The doyen of living English philosophers, by these reflections, took hold of and changed the outlook of a good many other philosophers, if not quite enough. He did so, essentially, by assuming that talk of freedom and responsibility is talk not of facts or truths, in a certain sense, but of our attitudes. His more explicit concern was to look again at the question of whether determinism and freedom are consistent with one another -- by shifting attention to certain personal (...) rather than moral attitudes, first of all gratitude and resentment. In the end, he arrived at a kind of Compatibilist or, as he says, Optimist conclusion. That is no doubt a recommendation but not the largest recommendation of this splendidly rich piece of philosophy. (shrink)

In information societies, operations, decisions and choices previously left to humans are increasingly delegated to algorithms, which may advise, if not decide, about how data should be interpreted and what actions should be taken as a result. More and more often, algorithms mediate social processes, business transactions, governmental decisions, and how we perceive, understand, and interact among ourselves and with the environment. Gaps between the design and operation of algorithms and our understanding of their ethical implications can have severe consequences (...) affecting individuals as well as groups and whole societies. This paper makes three contributions to clarify the ethical importance of algorithmic mediation. It provides a prescriptive map to organise the debate. It reviews the current discussion of ethical aspects of algorithms. And it assesses the available literature in order to identify areas requiring further work to develop the ethics of algorithms. (shrink)

1. Presumably the point of, say, inculcating a moral outlook lies in a concern with how people live. It may seem that the very idea of a moral outlook makes room for, and requires, the existence of moral theory, conceived as a discipline which seeks to formulate acceptable principles of conduct. It is then natural to think of ethics as a branch of philosophy related to moral theory, so conceived, rather as the philosophy of science is related to science. On (...) this view, the primary topic of ethics is the concept of right conduct, and the nature and justification of principles of behaviour. If there is a place for an interest in the concept of virtue, it is a secondary place. Virtue is a disposition to behave rightly; the nature of virtue is explained, as it were, from the outside in. (shrink)

This article reports the findings of AI4People, an Atomium—EISMD initiative designed to lay the foundations for a “Good AI Society”. We introduce the core opportunities and risks of AI for society; present a synthesis of five ethical principles that should undergird its development and adoption; and offer 20 concrete recommendations—to assess, to develop, to incentivise, and to support good AI—which in some cases may be undertaken directly by national or supranational policy makers, while in others may be led by other (...) stakeholders. If adopted, these recommendations would serve as a firm foundation for the establishment of a Good AI Society. (shrink)

This article considers the issue of opacity as a problem for socially consequential mechanisms of classification and ranking, such as spam filters, credit card fraud detection, search engines, news trends, market segmentation and advertising, insurance or loan qualification, and credit scoring. These mechanisms of classification all frequently rely on computational algorithms, and in many cases on machine learning algorithms to do this work. In this article, I draw a distinction between three forms of opacity: opacity as intentional corporate or state (...) secrecy, opacity as technical illiteracy, and an opacity that arises from the characteristics of machine learning algorithms and the scale required to apply them usefully. The analysis in this article gets inside the algorithms themselves. I cite existing literatures in computer science, known industry practices, and do some testing and manipulation of code as a form of lightweight code audit. I argue that recognizing the distinct forms of opacity that may be coming into play in a given application is a key to determining which of a variety of technical and non-technical solutions could help to prevent harm. (shrink)

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...

Inequality is widely regarded as morally objectionable: T. M. Scanlon investigates why it matters to us. He considers the nature and importance of equality of opportunity, whether the pursuit of greater equality involves objectionable interference with individual liberty, and whether the rich can be said to deserve their greater rewards.

