As arti® cial intelligence moves ever closer to the goal of producing fully autonomous agents, the question of how to design and implement an arti® cial moral agent (AMA) becomes increasingly pressing. Robots possessing autonomous capacities to do things that are useful to humans will also have the capacity to do things that are harmful to humans and other sentient beings. Theoretical challenges to developing arti® cial moral agents result both from controversies among ethicists about moral theory itself, and from (...) computational limits to the implementation of such theories. In this paper the ethical disputes are surveyed, the possibility of a `moral Turing Test ’ is considered and the computational di culties accompanying the diŒerent types of approach are assessed. Human-like performance, which is prone to include immoral actions, may not be acceptable in machines, but moral perfection may be computationally unattainable. The risks posed by autonomous machines ignorantly or deliberately harming people and other sentient beings are great. The development of machines with enough intelligence to assess the eŒects of their actions on sentient beings and act accordingly may ultimately be the most important task faced by the designers of arti® cially intelligent automata. (shrink)

In his famous 1950 paper where he presents what became the benchmark for success in artificial intelligence, Turing notes that "at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted" (Turing 1950, 442). Kurzweil (1990) suggests that Turing's prediction was correct, even if no machine has yet to pass the Turing Test. In the wake of the (...) computer revolution, research in artificial intelligence and cognitive science has pushed in the direction of interpreting "thinking" as some sort of computational process. On this understanding, thinking is something computers (in principle) and humans (in practice) can both do. It is difficult to say precisely when in history the meaning of the term "thinking" headed in this direction. Signs are already present in the mechanistic and mathematical tendencies of the early Modern period, and maybe even glimmers are apparent in the ancient Greeks themselves. But over the long haul, we somehow now consider "thinking" as separate from the categories of "thoughtfulness" (in the general sense of wondering about things), "insight" and "wisdom." Intelligent machines are all around us, and the world is populated with smart cars, smart phones and even smart (robotic) appliances. But, though my cell phone might be smart, I do not take that to mean that it is thoughtful, insightful or wise. So, what has become of these latter categories? They seem to be bygones left behind by scientific and computational conceptions of thinking and knowledge that no longer have much use for them. In 2000, Allen, Varner and Zinser addressed the possibility of a Moral Turing Test (MTT) to judge the success of an automated moral agent (AMA), a theme that is repeated in Wallach and Allen (2009).. (shrink)

In this paper I use philosophical accounts on the relationship between trust and knowledge in science to apprehend this relationship on the Web. I argue that trust and knowledge are fundamentally entangled in our epistemic practices. Yet despite this fundamental entanglement, we do not trust blindly. Instead we make use of knowledge to rationally place or withdraw trust. We use knowledge about the sources of epistemic content as well as general background knowledge to assess epistemic claims. Hence, although we may (...) have a default to trust, we remain and should remain epistemically vigilant; we look out and need to look out for signs of insincerity and dishonesty in our attempts to know. A fundamental requirement for such vigilance is transparency: in order to critically assess epistemic agents, content and processes, we need to be able to access and address them. On the Web, this request for transparency becomes particularly pressing if (a) trust is placed in unknown human epistemic agents and (b) if it is placed in non-human agents, such as algorithms. I give examples of the entanglement between knowledge and trust on the Web and draw conclusions about the forms of transparency needed in such systems to support epistemically vigilant behaviour, which empowers users to become responsible and accountable knowers. (shrink)

That the successful development of fully autonomous artificial moral agents (AMAs) is imminent is becoming the received view within artificial intelligence research and robotics. The discipline of Machines Ethics, whose mandate is to create such ethical robots, is consequently gaining momentum. Although it is often asked whether a given moral framework can be implemented into machines, it is never asked whether it should be. This paper articulates a pressing challenge for Machine Ethics: To identify an ethical framework that is both (...) implementable into machines and whose tenets permit the creation of such AMAs in the first place. Without consistency between ethics and engineering, the resulting AMAs would not be genuine ethical robots, and hence the discipline of Machine Ethics would be a failure in this regard. Here this challenge is articulated through a critical analysis of the development of Kantian AMAs, as one of the leading contenders for being the ethic that can be implemented into machines. In the end, however, the development of Kantian artificial moral machines is found to be anti-Kantian. The upshot of all this is that machine ethicists need to look elsewhere for an ethic to implement into their machines. (shrink)

