Search

Military AI optimists predict future AI assisting or making command decisions. We instead argue that, at a fundamental level, these predictions are dangerously wrong. The nature of war demands decisions based on abductive logic, whilst machine learning (or 'narrow AI') relies on inductive logic. The two forms of logic are not interchangeable, and therefore AI's limited utility in command – both tactical and strategic – is not something that can be solved by more data or more computing power. Many defence and government leaders are therefore proceeding with a false view of the nature of AI and of war itself. [ABSTRACT FROM AUTHOR]

A new solution strategy for the problem of induction has been developed based on a priori advantages of regret-weighted meta-induction (RW) in prediction tasks. These a priori advantages seem to contradict the no-free lunch (NFL) theorem. In this paper, the NFL challenge is dissolved by three novel results: (1) RW enjoys free lunches in the long run. (2) Yet, the NFL theorem applies to iterated prediction tasks, because the distribution underlying it assigns a zero probability to all possible worlds in which RW enjoys free lunches. (3) The a priori advantages of RW can even be demonstrated for the short run. [ABSTRACT FROM AUTHOR]

This is the introduction to the special issue "The Meta-Inductive Approach to Hume's Problem". The introduction includes introductory remarks and brief comments on each of the papers appearing in the special issue. [ABSTRACT FROM AUTHOR]

In this paper, the contributions to the account of meta-induction (Schurz 2019) collected in this volume are critically discussed and thereby, new insights are developed. How broad and expandable the program of meta-induction is can be learned from Ortner's contribution. New insights about the transition from the a priori justification of meta-induction to the a posteriori justification of object-induction emerge from the reflection of Shogenji's paper. How meta-induction may be applied also to religious prophecies and that their meta-inductive justification does not fail for a priori reasons but because of missing evidence for predictive success is learned from the discussion of Pitts' contribution. That meta-induction does not rely on a particular prior distribution, while the no free lunch theorem depends implicitly on a uniform prior, is the major conclusion drawn from the discussion of Wolpert's article. How the problem of induction is treated in different versions of the Bayesian account is learned from the discussion of Willliamson's paper. That meta-induction can also be employed for abduction, and that abductive theory-revision can offer meta-inductive aggregation methods is a new insight emerging from the reflection of Aliseda's contribution. [ABSTRACT FROM AUTHOR]

The subject of this essay is the classical problem of induction, which is sometimes attributed to David Hume and called “the Humean Problem of Induction.” Here, I examine a certain sort of Neo-Aristotelian solution to the problem, which appeals to the concept of natural kinds in its response to the inductive skeptic. This position is most notably represented by Howard Sankey and Marc Lange. The purpose of this paper is partly destructive and partly constructive. I raise two questions. The first is: Are the natural kind solutions to the problem successful? The first thesis of this paper is that they are not, and I will show how and why they fail. And the second question I raise here is: Is there nonetheless some alternative Neo-Aristotelian solution to the problem which is successful and can overcome the shortcomings endemic to the Sankey-Lange account? The second thesis is that there is, and I’ll attempt to sketch one. My stance here may be summarized by saying that, while I agree with Sankey and Lange that the problem of induction can be adequately resolved, and while I am on the whole sympathetic with the Aristotelian spirit of their account(s), I am, for all that, dissatisfied with the letter of them. Nothing short of a more thoroughgoing Aristotelianism about the epistemology of induction will do.

In this article, I seek to shed new light on a lesser-known stage of the development of Hans Reichenbach’s thought, namely his research, output and teaching activities at Istanbul University (1933–1938). I argue that the experience of Turkish exile was decisive in the elaboration of Reichenbach’s probability theory of meaning and knowledge. His work Experience and Prediction, produced while in Istanbul, should therefore be put in its Turkish context of elaboration and reception. To this end, I will take into consideration not only Reichenbach’s efforts to popularize and extend the Berlin Group’s program of scientific philosophy in Turkey and throughout Europe in the 1930s, but also the forgotten work of Reichenbach’s students—most of them women—at Istanbul University. [ABSTRACT FROM AUTHOR]

In this paper, I will discuss accounts to solve the problem of induction by introducing necessary connections. The basic idea is this: if we know that there are necessary connections between properties F and G such that F-ness necessarily brings about G-ness, then we are justified to infer that all, including future or unobserved, Fs will beGs. To solve the problem of induction with ontology has been proposed by David Armstrong and Brian Ellis. In this paper, I will argue that these attempts to solve the problem of induction fail. Necessary connections fail to reliably imply the respective regularities for two main reasons: Firstly, according to an argument originally presented by Helen Beebee, the respective necessary connections might be time-limited, and hence do not warrant inferences about future cases. As I will discuss, arguments against the possibility or explanatory power of time-limited necessary connections fail. Secondly, even time-unlimited necessary connections do not entail strict or non-strict regularities, and nor do they allow inferences about individual cases, which is an important function of inductive reasoning. Moreover, the proposed solution to the problem of induction would only apply to a tiny minority of inductive inferences. I argue that most inductive inferences are not easily reducible to the proposed inference pattern, as the vast majority of everyday inductive inferences do not involve necessary connections between fundamental physical properties or essences. [ABSTRACT FROM AUTHOR]