The United States Army’s Future Combat Systems Project, which aims to manufacture a “robot army” to be ready for deployment by 2012, is only the latest and most dramatic example of military interest in the use of artificially intelligent systems in modern warfare. This paper considers the ethics of a decision to send artificially intelligent robots into war, by asking who we should hold responsible when an autonomous weapon system is involved in an atrocity of the sort that would normally (...) be described as a war crime. A number of possible loci of responsibility for robot war crimes are canvassed; the persons who designed or programmed the system, the commanding officer who ordered its use, the machine itself. I argue that in fact none of these are ultimately satisfactory. Yet it is a necessary condition for fighting a just war, under the principle of jus in bellum, that someone can be justly held responsible for deaths that occur in the course of the war. As this condition cannot be met in relation to deaths caused by an autonomous weapon system it would therefore be unethical to deploy such systems in warfare. (shrink)

Although decision‐making algorithms are not new to medicine, the availability of vast stores of medical data, gains in computing power, and breakthroughs in machine learning are accelerating the pace of their development, expanding the range of questions they can address, and increasing their predictive power. In many cases, however, the most powerful machine learning techniques purchase diagnostic or predictive accuracy at the expense of our ability to access “the knowledge within the machine.” Without an explanation in terms of reasons or (...) a rationale for particular decisions in individual cases, some commentators regard ceding medical decision‐making to black box systems as contravening the profound moral responsibilities of clinicians. I argue, however, that opaque decisions are more common in medicine than critics realize. Moreover, as Aristotle noted over two millennia ago, when our knowledge of causal systems is incomplete and precarious—as it often is in medicine—the ability to explain how results are produced can be less important than the ability to produce such results and empirically verify their accuracy. (shrink)

We are sceptical of concerns over the opacity of algorithmic decision tools. While transparency and explainability are certainly important desiderata in algorithmic governance, we worry that automated decision-making is being held to an unrealistically high standard, possibly owing to an unrealistically high estimate of the degree of transparency attainable from human decision-makers. In this paper, we review evidence demonstrating that much human decision-making is fraught with transparency problems, show in what respects AI fares little worse or better and argue that (...) at least some regulatory proposals for explainable AI could end up setting the bar higher than is necessary or indeed helpful. The demands of practical reason require the justification of action to be pitched at the level of practical reason. Decision tools that support or supplant practical reasoning should not be expected to aim higher than this. We cast this desideratum in terms of Daniel Dennett’s theory of the “intentional stance” and argue that since the justification of action for human purposes takes the form of intentional stance explanation, the justification of algorithmic decisions should take the same form. In practice, this means that the sorts of explanations for algorithmic decisions that are analogous to intentional stance explanations should be preferred over ones that aim at the architectural innards of a decision tool. (shrink)

Many AI systems that make important decisions are black boxes: how they function is opaque even to their developers. This is due to their high complexity and to the fact that they are trained rather than programmed. Efforts to alleviate the opacity of black box systems are typically discussed in terms of transparency, interpretability, and explainability. However, there is little agreement about what these key concepts mean, which makes it difficult to adjudicate the success or promise of opacity alleviation methods. (...) I argue for a unified account of these key concepts that treats the concept of understanding as fundamental. This allows resources from the philosophy of science and the epistemology of understanding to help guide opacity alleviation efforts. A first significant benefit of this understanding account is that it defuses one of the primary, in-principle objections to post hoc explainable AI (XAI) methods. This “rationalization objection” argues that XAI methods provide mere rationalizations rather than genuine explanations. This is because XAI methods involve using a separate “explanation” system to approximate the original black box system. These explanation systems function in a completely different way than the original system, yet XAI methods make inferences about the original system based on the behavior of the explanation system. I argue that, if we conceive of XAI methods as idealized scientific models, this rationalization worry is dissolved. Idealized scientific models misrepresent their target phenomena, yet are capable of providing significant and genuine understanding of their targets. (shrink)

Journal of Political Philosophy, Volume 30, Issue 2, Page 209-229, June 2022.

Deep learning is currently the most prominent and widely successful method in artificial intelligence. Despite having played an active role in earlier artificial intelligence and neural network research, philosophers have been largely silent on this technology so far. This is remarkable, given that deep learning neural networks have blown past predicted upper limits on artificial intelligence performance—recognizing complex objects in natural photographs and defeating world champions in strategy games as complex as Go and chess—yet there remains no universally accepted explanation (...) as to why they work so well. This article provides an introduction to these networks as well as an opinionated guidebook on the philosophical significance of their structure and achievements. It argues that deep learning neural networks differ importantly in their structure and mathematical properties from the shallower neural networks that were the subject of so much philosophical reflection in the 1980s and 1990s. The article then explores several different explanations for their success and ends by proposing three areas of inquiry that would benefit from future engagement by philosophers of mind and science. (shrink)