Aristotle conceives of ethical theory as a field distinct from the theoretical sciences. Its methodology must match its subject matter—good action—and must respect the fact that in this field many generalizations hold only for the most part. We study ethics in order to improve our lives, and therefore its principal concern is the nature of human well-being. Aristotle follows Socrates and Plato in taking the virtues to be central to a well-lived life. Like Plato, he regards the ethical virtues (justice, (...) courage, temperance and so on) as complex rational, emotional and social skills. But he rejects Plato's idea that a training in the sciences and metaphysics is a necessary prerequisite for a full understanding of our good. What we need, in order to live well, is a proper appreciation of the way in which such goods as friendship, pleasure, virtue, honor and wealth fit together as a whole. In order to apply that general understanding to particular cases, we must acquire, through proper upbringing and habits, the ability to see, on each occasion, which course of action is best supported by reasons. Therefore practical wisdom, as he conceives it, cannot be acquired solely by learning general rules. We must also acquire, through practice, those deliberative, emotional, and social skills that enable us to put our general understanding of well-being into practice in ways that are suitable to each occasion. (shrink)

Artificial agents (AAs), particularly but not only those in Cyberspace, extend the class of entities that can be involved in moral situations. For they can be conceived of as moral patients (as entities that can be acted upon for good or evil) and also as moral agents (as entities that can perform actions, again for good or evil). In this paper, we clarify the concept of agent and go on to separate the concerns of morality and responsibility of agents (most (...) interestingly for us, of AAs). We conclude that there is substantial and important scope, particularly in Computer Ethics, for the concept of moral agent not necessarily exhibiting free will, mental states or responsibility. This complements the more traditional approach, common at least since Montaigne and Descartes, which considers whether or not (artificial) agents have mental states, feelings, emotions and so on. By focussing directly on mind-less morality we are able to avoid that question and also many of the concerns of Artificial Intelligence. A vital component in our approach is the Method of Abstraction for analysing the level of abstraction (LoA) at which an agent is considered to act. The LoA is determined by the way in which one chooses to describe, analyse and discuss a system and its context. The Method of Abstraction is explained in terms of an interface or set of features or observables at a given LoA. Agenthood, and in particular moral agenthood, depends on a LoA. Our guidelines for agenthood are: interactivity (response to stimulus by change of state), autonomy (ability to change state without stimulus) and adaptability (ability to change the transition rules by which state is changed) at a given LoA. Morality may be thought of as a threshold defined on the observables in the interface determining the LoA under consideration. An agent is morally good if its actions all respect that threshold; and it is morally evil if some action violates it. That view is particularly informative when the agent constitutes a software or digital system, and the observables are numerical. Finally we review the consequences for Computer Ethics of our approach. In conclusion, this approach facilitates the discussion of the morality of agents not only in Cyberspace but also in the biosphere, where animals can be considered moral agents without their having to display free will, emotions or mental states, and in social contexts, where systems like organizations can play the role of moral agents. The primary cost of this facility is the extension of the class of agents and moral agents to embrace AAs. (shrink)

Many industry leaders and academics from the field of machine ethics would have us believe that the inevitability of robots coming to have a larger role in our lives demands that robots be endowed with moral reasoning capabilities. Robots endowed in this way may be referred to as artificial moral agents. Reasons often given for developing AMAs are: the prevention of harm, the necessity for public trust, the prevention of immoral use, such machines are better moral reasoners than humans, and (...) building these machines would lead to a better understanding of human morality. Although some scholars have challenged the very initiative to develop AMAs, what is currently missing from the debate is a closer examination of the reasons offered by machine ethicists to justify the development of AMAs. This closer examination is especially needed because of the amount of funding currently being allocated to the development of AMAs coupled with the amount of attention researchers and industry leaders receive in the media for their efforts in this direction. The stakes in this debate are high because moral robots would make demands on society; answers to a host of pending questions about what counts as an AMA and whether they are morally responsible for their behavior or not. This paper shifts the burden of proof back to the machine ethicists demanding that they give good reasons to build AMAs. The paper argues that until this is done, the development of commercially available AMAs should not proceed further. (shrink)

