In this paper, we argue that there is at least a pro tanto reason to favor the control account of the right to privacy over the access account of the right to privacy. This conclusion is of interest due to its relevance for contemporary discussions related to surveillance policies. We discuss several ways in which the two accounts of the right to privacy can be improved significantly by making minor adjustments to their respective definitions. We then test the improved versions (...) of the two accounts on a test case, to see which account best explains the violation that occurs in the case. The test turns out in favor of the control account. (shrink)

In the literature on global justice, there has been a lengthy debate about what the world’s rich owe to the world’s poor. Some have argued that rich individuals have positive duties of beneficence to help the poor, while others have argued that rich individuals only have negative duties not to harm them. A common objection to the former view is that once it is accepted that positive duties exist, fulfilling these duties will be overdemanding since rich individuals can almost always (...) help a little more. Some have tried to overcome this overdemandingness objection by setting cut-off points for how demanding morality is. In this article, we aim to show that it is problematic to be committed to the following propositions: (1) Positive duties to aid exist; (2) The overdemandingness objection is a serious challenge for anyone who accepts that positive duties to aid exist; and (3) Setting cut-off points for how demanding morality is constitutes a plausible way to overcome the overdemandingness objection. Showing that a commitment to (1), (2) and (3) is problematic is of interest given that several influential theorists are committed to this set of views. This set of views is simply a bad theoretical cocktail. (shrink)

If I decide to disclose information about myself, this act can undermine other people’s ability to effectively conceal information about themselves. One case in point involves genetic information: if I share ‘my’ genetic information with others, I thereby also reveal genetic information about my biological relatives. Such dependencies are well-known in the privacy literature and are often referred to as ‘privacy dependencies’. Some take the existence of privacy dependencies to generate a moral duty to sometimes avoid sharing information about oneself. (...) If true, we argue, then it is sometimes justified for others to impose harm on the person sharing the information to prevent them from doing so. This is a highly revisionary implication. Hence, one must either endorse a highly revisionary view on what one may do to protect one’s privacy, or one must reject the view that privacy dependencies can be used to justify a moral duty that constrains choices about sharing information about oneself. (shrink)

Often, when we share information about ourselves, we contribute to people learning personal things about others. This may happen because what we share about ourselves can be used to infer personal information about others. Such dependencies have become known as privacy dependencies in the literature. It is sometimes claimed that the scope of the right to privacy should be expanded in light of such dependencies. For example, some have argued that inferring information about others can violate their right to privacy. (...) Others have argued that sharing personal information about yourself that license such inferences can by itself violate the right to privacy. In this paper, we argue that the latter view should be rejected. (shrink)

In this paper, we argue that the All Affected Principle generates a puzzle when applied to carbon tax referendums. According to recent versions of the All Affected Principle, people should have a say in a democratic decision in positive proportion to how much the decision affects them. Plausibly, one way of being affected by a carbon tax referendum is to bear the economic burden of paying the tax. On this metric of affectedness, then, people who pollute a lot are ceteris (...) paribus more affected than people who don’t, because the former stands to pay more in carbon taxes than the latter. Consequently, people who pollute a lot receive large voting power, while people who pollute a little receive little voting power. This is a puzzling distribution of voting power for several reasons. First, it is instrumentally counterproductive if the goal is to rectify damages to the climate. Second, it introduces prudential incentives for everyone to pollute more to gain voting power. Third, the relative difference in voting power between people who pollute a lot and people who don’t is puzzling from a desert-based point of view, since the former often make a serious effort to avoid polluting, while the latter don’t. (shrink)

In this paper, we argue that the widespread practice in the United States of busing voters to the polling station on Election Day is an instance of paying people to vote. We defend a definition of what it means to pay people to vote, and on this definition, busing voters to the polling station is an instance of paying people to vote. Paying people to vote is illegal according to United States federal election law. However, the United States courts have (...) historically considered the practice of busing voters to the polling station legally permissible. The United States legal system, therefore, faces a dilemma: either the courts must change their interpretation of current federal election law such that busing voters to the polling station is a violation of federal election law, or federal election law must be changed so that at least some instances of paying people to vote are legally permissible. We argue that choosing either horn of the dilemma has a controversial implication for the United States legal system. (shrink)

