Scientists depend on complex computational systems that are often ineliminably opaque, to the detriment of our ability to give scientific explanations and detect artifacts. Some philosophers have s...

Hunting Causes and Using Them argues that causation is not one thing, as commonly assumed, but many. There is a huge variety of causal relations, each with different characterizing features, different methods for discovery and different uses to which it can be put. In this collection of new and previously published essays, Nancy Cartwright provides a critical survey of philosophical and economic literature on causality, with a special focus on the currently fashionable Bayes-nets and invariance methods - and it exposes (...) a huge gap in that literature. Almost every account treats either exclusively how to hunt causes or how to use them. But where is the bridge between? It's no good knowing how to warrant a causal claim if we don't know what we can do with that claim once we have it. This book will interest philosophers, economists and social scientists. (shrink)

Everyone has heard the claim, "Correlation does not imply causation." What might sound like a reasonable dictum metastasized in the twentieth century into one of science's biggest obstacles, as a legion of researchers became unwilling to make the claim that one thing could cause another. Even two decades ago, asking a statistician a question like "Was it the aspirin that stopped my headache?" would have been like asking if he believed in voodoo, or at best a topic for conversation at (...) a cocktail party rather than a legitimate target of scientific inquiry. Scientists were allowed to posit only that the probability that one thing was associated with another. This all changed with Judea Pearl, whose work on causality was not just a victory for common sense, but a revolution in the study of the world. (shrink)

In his new foreword to this edition, Hilary Putnam forcefully rejects these nativist claims.

Many of the computing systems programmed using Machine Learning are opaque: it is difficult to know why they do what they do or how they work. Explainable Artificial Intelligence aims to develop analytic techniques that render opaque computing systems transparent, but lacks a normative framework with which to evaluate these techniques’ explanatory successes. The aim of the present discussion is to develop such a framework, paying particular attention to different stakeholders’ distinct explanatory requirements. Building on an analysis of “opacity” from (...) philosophy of science, this framework is modeled after accounts of explanation in cognitive science. The framework distinguishes between the explanation-seeking questions that are likely to be asked by different stakeholders, and specifies the general ways in which these questions should be answered so as to allow these stakeholders to perform their roles in the Machine Learning ecosystem. By applying the normative framework to recently developed techniques such as input heatmapping, feature-detector visualization, and diagnostic classification, it is possible to determine whether and to what extent techniques from Explainable Artificial Intelligence can be used to render opaque computing systems transparent and, thus, whether they can be used to solve the Black Box Problem. (shrink)

Simple idealized models seem to provide more understanding than opaque, complex, and hyper-realistic models. However, an increasing number of scientists are going in the opposite direction by utilizing opaque machine learning models to make predictions and draw inferences, suggesting that scientists are opting for models that have less potential for understanding. Are scientists trading understanding for some other epistemic or pragmatic good when they choose a machine learning model? Or are the assumptions behind why minimal models provide understanding misguided? In (...) this paper, using the case of deep neural networks, I argue that it is not the complexity or black box nature of a model that limits how much understanding the model provides. Instead, it is a lack of scientific and empirical evidence supporting the link that connects a model to the target phenomenon that primarily prohibits understanding. (shrink)

From a reliabilist point of view, our inferential practices make us into instruments for determining the truth value of hypotheses where, like all instruments, reliability is a central virtue. I apply this perspective to second-order inductions, the inductive assessments of inductive practices. Such assessments are extremely common, for example whenever we test the reliability of our instruments or our informants. Nevertheless, the inductive assessment of induction has had a bad name ever since David Hume maintained that any attempt to justify (...) induction by means of an inductive argument must beg the question. I will consider how the inductive justification of induction fares from the reliabilist point of view. I will also consider two other wellknown arguments that can be construed as inductive assessments of induction. One is the miracle argument, according to which the truth of scientific theories should be inferred as the best explanation of their predictive success; the other is the disaster argument, according to which we should infer that all present and future theories are false on the grounds that all past theories have been found to be false. (shrink)