The moral enhancement of human beings is a constant theme in the history of humanity. Today, faced with the threats of a new, globalised world, concern over this matter is more pressing. For this reason, the use of biotechnology to make human beings more moral has been considered. However, this approach is dangerous and very controversial. The purpose of this article is to argue that the use of another new technology, AI, would be preferable to achieve this goal. Whilst several (...) proposals have been made on how to use AI for moral enhancement, we present an alternative that we argue to be superior to other proposals that have been developed. (shrink)

Many authors have proposed constraining the behaviour of intelligent systems with ‘machine ethics’ to ensure positive social outcomes from the development of such systems. This paper critically analyses the prospects for machine ethics, identifying several inherent limitations. While machine ethics may increase the probability of ethical behaviour in some situations, it cannot guarantee it due to the nature of ethics, the computational limitations of computational agents and the complexity of the world. In addition, machine ethics, even if it were to (...) be ‘solved’ at a technical level, would be insufficient to ensure positive social outcomes from intelligent systems. (shrink)

Many industry leaders and academics from the field of machine ethics would have us believe that the inevitability of robots coming to have a larger role in our lives demands that robots be endowed with moral reasoning capabilities. Robots endowed in this way may be referred to as artificial moral agents. Reasons often given for developing AMAs are: the prevention of harm, the necessity for public trust, the prevention of immoral use, such machines are better moral reasoners than humans, and (...) building these machines would lead to a better understanding of human morality. Although some scholars have challenged the very initiative to develop AMAs, what is currently missing from the debate is a closer examination of the reasons offered by machine ethicists to justify the development of AMAs. This closer examination is especially needed because of the amount of funding currently being allocated to the development of AMAs coupled with the amount of attention researchers and industry leaders receive in the media for their efforts in this direction. The stakes in this debate are high because moral robots would make demands on society; answers to a host of pending questions about what counts as an AMA and whether they are morally responsible for their behavior or not. This paper shifts the burden of proof back to the machine ethicists demanding that they give good reasons to build AMAs. The paper argues that until this is done, the development of commercially available AMAs should not proceed further. (shrink)

The question of whether AI systems such as robots can or should be afforded moral agency or patiency is not one amenable either to discovery or simple reasoning, because we as societies constantly reconstruct our artefacts, including our ethical systems. Consequently, the place of AI systems in society is a matter of normative, not descriptive ethics. Here I start from a functionalist assumption, that ethics is the set of behaviour that maintains a society. This assumption allows me to exploit the (...) theoretical biology of sociality and autonomy to explain our moral intuitions. From this grounding I extend to consider possible ethics for maintaining either human- or of artefact-centred societies. I conclude that while constructing AI systems as either moral agents or patients is possible, neither is desirable. In particular, I argue that we are unlikely to construct a coherent ethics in which it it is ethical to afford AI moral subjectivity. We are therefore obliged not to build AI we are obliged to. (shrink)

This paper argues against the moral Turing test as a framework for evaluating the moral performance of autonomous systems. Though the term has been carefully introduced, considered, and cautioned about in previous discussions :251–261, 2000; Allen and Wallach 2009), it has lingered on as a touchstone for developing computational approaches to moral reasoning :98–109, 2015). While these efforts have not led to the detailed development of an MTT, they nonetheless retain the idea to discuss what kinds of action and reasoning (...) should be demanded of autonomous systems. We explore the flawed basis of an MTT in imitation, even one based on scenarios of morally accountable actions. MTT-based evaluations are vulnerable to deception, inadequate reasoning, and inferior moral performance vis a vis a system’s capabilities. We propose verification—which demands the design of transparent, accountable processes of reasoning that reliably prefigure the performance of autonomous systems—serves as a superior framework for both designer and system alike. As autonomous social robots in particular take on an increasing range of critical roles within society, we conclude that verification offers an essential, albeit challenging, moral measure of their design and performance. (shrink)

