of unfavorable case is meaningful only in connection with the concept of positive instance. Although the existence of an inductive method is disclaimed, it is maintained that the various patterns of plausible reasoning (inductive inference included) are worth being investigated. It is concluded that scientific research follows neither the advice of inductivism nor the injunction of deductivism, but takes a middle course in which induction is instrumental both heuristically and methodologically, although the over-all pattern of research is hypothetico-deductive. (shrink)

The purpose of this paper is to make a simple observation regarding the Johnson -Carnap continuum of inductive methods. From the outset, a common criticism of this continuum was its failure to permit the confirmation of universal generalizations: that is, if an event has unfailingly occurred in the past, the failure of the continuum to give some weight to the possibility that the event will continue to occur without fail in the future. The Johnson -Carnap continuum is the mathematical consequence (...) of an axiom termed Johnson 's sufficientness postulate, the thesis of this paper is that, properly viewed, the failure of the Johnson -Carnap continuum to confirm universal generalizations is not a deep fact, but rather an immediate consequence of the sufficientness postulate; and that if this postulate is modified in the minimal manner necessary to eliminate such an entailment, then the result is a new continuum that differs from the old one in precisely one respect: it enjoys the desideratum of confirming universal generalizations. (shrink)

Most philosophers of science nowadays hold a network or postulational view of the meaning of theoretical words. However, there are many nuances to this view, and after explicitly separating them, we show what we take to be wrong with each one. While we reject the postulational view we do not defend its traditional alternatives either; rather we show the pointlessness of insisting on a single source for the meaning of theoretical words. We also point out the shortcomings of Carnap's newest (...) meaning criterion which depends upon a network view. But, again, we suggest not only that this new rendition of the criterion is faulty but also that there is something misguided about any search at all for such a criterion. (shrink)

Framing systematics as a field consistent with scientific inquiry entails that inferences of phylogenetic hypotheses have the goal of producing accounts of past causal events that explain differentially shared characters among organisms. Linking observations of characters to inferences occurs by way of why-questions implied by data matrices. Because of their form, why-questions require the use of common-cause theories. Such theories in phylogenetic inferences include natural selection and genetic drift. Selection or drift can explain ‘morphological’ characters but selection cannot be causally (...) applied to sequences since fitness differences cannot be directly associated with individual nucleotides or amino acids. The relation of selection to sequence data is by way of downward or top-down causation from those phenotypes upon which selection occurs. The application of phylogenetic inference to explain sequence data is thus restricted to instances where drift is the relevant theory; those nucleotides or amino acids that can be explained via downward causation are precluded from inclusion in the data matrix. The restrictions on the inclusion of sequence data in phylogenetic inferences equally apply to species hypotheses, precluding the more restrictive approach known as DNA barcoding. Not being able to discern drift and selection as relevant causal mechanisms can severely constrain the inclusion and explanations of sequence data. Implications of such exclusion are discussed in relation to the requirement of total evidence. (shrink)

The question of whether or not to partition data for the purposes of inferring phylogenetic hypotheses remains controversial. Opinions have been especially divided since Kluge's (1989, Systematic Zoology 38, 7–25) claim that data partitioning violates the requirement of total evidence (RTE). Unfortunately, advocacy for or against the RTE has not been based on accurate portrayals of the requirement. The RTE is a basic maxim for non-deductive inference, stipulating that evidence must be considered if it has relevance to an inference. Evidence (...) is relevant if it has a positive or negative effect on a given conclusion. In the case of ℈partitioned’ phylogenetic inferences, the RTE is violated, and the basis for rational belief in any conclusion is compromised, unless it is shown that the partitions are evidentially irrelevant to one another. The goal of phylogenetic systematics is to hypothesize past causal conditions to account for observed shared similarities among two or more species. Such inferences are non-deductive, necessitating consideration of the RTE. Some phylogeneticists claim the parsimony criterion as justification for the RTE. There is no relation between the two – parsimony is a relation between a hypothesis and causal question(s). Parsimony does not dictate the content of premises prior to an inference. ℈Taxonomic congruence,’ ℈supertrees,’ and ℈conditional combination’ methods violate the RTE. Taxonomic congruence and supertree methods also fail to achieve the intended goal of phylogenetic inference, such that ℈consensus trees’ and ℈supertrees’ lack an empirical basis. ℈Conditional combination’ is problematic because hypotheses derived from partitioned data cannot be compared – a causal hypothesis inferred to account for a set of effects only has relevance to those effects, not any comparative relevance to other causal hypotheses. A similar problem arises in the comparisons of hypotheses derived from different causal theories. (shrink)

This paper has two objectives, neither previously attempted in the published literature—first, to outline J. M. Keynes's theory of knowledge in some detail, and, secondly, to justify the contention that his epistemology is a variety of rationalism, and not, as many have asserted, a form of empiricism. Keynes's attitude to empirical data is also analysed as well as his views on prediction and theory choice. 1This paper is partly based on ideas initially advanced in O'Donnell [1982], a revised and expanded (...) version of which is to be published as O'Donnell [1989]. I should like to thank an anonymous referee for helpful comment, and King's College, Cambridge for permission to quote from the Keynes Papers. (shrink)