[Peter Lipton] From a reliabilist point of view, our inferential practices make us into instruments for determining the truth value of hypotheses where, like all instruments, reliability is a central virtue. I apply this perspective to second-order inductions, the inductive assessments of inductive practices. Such assessments are extremely common, for example whenever we test the reliability of our instruments or our informants. Nevertheless, the inductive assessment of induction has had a bad name ever since David Hume maintained that any attempt (...) to justify induction by means of an inductive argument must beg the question. I will consider how the inductive justification of induction fares from the reliabilist point of view. I will also consider two other well-known arguments that can be construed as inductive assessments of induction. One is the miracle argument, according to which the truth of scientific theories should be inferred as the best explanation of their predictive success; the other is the disaster argument, according to which we should infer that all present and future theories are false on the grounds that all past theories have been found to be false. \\\ [John Worrall] Science seems in some ways to have been remarkably successful. What does this success tell us about the epistemological status of current scientific claims? Peter Lipton considers various meta-inductive arguments each of which start from premises about science's 'track record'. I show that his endorsements of the 'strongest' of these are, on analysis, remarkably weak. I argue that this is a reflection of difficulties within the general epistemological framework that he adopts-that of reliabilism. Finally, I briefly outline the quite different approach that I take to this issue, in the process responding to Lipton's criticisms of the 'pessimistic meta-induction'. (shrink)

From a reliabilist point of view, our inferential practices make us into instruments for determining the truth value of hypotheses where, like all instruments, reliability is a central virtue. I apply this perspective to second-order inductions, the inductive assessments of inductive practices. Such assessments are extremely common, for example whenever we test the reliability of our instruments or our informants. Nevertheless, the inductive assessment of induction has had a bad name ever since David Hume maintained that any attempt to justify (...) induction by means of an inductive argument must beg the question. I will consider how the inductive justification of induction fares from the reliabilist point of view. I will also consider two other well-known arguments that can be construed as inductive assessments of induction. One is the miracle argument, according to which the truth of scientific theories should be inferred as the best explanation of their predictive success; the other is the disaster argument, according to which we should infer that all present and future theories are false on the grounds that all past theories have been found to be false. \\\ [John Worrall] Science seems in some ways to have been remarkably successful. What does this success tell us about the epistemological status of current scientific claims? Peter Lipton considers various meta-inductive arguments each of which start from premises about science's 'track record'. I show that his endorsements of the 'strongest' of these are, on analysis, remarkably weak. I argue that this is a reflection of difficulties within the general epistemological framework that he adopts-that of reliabilism. Finally, I briefly outline the quite different approach that I take to this issue, in the process responding to Lipton's criticisms of the 'pessimistic meta-induction'. (shrink)

Complex systems research is becoming ever more important in both the natural and social sciences. It is commonly implied that there is such a thing as a complex system, different examples of which are studied across many disciplines. However, there is no concise definition of a complex system, let alone a definition on which all scientists agree. We review various attempts to characterize a complex system, and consider a core set of features that are widely associated with complex systems in (...) the literature and by those in the field. We argue that some of these features are neither necessary nor sufficient for complexity, and that some of them are too vague or confused to be of any analytical use. In order to bring mathematical rigour to the issue we then review some standard measures of complexity from the scientific literature, and offer a taxonomy for them, before arguing that the one that best captures the qualitative notion of the order produced by complex systems is that of the Statistical Complexity. Finally, we offer our own list of necessary conditions as a characterization of complexity. These conditions are qualitative and may not be jointly sufficient for complexity. We close with some suggestions for future work. (shrink)

In March 2016, DeepMind's computer programme AlphaGo surprised the world by defeating the world‐champion Go player, Lee Sedol. AlphaGo exhibits a novel, surprising and valuable style of play and has been recognised as “creative” by the artificial intelligence (AI) and Go communities. This article examines whether AlphaGo engages in creative problem solving according to the standards of comparative psychology. I argue that AlphaGo displays one important aspect of creative problem solving (namely mental scenario building in the form of Monte Carlo (...) tree search), while lacking another (domain generality). This analysis has consequences for how we think about creativity in humans and AI. (shrink)

When graduate students start their studies, they usually have sound knowledge of some areas of philosophy, but the overall map of their knowledge is often patchy and disjointed. There are a number of topics that any contemporary philosopher working in any part of the analytic tradition needs to grasp, and to grasp as a coherent whole rather than a rag-bag of interesting but isolated discussions. This book answers this need, by providing a overview of core topics in metaphysics and epistemology (...) that is at once accessible and nuanced. Ten core topics are explained, and their relation to each other is clearly set out. The book emphasizes the utility of the concepts and distinctions it covers for philosophy as a whole, not just for specialist discussions in metaphysics or epistemology. The text is highly readable and may be used as the basis of a course on these topics. Recommendations for reading are included at the end of each chapter, divided into essential and further readings. The text is also suitable for people approaching philosophy from other disciplines, as an accessible primer to the central topics, concepts and distinctions that are needed to engage meaningfully in contemporary philosophical debate. (shrink)

Receive free email alerts when new articles cite this article - sign up in the box at the top right corner of the article..