In artificial intelligence, recent research has demonstrated the remarkable potential of Deep Convolutional Neural Networks (DCNNs), which seem to exceed state-of-the-art performance in new domains weekly, especially on the sorts of very difficult perceptual discrimination tasks that skeptics thought would remain beyond the reach of artificial intelligence. However, it has proven difficult to explain why DCNNs perform so well. In philosophy of mind, empiricists have long suggested that complex cognition is based on information derived from sensory experience, often appealing to (...) a faculty of abstraction. Rationalists have frequently complained, however, that empiricists never adequately explained how this faculty of abstraction actually works. In this paper, I tie these two questions together, to the mutual benefit of both disciplines. I argue that the architectural features that distinguish DCNNs from earlier neural networks allow them to implement a form of hierarchical processing that I call “transformational abstraction”. Transformational abstraction iteratively converts sensory-based representations of category exemplars into new formats that are increasingly tolerant to “nuisance variation” in input. Reflecting upon the way that DCNNs leverage a combination of linear and non-linear processing to efficiently accomplish this feat allows us to understand how the brain is capable of bi-directional travel between exemplars and abstractions, addressing longstanding problems in empiricist philosophy of mind. I end by considering the prospects for future research on DCNNs, arguing that rather than simply implementing 80s connectionism with more brute-force computation, transformational abstraction counts as a qualitatively distinct form of processing ripe with philosophical and psychological significance, because it is significantly better suited to depict the generic mechanism responsible for this important kind of psychological processing in the brain. (shrink)

In a landmark essay, Andreas Matthias claimed that current developments in autonomous, artificially intelligent (AI) systems are creating a so-called responsibility gap, which is allegedly ever-widening and stands to undermine both the moral and legal frameworks of our society. But how severe is the threat posed by emerging technologies? In fact, a great number of authors have indicated that the fear is thoroughly instilled. The most pessimistic are calling for a drastic scaling-back or complete moratorium on AI systems, while the (...) optimists aim to show that the gap can be bridged nonetheless. Contrary to both camps, I argue against the prevailing assumption that there is a technology-based responsibility gap. I show how moral responsibility is a dynamic and flexible process, one that can effectively encompass emerging technological entities. (shrink)

Often machine learning programs inherit social patterns reflected in their training data without any directed effort by programmers to include such biases. Computer scientists call this algorithmic bias. This paper explores the relationship between machine bias and human cognitive bias. In it, I argue similarities between algorithmic and cognitive biases indicate a disconcerting sense in which sources of bias emerge out of seemingly innocuous patterns of information processing. The emergent nature of this bias obscures the existence of the bias itself, (...) making it difficult to identify, mitigate, or evaluate using standard resources in epistemology and ethics. I demonstrate these points in the case of mitigation techniques by presenting what I call ‘the Proxy Problem’. One reason biases resist revision is that they rely on proxy attributes, seemingly innocuous attributes that correlate with socially-sensitive attributes, serving as proxies for the socially-sensitive attributes themselves. I argue that in both human and algorithmic domains, this problem presents a common dilemma for mitigation: attempts to discourage reliance on proxy attributes risk a tradeoff with judgement accuracy. This problem, I contend, admits of no purely algorithmic solution. (shrink)

Previous research in Explainable Artificial Intelligence (XAI) suggests that a main aim of explainability approaches is to satisfy specific interests, goals, expectations, needs, and demands regarding artificial systems (we call these “stakeholders' desiderata”) in a variety of contexts. However, the literature on XAI is vast, spreads out across multiple largely disconnected disciplines, and it often remains unclear how explainability approaches are supposed to achieve the goal of satisfying stakeholders' desiderata. This paper discusses the main classes of stakeholders calling for explainability (...) of artificial systems and reviews their desiderata. We provide a model that explicitly spells out the main concepts and relations necessary to consider and investigate when evaluating, adjusting, choosing, and developing explainability approaches that aim to satisfy stakeholders' desiderata. This model can serve researchers from the variety of different disciplines involved in XAI as a common ground. It emphasizes where there is interdisciplinary potential in the evaluation and the development of explainability approaches. (shrink)