In this paper, I defend what I call the ‘Inference Principle’. This principle holds that if an agent obtains some information legitimately, then the agent can make any inference she wants based on the information, without violating anyone’s right to privacy. This principle is interesting for at least three reasons. First, it constitutes a novel answer to the timely question of whether the widespread use of ‘data analytics’ to infer personal information about individuals is morally permissible. Second, it contradicts what (...) seems to be a common view of inferences’ ability to violate privacy rights. Third, it offers an account of the theoretically underdeveloped issue of what duties are engendered by the moral right to privacy with regard to inferred information. (shrink)

In this paper, I offer an indirect argument for the Access Theory of privacy. First, I develop a new version of the rival Control Theory that is immune to all the classic objections against it. Second, I show that this new version of the Control Theory collapses into the Access Theory. I call the new version the ‘Negative Control Account’. Roughly speaking, the classic Control Theory holds that you have privacy if, and only if, you can control whether other people (...) know personal information about you. Critics of the Control Theory often give counterexamples, where privacy is either not diminished even though the claimant has lost control, or where privacy is diminished even though the claimant is in control. I argue that none of these alleged counterexamples work against the Negative Control Account. However, this is not a victory for the control theorist, because the Negative Control Account collapses into the Access Theory. The paper thus adds to the recent trend in the literature of favoring the Access Theory over the Control Theory. (shrink)

Artificial Intelligence (AI) often outperforms human doctors in terms of decisional speed. For some diseases, the expected benefit of a fast but less accurate decision exceeds the benefit of a slow but more accurate one. In such cases, we argue, it is often justified to rely on a medical AI to maximise decision speed – even if the AI is less accurate than human doctors.

This paper concerns the double standard debate in the ethics of AI literature. This debate essentially revolves around the question of whether we should subject AI systems to different normative standards than humans. So far, the debate has centered around the desideratum of transparency. That is, the debate has focused on whether AI systems must be more transparent than humans in their decision-making processes in order for it to be morally permissible to use such systems. Some have argued that the (...) same standards of transparency should be applied across the board, for AI systems and humans alike. Others have argued that we should hold AI systems to higher standards than humans in terms of transparency. In this paper, we first point out that there are structurally similar double standard debates to be had about other desiderata besides transparency, such as predictive accuracy. Second, we argue that when we focus on predictive accuracy, there are at least two reasons for holding AI systems to a lower standard than humans. (shrink)

In this paper, I shall discuss what I call the Argument From Exploitation. This argument has as its conclusion that for-profit markets in personal information are morally impermissible. The main premise given for this conclusion is that markets in personal information involve exploitation of vulnerable people, and appertaining inequalities. I try to show that at least one of the premises of this argument is false. I then entertain an objection to my argument that holds that adding the option for vulnerable (...) people to sell their personal information is in itself harmful, even if these people would be better off if they took the option. I try to demonstrate why that objection is not successful. In the end, I conclude that the Argument From Exploitation is not sound, and that until proven otherwise, we should thus think that markets in personal information are morally permissible. (shrink)

In this paper, we argue that because of the advent of Artificial Intelligence, the secret ballot is now much less effective at protecting voters from voting related instances of social ostracism and social punishment. If one has access to vast amounts of data about specific electors, then it is possible, at least with respect to a significant subset of electors, to infer with high levels of accuracy how they voted in a past election. Since the accuracy levels of Artificial Intelligence (...) are so high, the practical consequences of someone inferring one’s vote are identical to the practical consequences of having one’s vote revealed directly under an open voting regime. Therefore, if one thinks that the secret ballot is at least partly justified because it protects electors against voting related social ostracism and social punishment, one should be morally troubled by how Artificial Intelligence today can be used to infer individual electors’ past voting behaviour. (shrink)

