We propose a formal framework for interpretable machine learning. Combining elements from statistical learning, causal interventionism, and decision theory, we design an idealised explanation game in which players collaborate to find the best explanation for a given algorithmic prediction. Through an iterative procedure of questions and answers, the players establish a three-dimensional Pareto frontier that describes the optimal trade-offs between explanatory accuracy, simplicity, and relevance. Multiple rounds are played at different levels of abstraction, allowing the players to explore overlapping causal (...) patterns of variable granularity and scope. We characterise the conditions under which such a game is almost surely guaranteed to converge on a optimal explanation surface in polynomial time, and highlight obstacles that will tend to prevent the players from advancing beyond certain explanatory thresholds. The game serves a descriptive and a normative function, establishing a conceptual space in which to analyse and compare existing proposals, as well as design new and improved solutions. (shrink)

Machine learning algorithms may radically improve our ability to diagnose and treat disease. For moral, legal, and scientific reasons, it is essential that doctors and patients be able to understand and explain the predictions of these models. Scalable, customisable, and ethical solutions can be achieved by working together with relevant stakeholders, including patients, data scientists, and policy makers.

Artificial intelligence has historically been conceptualized in anthropomorphic terms. Some algorithms deploy biomimetic designs in a deliberate attempt to effect a sort of digital isomorphism of the human brain. Others leverage more general learning strategies that happen to coincide with popular theories of cognitive science and social epistemology. In this paper, I challenge the anthropomorphic credentials of the neural network algorithm, whose similarities to human cognition I argue are vastly overstated and narrowly construed. I submit that three alternative supervised learning (...) methods—namely lasso penalties, bagging, and boosting—offer subtler, more interesting analogies to human reasoning as both an individual and a social phenomenon. Despite the temptation to fall back on anthropomorphic tropes when discussing AI, however, I conclude that such rhetoric is at best misleading and at worst downright dangerous. The impulse to humanize algorithms is an obstacle to properly conceptualizing the ethical challenges posed by emerging technologies. (shrink)

The term ‘coherence’ (and its antonym ‘incoherence’) is used in a bewildering variety of ways in epistemology (and in philosophy more broadly). This entry attempts to bring some discipline to uses of the term by offering a taxonomy of notions of coherence (and incoherence), and then surveying which of the resulting notions is (or should be) at work in the various different contexts in which it is deployed.

This paper elaborates and defends a familiar ‘transparent’ account of knowledge of one's own beliefs, inspired by some remarks of Gareth Evans, and makes a case that the account can be extended to mental states in general, in particular to knowledge of one's intentions.

Many believe that we ought to save a large number from being permanently bedridden rather than save one from death. Many also believe that we ought to save one from death rather than a multitude from a very minor harm, no matter how large this multitude. I argue that a principle I call “Aggregate Relevant Claims” satisfactorily explains these judgments. I offer a rationale for this principle and defend it against objections.

The Unity of Perception offers a grand synoptic vision of how perception, consciousness and knowledge fit together. It is a remarkable achievement. A short comment can only address fragments of Schellenberg’s picture; naturally I will look for weak spots.

Are individuals morally responsible for their implicit biases? One reason to think not is that implicit biases are often advertised as unconscious, ‘introspectively inaccessible’ attitudes. However, recent empirical evidence consistently suggests that individuals are aware of their implicit biases, although often in partial and inarticulate ways. Here I explore the implications of this evidence of partial awareness for individuals’ moral responsibility. First, I argue that responsibility comes in degrees. Second, I argue that individuals’ partial awareness of their implicit biases makes (...) them (partially) morally responsible for them. I argue by analogy to a close relative of implicit bias: moods. (shrink)

Dispositional theories of colour have been attacked by McGinn and others on the ground that ‘Colours do not look like dispositions’. Langsam has argued that on the contrary they do, in ‘Why Colours Do Look Like Dispositions’, The Philosophical Quarterly, 50 , pp. 68–75. I make three claims. First, neither side has made its case. Secondly, it is true, at least on one interpretation, that colours do not look like dispositions. Thirdly, this does not show that dispositionalism about colours is (...) false. (shrink)