It is widely agreed that because animals feel pain we should not make them suffer gratuitously. Some ethical theories go even further: because of the capacities that they possess, animals have the right not to be harmed or killed. These views concern what not to do to animals, but we also face questions about when we should, and should not, assist animals that are hungry or distressed. Should we feed a starving stray kitten? And if so, does this commit us, (...) if we are to be consistent, to feeding wild animals during a hard winter? In this controversial book, Clare Palmer advances a theory that claims, with respect to assisting animals, that what is owed to one is not necessarily owed to all, even if animals share similar psychological capacities. Context, history, and relation can be critical ethical factors. If animals live independently in the wild, their fate is not any of our moral business. Yet if humans create dependent animals, or destroy their habitats, we may have a responsibility to assist them. Such arguments are familiar in human cases-we think that parents have special obligations to their children, for example, or that some groups owe reparations to others. Palmer develops such relational concerns in the context of wild animals, domesticated animals, and urban scavengers, arguing that different contexts can create different moral relationships. (shrink)

The essays in this volume represent the first steps by philosophers and artificial intelligence researchers toward explaining why it is necessary to add an ...

The usefulness of machine learning algorithms has led to their widespread adoption prior to the development of a conceptual framework for making sense of them. One common response to this situation is to say that machine learning suffers from a “black box problem.” That is, machine learning algorithms are “opaque” to human users, failing to be “interpretable” or “explicable” in terms that would render categorization procedures “understandable.” The purpose of this paper is to challenge the widespread agreement about the existence (...) and importance of a black box problem. The first section argues that “interpretability” and cognates lack precise meanings when applied to algorithms. This makes the concepts difficult to use when trying to solve the problems that have motivated the call for interpretability. Furthermore, since there is no adequate account of the concepts themselves, it is not possible to assess whether particular technical features supply formal definitions of those concepts. The second section argues that there are ways of being a responsible user of these algorithms that do not require interpretability. In many cases in which a black box problem is cited, interpretability is a means to a further end such as justification or non-discrimination. Since addressing these problems need not involve something that looks like an “interpretation” of an algorithm, the focus on interpretability artificially constrains the solution space by characterizing one possible solution as the problem itself. Where possible, discussion should be reformulated in terms of the ends of interpretability. (shrink)

This article examines the complaint that arbitrary algorithmic decisions wrong those whom they affect. It makes three contributions. First, it provides an analysis of what arbitrariness means in this context. Second, it argues that arbitrariness is not of moral concern except when special circumstances apply. However, when the same algorithm or different algorithms based on the same data are used in multiple contexts, a person may be arbitrarily excluded from a broad range of opportunities. The third contribution is to explain (...) why this systemic exclusion is of moral concern and to offer a solution to address it. (shrink)

Hardly anyone disputes that discrimination can be a grave moral wrong. Yet this consensus masks fundamental disagreements about what makes something discrimination, as well as precisely why acts of discrimination are wrong. Benjamin Eidelson develops systematic answers to those two questions. He claims that discrimination is a form of differential treatment distinguished by its special connection to the differential ascription of some property to different people, and goes on to argue that what makes some cases of discrimination intrinsically wrongful is (...) that they manifest an attitude of disrespect for the personhood of those who are disfavored. He endeavors to specify what this attitude consists in, and to demonstrate how attending to its character can help us to better understand the moral dimensions of different forms of wrongful discrimination. The book concludes with an extended discussion of racial profiling in law enforcement. (shrink)

We argue that explainable artificial intelligence (XAI), specifically reason-giving XAI, often constitutes the most suitable way of ensuring that someone can properly be held responsible for decisions that are based on the outputs of artificial intelligent (AI) systems. We first show that, to close moral responsibility gaps (Matthias 2004), often a human in the loop is needed who is directly responsible for particular AI-supported decisions. Second, we appeal to the epistemic condition on moral responsibility to argue that, in order to (...) be responsible for her decision, the human in the loop has to have an explanation available of the system’s recommendation. Reason explanations are especially well-suited to this end and we examine whether – and how – it might be possible to make such explanations fit with AI systems. We support our claims by focusing on a case of disagreement between human in the loop and AI system. (shrink)