In this article, we show how it is possible to lawfully buy an election. The method we describe for buying an election is novel. The key things that make it possible to buy an election are the existence of public voter registration lists where one can see whether a given elector has voted in a particular election, and the existence of Big Data Analytics that with a high degree of accuracy can predict what a given elector will vote in an (...) upcoming election. Someone interested in buying an election can enter an employment contract with all, or some of, the opponent electors where these electors are paid to do a job that prevents them from voting. By purchasing access to the public voter registration lists, it is possible to verify ex-post whether the opponent electors that one has signed a contract with have abstained. In the last two sections of the article, we discuss several barriers that can undermine an attempt to buy an election in the manner we identify. (shrink)

Many seem to think that AI-induced responsibility gaps are morally bad and therefore ought to be avoided. We argue, by contrast, that there is at least a pro tanto reason to welcome responsibility gaps. The central reason is that it can be bad for people to be responsible for wrongdoing. This, we argue, gives us one reason to prefer automated decision-making over human decision-making, especially in contexts where the risks of wrongdoing are high. While we are not the first to (...) suggest that responsibility gaps should sometimes be welcomed, our argument is novel. Others have argued that responsibility gaps should sometimes be welcomed because they can reduce or eliminate the psychological burdens caused by tragic moral choice-situations. By contrast, our argument explains why responsibility gaps should sometimes be welcomed even in the absence of tragic moral choice-situations, and even in the absence of psychological burdens. (shrink)

In this paper, we defend what we call the ‘Hybrid View’ of privacy. According to this view, an individual has privacy if, and only if, no one else forms an epistemically warranted belief about the individual’s personal matters, nor perceives them. We contrast the Hybrid View with what seems to be the most common view of what it means to access someone’s personal matters, namely the Belief-Based View. We offer a range of examples that demonstrate why the Hybrid View is (...) more plausible than the Belief-Based View. Finally, we show how the Hybrid View generates a more plausible fit between the concept of privacy, and a (morally objectionable) violation of privacy. (shrink)

In this paper, we defend what we call the ‘Hybrid View’ of privacy. According to this view, an individual has privacy if, and only if, no one else forms an epistemically warranted belief about the individual’s personal matters, nor perceives them. We contrast the Hybrid View with what seems to be the most common view of what it means to access someone’s personal matters, namely the Belief-Based View. We offer a range of examples that demonstrate why the Hybrid View is (...) more plausible than the Belief-Based View. Finally, we show how the Hybrid View generates a more plausible fit between the concept of privacy, and the concept of a (morally objectionable) violation of privacy. (shrink)

The stakes associated with an algorithmic decision are often said to play a role in determining whether the decision engenders a right to an explanation. More specifically, “high stakes” decisions are often said to engender such a right to explanation whereas “low stakes” or “non-high” stakes decisions do not. While the overall gist of these ideas is clear enough, the details are lacking. In this paper, we aim to provide these details through a detailed investigation of what we will call (...) the “Simple Stakes Thesis”. The Simple Stakes Thesis, as it will turn out, is too simple. For even if the stakes associated with a specific one-off decision are low—and hence does not engender a right to an explanation—such decisions may nevertheless form part of a high stakes pattern or aggregate of decisions. In such cases, we argue, even a low stakes decision may engender a right to explanation. Not only does this show that the right to explanation is more demanding than so far recognized. It also shows that the stakes thesis is significantly harder to apply in practice. (shrink)