Copyright of Artefactos: Revista de Estudios Sobre La Ciencia Y La Tecnologia is the property of Ediciones Universidad de Salamanca and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The classical Indian school called Nyāya (literally "logic" or "right reasoning"), is arguably the leading anti-skeptical tradition within all of Indian philosophy. Defending a realist metaphysics and an epistemology of "knowledge sources" (pramāṇa), its responses to skepticism are often appropriated by other schools of thought. This paper examines its responses to skeptical arguments from dreams, from "the three times," from justificatory regress, and over the problem of induction. [ABSTRACT FROM AUTHOR]

The classical Indian Cārvāka ("Materialist") tradition contains three branches with regard to the means of knowledge (pramāṇas). First, the standard Cārvākas accept a single means of knowledge, perception, supporting this view with a critique of the reliability and coherence of inference (anumāna). Second, the "more educated" Cārvākas as well as Purandara endorse a form of inference limited to empirical matters. Third, radical skeptical Cārvākas like Jayarāśi attempt to undermine all accounts or technical definitions of the means of knowledge (even perception) in order to enjoy a life free from philosophical and religious speculation. These branches respectively present something akin to David Hume's problem of induction, endorse a fallibilistic, mitigated skepticism, and embody a thoroughgoing skepticism about philosophy itself. While all three branches are skeptics about religious matters, each branch exemplifies a different variety of epistemological skepticism. [ABSTRACT FROM AUTHOR]

In this paper Reichenbach's best alternative account (BAA) to induction is examined. In the first section, three versions of the BAA are distinguished that have been discussed in the literature. The major objections against all three versions are presented. In the second section it is shown by a text analysis that Reichenbach (The theory of probability, University of California Press, California, 1949) argues for all three versions of the BAA and does not sufficiently distinguish between them. In the third section it is explained how Reichenbach's third version of the BAA can be transformed into a provable optimality theorem within the account of meta-induction. [ABSTRACT FROM AUTHOR]

The no-free-lunch theorems promote a skeptical conclusion that all possible machine learning algorithms equally lack justification. But how could this leave room for a learning theory, that shows that some algorithms are better than others? Drawing parallels to the philosophy of induction, we point out that the no-free-lunch results presuppose a conception of learning algorithms as purely data-driven. On this conception, every algorithm must have an inherent inductive bias, that wants justification. We argue that many standard learning algorithms should rather be understood as model-dependent: in each application they also require for input a model, representing a bias. Generic algorithms themselves, they can be given a model-relative justification. [ABSTRACT FROM AUTHOR]

Classic inductive skepticism–the epistemological claim that we have no good reason to believe that the unobserved resembles the observed–is plausibly everyone's lot, whether or not they embrace Hume's metaphysical claim that distinct existents are "entirely loose and separate". But contemporary advocates of a Humean metaphysic accept a metaphysical claim stronger than Hume's own. I argue that their view plausibly gives rise to a radical inductive skepticism–according to which we are downright irrational in believing as we do about the unobserved–that we don't otherwise have reason to accept. The Metaphysical Neo-Humean is in an epistemological quagmire all her own. [ABSTRACT FROM AUTHOR]

This article discusses the problem of non-zero probabilities for non-tautologous universal generalizations in Rudolf Carnap's inductive logic (1950, 1952) when the domain of discourse is infinite. A solution is provided for a generalization of the form "all Xs are Ys", for example "all ravens all black". The solution is based on assuming that a significant part of the domain consists of non-Xs. This assumption can often be justified as a kind of ceteris paribus principle. [ABSTRACT FROM AUTHOR]

In the first chapter of his book Realism and the Aim of Science, Karl Popper establishes a close connection between the problem of induction and corroboration. In doing so, he formulates a family of four problems of induction. The presented paper after short insight to some Popper's methodological rules analyzes these four problems of induction, namely how can we distinguish between good and bad theories; the problem of rational belief; the problem of tomorrow; and the problem of regularities. Although Popper was convinced that he had solved the problem of induction completely logically, methodically as well as epistemologically, and a metaphysical solution is not necessary, we tried to show that neither did he solve the problem as he claims, nor he implicitly use a principle of induction, as some of his critics claims. [ABSTRACT FROM AUTHOR]