The operations of deep networks are widely acknowledged to be inscrutable. The growing field of Explainable AI has emerged in direct response to this problem. However, owing to the nature of the opacity in question, XAI has been forced to prioritise interpretability at the expense of completeness, and even realism, so that its explanations are frequently interpretable without being underpinned by more comprehensive explanations faithful to the way a network computes its predictions. While this has been taken to be a (...) shortcoming of the field of XAI, I argue that it is broadly the right approach to the problem. (shrink)

Responsibility gaps concern the attribution of blame for harms caused by autonomous machines. The worry has been that, because they are artificial agents, it is impossible to attribute blame, even though doing so would be appropriate given the harms they cause. We argue that there are no responsibility gaps. The harms can be blameless. And if they are not, the blame that is appropriate is indirect and can be attributed to designers, engineers, software developers, manufacturers or regulators. The real problem (...) lies elsewhere: autonomous machines should be built so as to exhibit a level of risk that is morally acceptable. If they fall short of this standard, they exhibit what we call ‘a control gap.’ The causal control that autonomous machines have will then fall short of the guidance control they should emulate. (shrink)

Statistical criteria of fairness are formal measures of how an algorithm performs that aim to help us determine whether an algorithm would be fair to use in decision-making. In this paper, I introduce a new version of the criterion known as “Equalized Odds,” argue that it is a requirement of procedural fairness, and show that it is immune to a number of objections to the standard version.

We propose that the prevalent moral aversion to AWS is supported by a pair of compelling objections. First, we argue that even a sophisticated robot is not the kind of thing that is capable of replicating human moral judgment. This conclusion follows if human moral judgment is not codifiable, i.e., it cannot be captured by a list of rules. Moral judgment requires either the ability to engage in wide reflective equilibrium, the ability to perceive certain facts as moral considerations, moral (...) imagination, or the ability to have moral experiences with a particular phenomenological character. Robots cannot in principle possess these abilities, so robots cannot in principle replicate human moral judgment. If robots cannot in principle replicate human moral judgment then it is morally problematic to deploy AWS with that aim in mind. Second, we then argue that even if it is possible for a sufficiently sophisticated robot to make ‘moral decisions’ that are extensionally indistinguishable from (or better than) human moral decisions, these ‘decisions’ could not be made for the right reasons. This means that the ‘moral decisions’ made by AWS are bound to be morally deficient in at least one respect even if they are extensionally indistinguishable from human ones. Our objections to AWS support the prevalent aversion to the employment of AWS in war. They also enjoy several significant advantages over the most common objections to AWS in the literature. (shrink)

Statistical evidence—say, that 95% of your co-workers badmouth each other—can never render resenting your colleague appropriate, in the way that other evidence (say, the testimony of a reliable friend) can. The problem of statistical resentment is to explain why. We put the problem of statistical resentment in several wider contexts: The context of the problem of statistical evidence in legal theory; the epistemological context—with problems like the lottery paradox for knowledge, epistemic impurism and doxastic wrongdoing; and the context of a (...) wider set of examples of responses and attitudes that seem not to be appropriately groundable in statistical evidence. Regrettably, we do not come up with a fully general, fully adequate, fully unified account of all the phenomena discussed. But we give reasons to believe that no such account is forthcoming, and we sketch a somewhat messier account that may be the best that can be had here. (shrink)

If someone says, “You’ve stereotyped me,” we hear the statement as an accusation. One way to interpret the accusation is as follows: you haven’t seen or treated me as an individual. In this essay, I interpret and evaluate a theory of wrongful stereotyping inspired by this thought, which I call the failure-to-individualize theory of wrongful stereotyping. According to this theory, stereotyping is wrong if and only if it involves failing to treat persons as individuals. I argue that the theory—however one (...) interprets it—is inadequate. Either the theory will not reliably identify all cases of wrongful stereotyping or it will fail to adequately explain why they are wrong. I conclude that it does not follow that we must entirely jettison the objection that stereotyping fails to treat persons as individuals. What follows is only that the objection must play a more circumscribed role in a theory of when and why stereotyping is wrong. (shrink)