Many philosophers and psychologists have attempted to elucidate the nature of mental representation by appealing to notions like isomorphism or abstract structural resemblance. The ‘structural representations’ that these theorists champion are said to count as representations by virtue of functioning as internal models of distal systems. In his 2007 book, Representation Reconsidered, William Ramsey endorses the structural conception of mental representation, but uses it to develop a novel argument against representationalism, the widespread view that cognition essentially involves the manipulation of (...) mental representations. Ramsey argues that although theories within the ‘classical’ tradition of cognitive science once posited structural representations, these theories are being superseded by newer theories, within the tradition of connectionism and cognitive neuroscience, which rarely if ever appeal to structural representations. Instead, these theories seem to be explaining cognition by invoking so-called ‘receptor representations’, which, Ramsey claims, aren’t genuine representations at all—despite being called representations, these mechanisms function more as triggers or causal relays than as genuine stand-ins for distal systems. I argue that when the notions of structural and receptor representation are properly explicated, there turns out to be no distinction between them. There only appears to be a distinction between receptor and structural representations because the latter are tacitly conflated with the ‘mental models’ ostensibly involved in offline cognitive processes such as episodic memory and mental imagery. While structural representations might count as genuine representations, they aren’t distinctively mental representations, for they can be found in all sorts of non-intentional systems such as plants. Thus to explain the kinds of offline cognitive capacities that have motivated talk of mental models, we must develop richer conceptions of mental representation than those provided by the notions of structural and receptor representation. (shrink)

The remaining seven papers (eight, if you count this introductory piece) in this volume of Oxford Studies in Epistemology constitute a special issue on applied epistemology, an exciting, novel, and currently burgeoning subfield of epistemology. The term ‘applied epistemology’ is a relatively recent one, however, and anecdotally, many people I’ve encountered are not quite sure what it denotes, or what different works within the field have in common. In this introductory piece, I’ll venture some views about these questions, and about (...) why applied epistemology is worth doing, as well as about its dangers. Doing so will set the state for me to situate the papers in this volume within the subfield. (shrink)

Necessity and sufficiency are the building blocks of all successful explanations. Yet despite their importance, these notions have been conceptually underdeveloped and inconsistently applied in explainable artificial intelligence (XAI), a fast-growing research area that is so far lacking in firm theoretical foundations. Building on work in logic, probability, and causality, we establish the central role of necessity and sufficiency in XAI, unifying seemingly disparate methods in a single formal framework. We provide a sound and complete algorithm for computing explanatory factors (...) with respect to a given context, and demonstrate its flexibility and competitive performance against state of the art alternatives on various tasks. (shrink)

Recent work on rationality has been increasingly attentive to “coherence requirements”, with heated debates about both the content of such requirements and their normative status (e.g., whether there is necessarily reason to comply with them). Yet there is little to no work on the metanormative status of coherence requirements. Metaphysically: what is it for two or more mental states to be jointly incoherent, such that they are banned by a coherence requirement? In virtue of what are some putative requirements genuine (...) and others not? Epistemologically: how are we to know which of the requirements are genuine and which aren’t? This paper tries to offer an account that answers these questions. On my account, the incoherence of a set of attitudinal mental states is a matter of its being (partially) constitutive of the mental states in question that, for any agent that holds these attitudes jointly, the agent is disposed, when conditions of full transparency are met, to give up at least one of the attitudes. (shrink)

How do I know that I see a cat? A curiously under-asked question. The paper tries to answer it.

This paper offers a positive account of an important but under-explored class of mental states, non-propositional attitudes such as loving one’s department, liking lattice structures, fearing Freddy Krueger, and hating Sherlock Holmes. In broadest terms, the view reached is a representationalist account guided by two puzzles. The proposal allows one to say in an elegant way what differentiates a propositional attitude from an attitude merely about a proposition. The proposal also allows one to offer a unified account of the non-propositional (...) attitudes that captures both empty and non-empty cases by properly locating the posited representations in the metaphysical structure of the attitudes. (shrink)

Should we understand implicit attitudes on the model of belief? I argue that implicit attitudes are (probably) members of a different psychological kind altogether, because they seem to be insensitive to the logical form of an agent’s thoughts and perceptions. A state is sensitive to logical form only if it is sensitive to the logical constituents of the content of other states (e.g., operators like negation and conditional). I explain sensitivity to logical form and argue that it is a necessary (...) condition for belief. I appeal to two areas of research that seem to show that implicit attitudes fail spectacularly to satisfy this condition—although persistent gaps in the empirical literature leave matters inconclusive. I sketch an alternative account, according to which implicit attitudes are sensitive merely to spatiotemporal relations in thought and perception, i.e., the spatial and temporal orders in which people think, see, or hear things. (shrink)