A new theory is taking hold in neuroscience. It is the theory that the brain is essentially a hypothesis-testing mechanism, one that attempts to minimise the error of its predictions about the sensory input it receives from the world. It is an attractive theory because powerful theoretical arguments support it, and yet it is at heart stunningly simple. Jakob Hohwy explains and explores this theory from the perspective of cognitive science and philosophy. The key argument throughout The Predictive Mind (...) is that the mechanism explains the rich, deep, and multifaceted character of our conscious perception. It also gives a unified account of how perception is sculpted by attention, and how it depends on action. The mind is revealed as having a fragile and indirect relation to the world. Though we are deeply in tune with the world we are also strangely distanced from it. (shrink)

Climate change mitigation has become a paradigm case both for externalities in general and for the game-theoretic model of the Tragedy of the Commons (ToC) in particular. This situation is worrying, as we have reasons to suspect that some models in the social sciences are apt to be performative to the extent that they can become self-fulfilling prophecies. Framing climate change mitigation as a hardly solvable coordination problem may force us into a worse situation, by changing real-world behaviour to fit (...) our model, rather than the other way around. But while this problem of the performativity of the ToC has been noted in a recent paper in this journal by Matthew Kopec, his proposed strategies for dealing with their self-fulfilling nature fall short of providing an adequate solution. Instead of relying on the idea that modelling assumptions are always strictly speaking false, this paper shows that the problem may be better framed as a problem of underdetermination between competing explanations. Our goal here is to provide a framework for choosing between this set of competing models that allows us to avoid a ‘Russian Roulette’-like situation in which we gamble with existential risk. (shrink)

A series of recent developments points towards auditing as a promising mechanism to bridge the gap between principles and practice in AI ethics. Building on ongoing discussions concerning ethics-based auditing, we offer three contributions. First, we argue that ethics-based auditing can improve the quality of decision making, increase user satisfaction, unlock growth potential, enable law-making, and relieve human suffering. Second, we highlight current best practices to support the design and implementation of ethics-based auditing: To be feasible and effective, ethics-based auditing (...) should take the form of a continuous and constructive process, approach ethical alignment from a system perspective, and be aligned with public policies and incentives for ethically desirable behaviour. Third, we identify and discuss the constraints associated with ethics-based auditing. Only by understanding and accounting for these constraints can ethics-based auditing facilitate ethical alignment of AI, while enabling society to reap the full economic and social benefits of automation. (shrink)

A series of recent developments points towards auditing as a promising mechanism to bridge the gap between principles and practice in AI ethics. Building on ongoing discussions concerning ethics-based auditing, we offer three contributions. First, we argue that ethics-based auditing can improve the quality of decision making, increase user satisfaction, unlock growth potential, enable law-making, and relieve human suffering. Second, we highlight current best practices to support the design and implementation of ethics-based auditing: To be feasible and effective, ethics-based auditing (...) should take the form of a continuous and constructive process, approach ethical alignment from a system perspective, and be aligned with public policies and incentives for ethically desirable behaviour. Third, we identify and discuss the constraints associated with ethics-based auditing. Only by understanding and accounting for these constraints can ethics-based auditing facilitate ethical alignment of AI, while enabling society to reap the full economic and social benefits of automation. (shrink)

Important decisions that impact humans lives, livelihoods, and the natural environment are increasingly being automated. Delegating tasks to so-called automated decision-making systems can improve efficiency and enable new solutions. However, these benefits are coupled with ethical challenges. For example, ADMS may produce discriminatory outcomes, violate individual privacy, and undermine human self-determination. New governance mechanisms are thus needed that help organisations design and deploy ADMS in ways that are ethical, while enabling society to reap the full economic and social benefits of (...) automation. In this article, we consider the feasibility and efficacy of ethics-based auditing as a governance mechanism that allows organisations to validate claims made about their ADMS. Building on previous work, we define EBA as a structured process whereby an entity’s present or past behaviour is assessed for consistency with relevant principles or norms. We then offer three contributions to the existing literature. First, we provide a theoretical explanation of how EBA can contribute to good governance by promoting procedural regularity and transparency. Second, we propose seven criteria for how to design and implement EBA procedures successfully. Third, we identify and discuss the conceptual, technical, social, economic, organisational, and institutional constraints associated with EBA. We conclude that EBA should be considered an integral component of multifaced approaches to managing the ethical risks posed by ADMS. (shrink)