Algorithms influence every facet of modern life: criminal justice, education, housing, entertainment, elections, social media, news feeds, work… the list goes on. Delegating important decisions to machines, however, gives rise to deep moral concerns about responsibility, transparency, freedom, fairness, and democracy. Algorithms and Autonomy connects these concerns to the core human value of autonomy in the contexts of algorithmic teacher evaluation, risk assessment in criminal sentencing, predictive policing, background checks, news feeds, ride-sharing platforms, social media, and election interference. Using these (...) case studies, the authors provide a better understanding of machine fairness and algorithmic transparency. They explain why interventions in algorithmic systems are necessary to ensure that algorithms are not used to control citizens' participation in politics and undercut democracy. This title is also available as Open Access on Cambridge Core. (shrink)

In some cases, there appears to be an asymmetry in the evidential value of statistical and more individualized evidence. For example, while I may accept that Alex is guilty based on eyewitness testimony that is 80% likely to be accurate, it does not seem permissible to do so based on the fact that 80% of a group that Alex is a member of are guilty. In this paper I suggest that rather than reflecting a deep defect in statistical evidence, this (...) asymmetry might arise from a general constraint on rational inquiry. Plausibly the degree of evidential support needed to justify taking a proposition to be true depends on the stakes of error. While relying on statistical evidence plausibly raises the stakes by introducing new kinds of risk to members of the reference class, paradigmatically `individualized' evidence---evidence tracing back to A's voluntary behavior---can lower the stakes. The net result explains the apparent evidential asymmetry without positing a deep difference in the brute justificatory power of different types of evidence. (shrink)

Several influential characterizations of paternalism or its distinctive wrongness emphasize a belief or judgement that it typically involves—namely, 10 the judgement that the paternalized is likely to act irrationally, or some such. But it's not clear what about such a belief can be morally objectionable if it has the right epistemic credentials (if it is true, say, and is best supported by the evidence). In this paper, I elaborate on this point, placing it in the context of the relevant epistemological (...) discussions. I explain how 15 evidentialism is opposed to such thoughts; I show that possible ways of rejecting evidentialism (along lines analogous to those of pragmatic encroachment) won't work; and I sketch an account of the wrongness of paternalism that doesn't depend on any flaw in the belief about others' likely behaviour. (shrink)

Data-driven, decision-making technologies used in the justice system to inform decisions about bail, parole, and prison sentencing are biased against historically marginalized groups (Angwin, Larson, Mattu, & Kirchner 2016). But these technologies’ judgments—which reproduce patterns of wrongful discrimination embedded in the historical datasets that they are trained on—are well-evidenced. This presents a puzzle: how can we account for the wrong these judgments engender without also indicting morally permissible statistical inferences about persons? I motivate this puzzle and attempt an answer.

It is a philosophical truism that we must think of others as moral agents, not merely as causal or statistical objects. But why? I argue that this follows from the best resolution of an antinomy between our experience of morality as necessarily binding on the will and our knowledge that all moral beliefs originate in contingent histories. We can address this antinomy only by understanding moral deliberation via interpersonal relationships, which simultaneously vindicate and constrains morality’s bind on the will. This (...) means that moral agency is fundamentally social. I model an attitude toward our causal nature on sociologist Erving Goffman’s concept of ‘civil inattention’; our social practice of agency requires that we give minimal attention to the contingent origins of moral judgments in ourselves and others. Understood this way, seeing ourselves as moral agents requires avoiding appeal to causal aetiology to settle substantive moral disagreement. (shrink)

Traditionally, the manufacturer/operator of a machine is held (morally and legally) responsible for the consequences of its operation. Autonomous, learning machines, based on neural networks, genetic algorithms and agent architectures, create a new situation, where the manufacturer/operator of the machine is in principle not capable of predicting the future machine behaviour any more, and thus cannot be held morally responsible or liable for it. The society must decide between not using this kind of machine any more (which is not a (...) realistic option), or facing a responsibility gap, which cannot be bridged by traditional concepts of responsibility ascription. (shrink)