C. D. Broad famously labelled the problem of providing our inductive practices with a proper justification "the scandal of philosophy" (Broad, 1952). Recently, John Norton has provided a dissolution of this problem (2014). According to Norton, inductive inference is grounded in particular facts obtaining within particular domains (J. Norton, 2003b, 2010, 2014). Because the material theory does not involve a universal schema of induction, Norton claims it dissolves the problem of induction (which implies that such universal schemas cannot be justified). In this paper, I critically evaluate Norton's dissolution. In particular, I argue that the problem of induction is an epistemological problem, that Norton's material theory entails an externalist epistemology, and that it is a common feature of such epistemologies that they dissolve the problem of induction. The upshot is that the material theory is not unique in its ability to reap the specifically epistemic benefits of dissolving the problem of induction, and thus that the epistemic advantages of the material theory over extant alternatives in this regard are fewer than it may appear at first sight. • Inductive logic influences inductive epistemology. • Material theory leads to externalist epistemology. • Externalist epistemology generally dissolves problem of induction. • Epistemic benefits of the material theory of induction unclear. [ABSTRACT FROM AUTHOR]

In the first section, five major attempts to solve the problem of induction and their failures are discussed. In the second section, an account of meta-induction is introduced. It offers a novel solution to the problem of induction, based on mathematical theorems about the predictive optimality of attractivity-weighted meta-induction. In the third section, how the a priori justification of meta-induction provides a non-circular a posteriori justification of object-induction, based on its superior track record, is explained. In the fourth section, four important extensions and refinements of the method of meta-induction are presented. The final section, summarizes the major impacts of the program of meta-induction for epistemology, the philosophy of science and cognitive science.

Over 40 years ago, I put forward a philosophy of science based on the argument that physics, in only ever accepting unified theories, thereby makes a substantial metaphysical presupposition about the universe, to the effect that it possesses an underlying unity. I argued that a new conception of scientific method is required to subject this problematic presupposition to critical attention so that it may be improved as science proceeds. This view has implications for the study of the metaphysics of science. The view has however been ignored by recent contributions to the field. Broader implications of the view are indicated. Finally, reasons for the neglect of the view are considered. [ABSTRACT FROM AUTHOR]

The text focuses on the so called "paradox of ravens" formulated in the 1940s by C. G. Hempel, which challenges the possibility of confirming general hypotheses by observation of particular instances of these hypotheses. After a brief presentation of the paradox and some of the possible solutions of the paradox based on the works of C. G. Hempel, K. R. Popper and W. V. O. Quine, the author presents his own analysis of the paradox that is based on understanding hypotheses challenged by the paradox not as describing general natural laws, but as concerning merely actual instances of such hypotheses. The author claims that on this interpretation of the paradox it is possible to avoid the problematic conclusion if we abandon ontological position known as "metaphysical realism", because refusing this position will disqualify the terms like "non-black" or "non-raven" from being applicable in the process of confirmation - and without these terms the paradox cannot even arise in the first place. [ABSTRACT FROM AUTHOR]

Is there any kind of justification fora belief that the future will be largely like the past? In the eighteenth century David Hume formulatedthe problem of induction in a similar spirit, andhis version is now conventional. Even two hundred and fifty years after Hume’s death, the ef-fort to justify inductive inferences still preoccupiesphilosophers. Sir Karl Popper believed that he hadsuccessfully solved this fundamental philosophicalproblem. The present article introduces the problematic situation formulated by David Hume, followed by Popper’s original solution to the problem. Along with that, relevant criticism from John Worall and Wes Salmon will be presented. It concentrates on Popper’s rejection of induction and therelated inability to implement any prediction aboutthe future within critical rationalism. The validityor failure of Popper’s solution is discussed towardsthe end of the article, together with a reconsideration of the current state of the debate. The lastpart of the study is devoted to the author’s proposal, which should be able to stand up againstanti-Popperian objections while remaining withinthe boundaries of critical-rationalist methodology.

Military AI optimists predict future AI assisting or making command decisions. We instead argue that, at a fundamental level, these predictions are dangerously wrong. The nature of war demands decisions based on abductive logic, whilst machine learning (or 'narrow AI') relies on inductive logic. The two forms of logic are not interchangeable, and therefore AI's limited utility in command - both tactical and strategic - is not something that can be solved by more data or more computing power. Many defence and government leaders are therefore proceeding with a false view of the nature of AI and of war itself., Competing Interests: No potential conflict of interest was reported by the author(s)., (© 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.)

I here aim to show that a particular approach to the problem of induction, which I will call "induction by direct inference", comfortably handles Goodman's problem of induction. I begin the article by describing induction by direct inference. After introducing induction by direct inference, I briefly introduce the Goodman problem, and explain why it is, prima facie, an obstacle to the proposed approach. I then show how one may address the Goodman problem, assuming one adopts induction by direct inference as an approach to the problem of induction. In particular, I show that a relatively standard treatment of what some have called the "Reference Class Problem" addresses the Goodman Problem. Indeed, plausible and relatively standard principles of direct inference yield the conclusion that the Goodman inference (involving the grue predicate) is defeated, so it is unnecessary to invoke considerations of 'projectibility' in order to address the Goodman problem. I conclude the article by discussing the generality of the proposed approach, in dealing with variants of Goodman's example. [ABSTRACT FROM AUTHOR]