Are women (simply) adult human females? Dictionaries suggest that they are. However, philosophers who have explicitly considered the question invariably answer no. This paper argues that they are wrong. The orthodox view is that the category *woman* is a social category, like the categories *widow* and *police officer*, although exactly what this social category consists in is a matter of considerable disagreement. In any event, orthodoxy has it that *woman* is definitely not a biological category, like the categories *amphibian* or (...) *adult human female*. -/- In the first part, a number of arguments are given for the view that women are adult human females; the second part turns to rebutting the main objections. Finally, a couple of morals are briefly noted: one for activist sloganeering, and one for ameliorative projects that seek to change the meaning of ‘woman’. (shrink)

According to orthodoxy, perceptual beliefs are caused by perceptual experiences. The paper argues that this view makes it impossible to explain how experiences can be epistemically significant. A rival account, on which experiences in the “good case” are ways of knowing, is set out and defended.

It’s interesting and a bit surprising how little attention philosophy has given to the status of emoji, those funny little symbols that punctuate text messages, Twitter, and other digital spaces. They have become ubiquitous, but maybe because they’re seen as frivolous or a “lower” form of communication, philosophy hasn’t paid them much mind. But they are an interesting aesthetic phenomenon. They are part language, part representational image. They are phenomenologically interesting in their effect on how we experience the written word. (...) They punctuate, accentuate, emphasize, and add flavor to our communication in ways that are difficult to achieve otherwise. It would not be ridiculous to say that they represent a genuine linguistic development—a change in conventional orthography, and of an almost unbelievably sudden and dramatic, even revolutionary, kind. (shrink)

Comment on David Papineau's _The Metaphysics of Sensory Experience_.

In this paper it is argued that we should amend the traditional understanding of the view known as the guise of the good. The guise of the good is traditionally understood as the view that we only want to act in ways that we believe to be good in some way. But it is argued that a more plausible view is that we only want to act in ways that we believe we have normative reason to act in. This change (...) – from formulating the view in terms of goodness to formulating it in terms of reasons – is significant because the revised view avoids various old and new counterexamples to the traditional view, because the revised view is better motivated than the traditional view, and because the revised view is better placed to explain certain features of desire than the traditional view. The paper finishes by showing that the conclusions reached are compatible with theories such as the buck passing account of value. (shrink)

We offer a summary of the WHO Report "Making Fair Choices on the Path to Universal Health Coverage".

A Letter to the Editor about the origin of "gender identity" and deficiencies in its current definition.

The foundations of research ethics are riven with fault lines emanating from a fear that if research is too closely connected to weighty social purposes an imperative to advance the common good through research will justify abrogating the rights and welfare of study participants. The result is an impoverished conception of the nature of research, an incomplete focus on actors who bear important moral responsibilities, and a system of ethics and oversight highly attuned to the dangers of research but largely (...) silent about threats of ineffective, inefficient, and inequitable medical practices and health systems. -/- In For the Common Good: Philosophical Foundations of Research Ethics, Alex John London defends a conception of the common good that grounds a moral imperative with two requirements. The first is to promote research that generates the information necessary to enable key social institutions to effectively, efficiently, and equitably safeguard the basic interests of individuals. The second is to ensure that research is organized as a voluntary scheme of social cooperation that respects its various contributors' moral claims to be treated as free and equal. Connecting research to the goals of a just social order grounds a framework for assessing and managing research risk that reconciles these requirements and justifies key oversight practices in non-paternalistic terms. Reconceiving research ethics as resolving coordination problems and providing credible assurance that these requirements are being met expands the issues and actors that fall within the purview of the field and provides the foundation for a more unified and coherent approach to domestic and international research. -/- This is an open access title available under the terms of a CC BY-NC-ND 4.0 license. It is free to read at Oxford Scholarship Online and offered as a free PDF download from OUP and selected open access locations. (shrink)