Dr Thrane makes an original contribution to one of the central topics in syntax and semantics: the nature and mechanisms of reference in natural language. He makes a fundamental distinction between syntactic analyses that are internal to the structure of a language and analyses of the referential properties that connect a language with the 'outside world' - and therefore derive in some sense from common human capacities for perceptual discrimination. Dr Thrane argues that the failure to make this (...) distinction and to attend separately to both kinds of analysis has vitiated previous general accounts of linguistic structure. The book focuses particularly on pronouns and on the role of determiners, quantifiers and other components of the noun phrase. Most of the data come from the modern Germanic languages, especially English, but Dr Thrane considers also the structural peculiarities of 'classifier languages' like Vietnamese. The book will be important for students of English language as well as for general linguists. (shrink)

In lieu of an abstract, here is a brief excerpt of the content:Philosophy, Psychiatry, & Psychology 11.1 (2004) 65-70 [Access article in PDF] Top-Down and Bottom-Up in Delusion Formation Jakob Hohwy Keywords delusions, top-down, bottom-up, predictive coding Some delusions may arise as responses to unusual experiences (Davies et al. 2001; Maher 1974;). The implication is that delusion formation in some cases involves some kind of bottom-up mechanism—roughly, from perception to belief. Delusion formation may also involve some kind of top-down (...) mechanism. This could be in the shape of a patient's flawed background beliefs, or biases, which somehow modulate perception or ensure delusional interpretations of experience. There is little agreement about what kind of mechanism is primary in delusion formation, or, indeed, what the particular mechanisms and their neural substrates might be.In the evaluation of top-down and bottom-up mechanisms in delusion formation, a number of conceptual questions should be addressed (Davies and Coltheart 2000). For example, if there is an experiential component, why is a hypothesis with a delusional explanation of the unusual experience seen as relevant? Why do the subjects come to believe it tenaciously? How come some unusual experiences do not develop into delusions?A number of empirical findings also need to be taken into account: for example, the damage to the autonomic system in Capgras delusion (Ellis et al. 1997), the decreased ability for fast error correction in schizophrenia (Frith and Done 1989), modulation of activity in parietal cortex in delusions of alien control and other passivity experiences (Blakemore, Oakley, and Frith 2003), and also in schizophrenia, the increased sensitivity to self-produced stimuli, such as tickling oneself (Blakemore et al. 2000) and the sound of one's own voice (Ford et al. 2001).Theory choice can also be guided by philosophical doctrines about, for example, the nature of belief, the constitutive role of rationality, and intersubjectivity and the nature of judgements about the external world. It is natural, however, to let beliefs about philosophical doctrine be secondary to empirical findings and conceptual issues.The distinction between bottom-up and top-down mechanisms is particularly sharp in John Campbell's (2001) account of delusions and in Tim Bayne and Elisabeth Pacherie's (2004) discussion of Campbell. In this discussion, top-down approaches are associated with the general philosophical doctrine of rationalism, and bottom-up approaches with the philosophical doctrine of empiricism. Campbell's account is top-down because he posits delusional framework beliefs that arise directly as a result of brain damage or malfunction, and that work in some top-down fashion on the interpretation of experience. Bayne [End Page 65] and Pacherie convincingly defend empiricism against Campbell's rationalist account. Among their many other objections, they briefly puzzle about how a delusional framework belief (e.g., that one's spouse is an impostor) in a top-down fashion could cause the damage to the autonomic system seen in the Capgras and Cotard delusions (2004, 1). I think this is a very important objection because skin conductance data are part of the evidence for the bottom-up hypothesis concerning the Capgras delusion. We could add to this, for example, that it is mysterious how belief in a top-down fashion could cause people with schizophrenia to be worse at fast error correction than normal controls, to be better at tickling themselves than normal controls, and to have an increased response to the sound of their own voice.On the other hand, although the skin conductance response (SCR) evidence suggests a bottom-up approach to the Capgras delusion, it does not explain the character of the subjective experience that is hypothesized to give rise to the delusional belief: reduced SCR is not part of the experience itself (thanks to Ian Gold for discussion of this). Something more is required to give a full account of the type of experience that gives rise to the delusion. Further, as they stand, most bottom-up accounts are unable to account for why reality testing fails for beliefs based on these experiences.Neither rationalism nor empiricism on their own seem able to address the conceptual questions while paying... (shrink)