This paper explores the ambiguous impact of new information and communications technologies on the cultivation of moral skills in human beings. Just as twentieth century advances in machine automation resulted in the economic devaluation of practical knowledge and skillsets historically cultivated by machinists, artisans, and other highly trained workers , while also driving the cultivation of new skills in a variety of engineering and white collar occupations, ICTs are also recognized as potential causes of a complex pattern of economic deskilling, (...) reskilling, and upskilling. In this paper, I adapt the conceptual apparatus of sociological debates over economic deskilling to illuminate a different potential for technological deskilling/upskilling, namely the ability of ICTs to contribute to the moral deskilling of human users, a potential that exists alongside rich but currently underrealized possibilities for moral reskilling and/or upskilling. I flesh out this general hypothesis by means of examples involving automated weapons technology, new media practices, and social robotics. I conclude that since moral skills are essential prerequisites for the effective development of practical wisdom and virtuous character, and since market and cultural forces are not presently aligned to bring about the more salutary of the ambiguous potentials presented here, the future shape of these developments warrants our close attention—and perhaps active intervention. (shrink)

Overview of the main issues about value pluralism.

We describe a form of moral artificial intelligence that could be used to improve human moral decision-making. We call it the “artificial moral advisor”. The AMA would implement a quasi-relativistic version of the “ideal observer” famously described by Roderick Firth. We describe similarities and differences between the AMA and Firth’s ideal observer. Like Firth’s ideal observer, the AMA is disinterested, dispassionate, and consistent in its judgments. Unlike Firth’s observer, the AMA is non-absolutist, because it would take into account the human (...) agent’s own principles and values. We argue that the AMA would respect and indeed enhance individuals’ moral autonomy, help individuals achieve wide and a narrow reflective equilibrium, make up for the limitations of human moral psychology in a way that takes conservatives’ objections to human bioenhancement seriously, and implement the positive functions of intuitions and emotions in human morality without their downsides, such as biases and prejudices. (shrink)

Moral judgements are based on automatic processes. Moral judgements are based on reason. In this paper, I argue that both of these claims are true, and show how they can be reconciled. Neither the automaticity of moral judgement nor the post hoc nature of conscious moral reasoning pose a threat to rationalist models of moral cognition. The relation moral reasoning bears to our moral judgements is not primarily mediated by episodes of conscious reasoning, but by the acquisition, formation and maintenance (...) – in short: education – of our moral intuitions. (shrink)

This paper examines the ethical pitfalls and challenges that non-ethicists, such as researchers and programmers in the fields of computer science, artificial intelligence and robotics, face when building moral machines. Whether ethics is “computable” depends on how programmers understand ethics in the first place and on the adequacy of their understanding of the ethical problems and methodological challenges in these fields. Researchers and programmers face at least two types of problems due to their general lack of ethical knowledge or expertise. (...) The first type is so-called rookie mistakes, which could be addressed by providing these people with the necessary ethical knowledge. The second, more difficult methodological issue concerns areas of peer disagreement in ethics, where no easy solutions are currently available. This paper examines several existing approaches to highlight the ethical pitfalls and challenges involved. Familiarity with these and similar problems can help programmers to avoid pitfalls and build better moral machines. The paper concludes that ethical decisions regarding moral robots should be based on avoiding what is immoral in combination with a pluralistic ethical method of solving moral problems, rather than relying on a particular ethical approach, so as to avoid a normative bias. (shrink)

In his 1961 monograph, Totality and Infinity: An Essay on Exteriority , the late phenomenologist, Emmanuel Levinas, noted that “everyone will readily agree that it is of the highest importance to know whether we are not duped by morality” (1961/1969, p. 21). What follows thereafter is an extensive attempt to ground a quasi-Kantian existential ethics based on interpersonal, face to face, relations (Beavers 2001). That philosophy should invite such an attempt already signifies that we might be in trouble where ethics (...) are concerned. After all, when one looks back over history, it appears that we have not made much progress in this area. (shrink)