Most contemporary philosophical discussions of intentionality start and end with a treatment of the propositional attitudes. In fact, many theorists hold that all attitudes are propositional attitudes. Our folk-psychological ascriptions suggest, however, that there are non-propositional attitudes: I like Sally, my brother fears snakes, everyone loves my grandmother, and Rush Limbaugh hates Obama. I argue that things are as they appear: there are non-propositional attitudes. More specifically, I argue that there are attitudes that relate individuals to non-propositional objects and do (...) so not in virtue of relating them to propositions. I reach this conclusion by not only showing that attempted analyses of apparently non-propositional attitudes in terms of the propositional fail, but that some non-propositional attitudes don’t even supervene on propositional attitudes. If this is correct, then the common discussions of intentionality that address only propositional attitudes are incomplete and those who hold that all intentional states are propositional are mistaken. (shrink)

In this paper, I defend a new theory of normative reasons called reasons as good bases, according to which a normative reason to φ is something that is a good basis for φing. The idea is that the grounds on which we do things—bases—can be better or worse as things of their kind, and a normative reason—a good reason—is something that is just a good instance of such a ground. After introducing RGB, I clarify what it is to be a (...) good basis, and argue that RGB has various attractive features: it has intuitive implications, makes good sense of the weights of reasons, and attractively explains the relationship between normative reasons and motivating reasons. I then briefly defend the view from objections and compare it to rivals. Finally, I sketch two possible implications of RGB: some kind of constitutivism, according to which the norms that govern us are explained by the nature of agency, and second, the claim that agents who do things for reasons generally do them for good reasons. (shrink)

This chapter is centered around an apparent tension that research on implicit bias raises between virtue and social knowledge. Research suggests that simply knowing what the prevalent stereotypes are leads individuals to act in prejudiced ways—biasing decisions about whom to trust and whom to ignore, whom to promote and whom to imprison—even if they reflectively reject those stereotypes. Because efforts to combat discrimination obviously depend on knowledge of stereotypes, a question arises about what to do next. This chapter argues that (...) the obstacle to virtue is not knowledge of stereotypes as such, but the “accessibility” of such knowledge to the agent who has it. “Accessibility” refers to how easily knowledge comes to mind. Social agents can acquire the requisite knowledge of stereotypes while resisting their pernicious influence, so long as that knowledge remains, in relevant contexts, relatively inaccessible. (shrink)

When one visually hallucinates, the object of one’s hallucination is not before one’s eyes. On the standard view, that is because the object of hallucination does not exist, and so is not anywhere. Many different defenses of the standard view are on offer; each have problems. This paper defends the view that there is always an object of hallucination—a physical object, sometimes with spatiotemporally scattered parts.

Priority monism is the view that the cosmos is the only independent concrete object. The paper argues that, pace its proponents, Priority monism is in conflict with the dependence of any whole on any of its parts: if the cosmos does not depend on its parts, neither does any smaller composite.

Comment on Brogaard's Seeing and Saying (OUP 2018).

Standard responses to the knowledge argument grant that Mary could know all of the physical facts even while trapped inside her black‐and‐white room. What they deny is that upon leaving her black‐and‐white room and experiencing red for the first time, Mary learns a genuinely new fact. This paper develops an alternate response in a grounding physicalist framework, on which Mary does not know all of the physical facts while trapped inside the room. The main thesis is that Mary does not (...) know certain phenomenal facts while trapped inside the room, whereby these facts classify as physical due to being wholly and fully metaphysically grounded in the underlying fundamental facts which are themselves entirely physical. (shrink)

Remembering a cat sleeping (specifically, recollecting the way the cat looked), perceiving (specifically, seeing) a cat sleeping, and imagining (specifically, visualizing) a cat sleeping are of course importantly different. Nonetheless, from the first-person perspective they are palpably alike. The paper addresses two questions: Q1. What are these similarities (and differences)? Q2. How does one tell that one is recalling (and so not perceiving or imagining)?