An exciting theory in neuroscience is that the brain is an organ for prediction error minimization. This theory is rapidly gaining influence and is set to dominate the science of mind and brain in the years to come. PEM has extreme explanatory ambition, and profound philosophical implications. Here, I assume the theory, briefly explain it, and then I argue that PEM implies that the brain is essentially self-evidencing. This means it is imperative to identify an evidentiary boundary between the brain (...) and its environment. This boundary defines the mind-world relation, opens the door to skepticism, and makes the mind transpire as more inferentially secluded and neurocentrically skull-bound than many would nowadays think. Therefore, PEM somewhat deflates contemporary hypotheses that cognition is extended, embodied and enactive; however, it can nevertheless accommodate the kinds of cases that fuel these hypotheses. (shrink)

Virtue ethics is traditionally a conservative project. It analyses the virtues that humanity has been relying on since antiquity. This conservatism unduly limits the potential of virtue ethics to contribute to moral progress. Instead, we should pay more attention to constructionist virtue ethics with the help of conceptual engineering. I will argue that revising and ameliorating the virtue concepts which a community uses directly and indirectly leads to a change of the virtues that exist in this community. By revising and (...) innovating virtue concepts, we can re-make and improve the virtues that we have. (shrink)

Berichte zur Wissenschaftsgeschichte, Volume 45, Issue 1-2, Page 87-111, June 2022.

The notion that the brain is a prediction error minimizer entails, via the notion of Markov blankets and self-evidencing, a form of global scepticism — an inability to rule out evil demon scenarios. This type of scepticism is viewed by some as a sign of a fatally flawed conception of mind and cognition. Here I discuss whether this scepticism is ameliorated by acknowledging the role of action in the most ambitious approach to prediction error minimization, namely under the free energy (...) principle. I argue that the scepticism remains but that the role of action in the free energy principle constrains the demon’s work. This yields new insights about the free energy principle, epistemology, and the place of mind in nature. (shrink)

Virtue ethics has many times been suggested as a promising recipe for the construction of artificial moral agents due to its emphasis on moral character and learning. However, given the complex nature of the theory, hardly any work has de facto attempted to implement the core tenets of virtue ethics in moral machines. The main goal of this paper is to demonstrate how virtue ethics can be taken all the way from theory to machine implementation. To achieve this goal, we (...) critically explore the possibilities and challenges for virtue ethics from a computational perspective. Drawing on previous conceptual and technical work, we outline a version of artificial virtue based on moral functionalism, connectionist bottom–up learning, and eudaimonic reward. We then describe how core features of the outlined theory can be interpreted in terms of functionality, which in turn informs the design of components necessary for virtuous cognition. Finally, we present a comprehensive framework for the technical development of artificial virtuous agents and discuss how they can be implemented in moral environments. (shrink)

Some monothematic types of delusions may arise because subjects have unusual experiences. The role of this experiential component in the pathogenesis of delusion is still not understood. Focussing on delusions of alien control, we outline a model for reality testing competence on unusual experiences. We propose that nascent delusions arise when there are local failures of reality testing performance, and that monothematic delusions arise as normal responses to these. In the course of this we address questions concerning the tenacity with (...) which delusions are maintained, their often bizarre content, the patients’ inability to dismiss them, and their often circumscribed character. (shrink)