In 'Rethinking the Good', Larry Temkin makes two core claims. First, the goodness of a distribution is sometimes ‘essentially comparative’ – it sometimes depends on which alternative distribution(s) it is compared to. Second, such cases threaten the transitivity of ‘all things considered better than’. I argue that the goodness of a distribution may indeed depend on what other distributions are feasible. But contrary to Temkin, I also argue that transitivity holds even when the goodness of a distribution depends on the (...) feasible set of which it is a member. (shrink)

What might it mean to read video games as a source for critical theological reflection? This project explores the extent to which FromSoftware’s 2015 classic Bloodborne can be read in just this way for at least two reasons. Firstly, the game makes visible foundational but often taken-for-granted Christian themes, as well as their internal paradoxes. Secondly, Bloodborne’s gameplay is structured through these themes in ways that invite the player to take on a mode of theological reflection which allows them to (...) think creatively beyond the conditions giving rise to these paradoxes. (shrink)

Disagreement is a hot topic in epistemology. A fast-growing literature centers around a dispute between the ‘steadfast’ view, on which one may maintain one’s beliefs even in the light of disagreement with epistemic peers who have all the same evidence, and the ‘conciliationist’ view, on which such disagreement requires a revision of attitudes. In this paper, however, I argue that there is less separating the main rivals in the debate about peer disagreement than is commonly thought. The extreme versions of (...) both views are clearly indefensible, while more moderate versions of the views converge on the idea that how much revision of belief is called for by an instance of peer disagreement varies from case to case. Those tempted by this diagnosis are sometimes pessimistic about the prospects for giving a unified account which clearly predicts when more or less extensive revisions will be called for. By contrast, in this paper I give an account that aspires to such unity and predictive power, centering on the notion of the net resilience of your estimate of your own reliability against your estimate of your interlocutor’s reliability. The view I present thus amounts to a new, moderate theory of how one should respond to disagreement. I argue that ultimately, when we weaken conciliationism and the steadfast view to account for exception cases and to make them adequately plausible, they end up converging on the moderate view I present. Much of the seeming disagreement about disagreement is, then, illusory. (shrink)

Research suggests that interventions involving extensive training or counterconditioning can reduce implicit prejudice and stereotyping, and even susceptibility to stereotype threat. This research is widely cited as providing an “existence proof” that certain entrenched social attitudes are capable of change, but is summarily dismissed—by philosophers, psychologists, and activists alike—as lacking direct, practical import for the broader struggle against prejudice, discrimination, and inequality. Criticisms of these “debiasing” procedures fall into three categories: concerns about empirical efficacy, about practical feasibility, and about the (...) failure to appreciate the underlying structural-institutional nature of discrimination. I reply to these criticisms of debiasing, and argue that a comprehensive strategy for combating prejudice and discrimination should include a central role for training our biases away. (shrink)

The difference between the unity of the individual and the separateness of persons requires that there be a shift in the moral weight that we accord to changes in utility when we move from making intrapersonal tradeoffs to making interpersonal tradeoffs. We examine which forms of egalitarianism can, and which cannot, account for this shift. We argue that a form of egalitarianism which is concerned only with the extent of outcome inequality cannot account for this shift. We also argue that (...) a view which is concerned with both outcome inequality and with the unfairness of inequality in individuals‘ expected utilities can account for this shift. Finally, we limn an alternative view, on.. (shrink)

Hill's views on visual experience are critically examined.

This chapter explores the idea that causal inference is warranted if and only if the mechanism underlying the inferred causal association is identified. This mechanistic stance is discernible in the epidemiological literature, and in the strategies adopted by epidemiologists seeking to establish causal hypotheses. But the exact opposite methodology is also discernible, the black box stance, which asserts that epidemiologists can and should make causal inferences on the basis of their evidence, without worrying about the mechanisms that might underlie their (...) hypotheses. I argue that the mechanistic stance is indeed a bad methodology for causal inference. However, I detach and defend a mechanistic interpretation of causal generalisations in epidemiology as existence claims about underlying mechanisms. (shrink)

Although philosophers of art and aesthetics regularly appeal to a notion of ‘pictorial content’, there is little agreement over its nature. The present paper argues that pictures have propositional contents. This conclusion is reached by considering a style of argument having to do with the phenomenon of negation intended to show that pictures must have some kind of non-propositional content. I first offer reasons for thinking that arguments of that type fail. Second, I show that when properly understood, such arguments (...) can in fact be turned on their heads and shown to support the propositionalist position. (shrink)

In this paper, I defend the view that it is wrong for us to consume only, or overwhelmingly, media that broadly aligns with our own political viewpoints: that is, it is wrong to be politically “partisan” in our decisions about what media to consume. We are obligated to consume media that aligns with political viewpoints other than our own – to “diversify our sources”. This is so even if our own views are, as a matter of fact, substantively correct.

Response-dependence theories have historically been very popular in aesthetics, and aesthetic response-dependence has motivated response-dependence in ethics. This chapter closely examines the prospects for such theories. It breaks this category down into dispositional and fittingness strands of response-dependence, corresponding to descriptive and normative ideal observer theories. It argues that the latter have advantages over the former but are not themselves without issue. Special attention is paid to the relationship between hedonism and response-dependence. The chapter also introduces two aesthetic properties that (...) lead to wrong kinds of reasons problems for aesthetic response-dependence: insightfulness and the capacity to change one’s perspective. These properties do not have obvious parallels in the ethical domain, and so present an obstacle for response-dependence even in aesthetics. The chapter ends by examining replies on behalf of the response-dependence theorist, ultimately suggesting that a restricted form of response-dependence is the most promising way forward for fans of such theories. (shrink)

Recent years have seen a surge in online collaboration between experts and amateurs on scientific research. In this article, we analyse the epistemological implications of these crowdsourced projects, with a focus on Zooniverse, the world’s largest citizen science web portal. We use quantitative methods to evaluate the platform’s success in producing large volumes of observation statements and high impact scientific discoveries relative to more conventional means of data processing. Through empirical evidence, Bayesian reasoning, and conceptual analysis, we show how information (...) and communication technologies enhance the reliability, scalability, and connectivity of crowdsourced e-research, giving online citizen science projects powerful epistemic advantages over more traditional modes of scientific investigation. These results highlight the essential role played by technologically mediated social interaction in contemporary knowledge production. We conclude by calling for an explicitly sociotechnical turn in the philosophy of science that combines insights from statistics and logic to analyse the latest developments in scientific research. (shrink)

This paper examines the view that desires are beliefs about normative reasons for action. It describes the view, and briefly sketches three arguments for it. But the focus of the paper is defending the view from objections. The paper argues that the view is consistent with the distinction between the direction of fit of beliefs and desires, that it is consistent with the existence of appetites such as hunger, that it can account for counterexamples that aim to show that beliefs (...) about reasons are not sufficient for desire, such as weakness of will, and that it can account for counterexamples that aim to show that beliefs about reasons are not necessary for desire, such as addiction. The paper also shows how it is superior to the view that desires are appearances of the good. (shrink)

This paper explores the implications of empirical theories of migration for normative accounts of migration and distributive justice. It examines neo-classical economics, world-systems theory, dual labor market theory, and feminist approaches to migration and contends that neo-classical economic theory in isolation provides an inadequate understanding of migration. Other theories provide a fuller account of how national and global economic, political, and social institutions cause and shape migration flows by actively affecting people's opportunity sets in source countries and by admitting people (...) according to social categories such as class and gender. These empirical theories reveal the causal impact of institutions regulating migration and clarify moral obligations frequently overlooked by normative theorists. (shrink)

The political theory of migration has largely occurred within a paradigm of methodological nationalism and this has led to the neglect of morally salient agents and causes. This article draws on research from the social sciences on the transnationalism, globalization and migration systems theory to show how methodological nationalist assumptions have affected the views of political theorists on membership, culture and distributive justice. In particular, it is contended that methodological nationalism has prevented political theorists of migration from addressing the roles (...) of non-state agents and of transnational economic, social and political structures. These agents and structures contribute to the asymmetrical distribution of goods and opportunities and thus have important implications for debates about migration and distributive justice. (shrink)

In the summer of 1514, Desiderius Erasmus (c.1467-1536) was beginning to establish his name as the leading humanist scholar of his age, when he was recalled to his monastery in his native Holland.

This paper sets out a novel response to the ‘screening off problem’ for naïve realism. The aim is to resist the claim (which many naïve realists accept) that the kind of experience involved in hallucinating also occurs during perception, by arguing that there are causal constraints that must be met if an hallucinatory experience is to occur that are never met in perceptual cases. Notably, given this response, it turns out that, contra current orthodoxy, naïve realists need not adopt any (...) particular view about the psychological nature of hallucinatory experience to handle the screening off problem. Consequently, room opens up for naïve realists to endorse whatever theory of hallucinatory experience seems to best capture the distinctive nature of such episodes. (shrink)