The proposed European Artificial Intelligence Act (AIA) is the first attempt to elaborate a general legal framework for AI carried out by any major global economy. As such, the AIA is likely to become a point of reference in the larger discourse on how AI systems can (and should) be regulated. In this article, we describe and discuss the two primary enforcement mechanisms proposed in the AIA: the _conformity assessments_ that providers of high-risk AI systems are expected to conduct, and (...) the _post-market monitoring plans_ that providers must establish to document the performance of high-risk AI systems throughout their lifetimes. We argue that the AIA can be interpreted as a proposal to establish a Europe-wide ecosystem for conducting AI auditing, albeit in other words. Our analysis offers two main contributions. First, by describing the enforcement mechanisms included in the AIA in terminology borrowed from existing literature on AI auditing, we help providers of AI systems understand how they can prove adherence to the requirements set out in the AIA in practice. Second, by examining the AIA from an auditing perspective, we seek to provide transferable lessons from previous research about how to refine further the regulatory approach outlined in the AIA. We conclude by highlighting seven aspects of the AIA where amendments (or simply clarifications) would be helpful. These include, above all, the need to translate vague concepts into verifiable criteria and to strengthen the institutional safeguards concerning conformity assessments based on internal checks. (shrink)

I argue that virtue reliabilism and virtue responsibilism are complementary. They do not give competing accounts of epistemic virtue. Rather they explain the excellent functioning of different parts of our cognitive apparatus. Reliabilist virtue designates the excellent functioning of fast and context-specific Type 1 cognitive processes, while responsibilist virtue means an excellent functioning of effortful and reflective Type 2 cognitive processes. This account unifies reliabilist and responsibilist virtue theory. But the virtues are not unified by designating some epistemic norm that (...) both aim at. Instead, I unify them through their cognitive foundations. Because Type 1 and Type 2 cognition are complementary, reliabilist and responsibilist virtues are complementary. Thereby, this dual-process theory of epistemic virtue gives a naturalised account of virtues as well as an explanation of how reliabilism and responsibilism relate. This approach offers a solution for both the generality problem and the situationist challenge to virtue epistemology; additionally it preserves the epistemological autonomy of each virtue type. (shrink)

Virtue ethics has many times been suggested as a promising recipe for the construction of artificial moral agents due to its emphasis on moral character and learning. However, given the complex nature of the theory, hardly any work has de facto attempted to implement the core tenets of virtue ethics in moral machines. The main goal of this paper is to demonstrate how virtue ethics can be taken all the way from theory to machine implementation. To achieve this goal, we (...) critically explore the possibilities and challenges for virtue ethics from a computational perspective. Drawing on previous conceptual and technical work, we outline a version of artificial virtue based on moral functionalism, connectionist bottom–up learning, and eudaimonic reward. We then describe how core features of the outlined theory can be interpreted in terms of functionality, which in turn informs the design of components necessary for virtuous cognition. Finally, we present a comprehensive framework for the technical development of artificial virtuous agents and discuss how they can be implemented in moral environments. (shrink)

Recent advancements in artificial intelligence have fueled widespread academic discourse on the ethics of AI within and across a diverse set of disciplines. One notable subfield of AI ethics is machine ethics, which seeks to implement ethical considerations into AI systems. However, since different research efforts within machine ethics have discipline-specific concepts, practices, and goals, the resulting body of work is pestered with conflict and confusion as opposed to fruitful synergies. The aim of this paper is to explore ways to (...) alleviate these issues, both on a practical and theoretical level of analysis. First, we describe two approaches to machine ethics: the philosophical approach and the engineering approach and show how tensions between the two arise due to discipline specific practices and aims. Using the concept of disciplinary capture, we then discuss potential promises and pitfalls to cross-disciplinary collaboration. Drawing on recent work in philosophy of science, we finally describe how metacognitive scaffolds can be used to avoid epistemological obstacles and foster innovative collaboration in AI ethics in general and machine ethics in particular. (shrink)