Artificial Intelligence (AI) is increasingly adopted in society, creating numerous opportunities but at the same time posing ethical challenges. Many of these are familiar, such as issues of fairness, responsibility and privacy, but are presented in a new and challenging guise due to our limited ability to steer and predict the outputs of AI systems. This chapter first introduces these ethical challenges, stressing that overviews of values are a good starting point but frequently fail to suffice due to the context (...) sensitivity of ethical challenges. Frequently, additional (ethical) values emerge for specific applications, as e.g. challenges with fraud detection are very different from those around language technologies. Second, methods to tackle these challenges are discussed. Main ethical theories (virtue ethics, consequentialism, and deontology) are shown to provide a starting point, but often lack the details needed for actionable AI Ethics. Instead, mid-level philosophical theories coupled to design-approaches such as Design for Values, together with interdisciplinary working methods, offer the best way forwards. The chapter aims to show how these approaches can lead to an ethics of AI that is actionable and that can be proactively integrated in the design of AI systems. (shrink)

The crucial premise of the standard argument for two-boxing in Newcomb's problem, a causal dominance principle, is false. We present some counterexamples. We then offer a metaethical explanation for why the counterexamples arise. Our explanation reveals a new and superior argument for two-boxing, one that eschews the causal dominance principle in favor of a principle linking rational choice to guidance and actual value maximization.

Twentieth century philosophers introduced the distinction between “objective rightness” and “subjective rightness” to achieve two primary goals. The first goal is to reduce the paradoxical tension between our judgments of (i) what is best for an agent to do in light of the actual circumstances in which she acts and (ii) what is wisest for her to do in light of her mistaken or uncertain beliefs about her circumstances. The second goal is to provide moral guidance to an agent who (...) may be uncertain about the circumstances in which she acts, and hence is unable to use her standard moral principle directly in deciding what to do. This paper distinguishes two important senses of “moral guidance”; proposes criteria of adequacy for accounts of subjective rightness; canvasses existing definitions for “subjective rightness”; finds them all deficient; and proposes a new and more successful account. It argues that each comprehensive moral theory must include multiple principles of subjective rightness to address the epistemic situations of the full range of moral decision-makers, and shows that accounts of subjective rightness formulated in terms of what it would reasonable for the agent to believe cannot provide that guidance. -/- . (shrink)

This paper argues that non-identity actions are wrong because they _cause_ harm to people. While non-identity actions also typically benefit people, failure to act would similarly benefit someone, so considerations of benefit are ineligible to justify the harm. However, in some non-identity cases, failure to act would not benefit anyone: cases where one is choosing whether to procreate at all. These are the _hard_ non-identity cases. Not all "different-number" cases are hard. In some cases, we don't know whether acting would (...) result in more or fewer people; this paper argues that this _epistemic_ factor makes acting in these cases wrong. (shrink)

Deontologists have long been upbraided for lacking an account of justified decision- making under risk and uncertainty. One response is to develop a deontological decision theory—a set of necessary and sufficient conditions for an act’s being permissible given an agent’s imperfect information. In this article, I show that deontologists can make more use of regular decision theory than some might have thought, but that we must adapt decision theory to accommodate agent- centered options—permissions to favor or sacrifice our own interests, (...) when doing so is overall morally worse. Accommodating options requires more than just amend- ing the decision-theoretic ‘value function’. We must change the decision rule as well. (shrink)

How should deontologists concerned with the ethics of killing apply their moral theory when we don’t know all the facts relevant to the permissibility of our action? Though the stakes couldn’t be higher, and uncertainty is endemic where killing is concerned, few deontologists have an answer to this question. In this paper I canvass two possibilities: that we should apply a threshold standard, equivalent to the ‘beyond a reasonable doubt’ standard applied for criminal punishment; and that we should fit our (...) deontological ethical theory into the apparatus of decision theory. I show that the first approach faces insurmountable obstacles, while the second holds much more promise for deontologists than they might first have assumed. (shrink)

Decision theory is founded on the principle that we ought to take the action that has the maximum expected value from among actions we are in a position to take. But prior to the notion of expected value is the notion of the actual value of that action: roughly, a measure of the good outcomes you would in fact procure if you were to take it. Surprisingly many decision theories operate without an analysis of actual value. I offer a definition (...) of actual value, and show that a form of decision theory due to Stalnaker can be reformulated so as to be in line with the edict to maximize expected value. By contrast, I show that there is no quantity — given by my definition or otherwise — that plays the role of actual value in Jeffrey's decision theory. (shrink)

Consequentialists often assume rational monism: the thesis that options are always made rationally permissible by the maximization of the selfsame quantity. This essay argues that consequentialists should reject rational monism and instead accept rational pluralism: the thesis that, on different occasions, options are made rationally permissible by the maximization of different quantities. The essay then develops a systematic form of rational pluralism which, unlike its rivals, is capable of handling both the Newcomb problems that challenge evidential decision theory and the (...) unstable problems that challenge causal decision theory. (shrink)

In this paper I present a new argument for prospectivism: the view that, for a consequentialist, rightness depends on what is prospectively best rather than what would actually be best. Prospective bestness depends on the agent’s epistemic position, though exactly how that works is not straightforward. I clarify various possible versions of prospectivism, which differ in how far they go in relativizing to the agent’s limitations. My argument for prospectivism is an argument for moderately objective prospectivism, according to which the (...) right thing to do is what would make sense given reasonable beliefs, reasonable probability estimates and a reasonable understanding of value. My argument is an argument for this form of prospectivism over objectivism. Arguments about prospectivism and objectivism usually use an example with the following form: an agent has a choice between options, one she knows would be acceptable, while the other could either be catastrophic or very good. Objectivists argue that the right thing to do is what would in fact be best (though the agent cannot know which option that is) while prospectivists argue that the agent’s ignorance is relevant, and the right thing to do is to compromise. The question is how we should understand the underlying argument. It is not about action guidance. Moderately objective prospectivism is not action guiding, because an actual agent may not have access to reasonable beliefs, probability estimates and so on. Another common argument is that the objective notion is the primary one. I show that there are no good grounds for this claim. My argument uses the distinction between rightness and goodness to show that a consequentialist theory, that bases rightness on goodness, should take into account how much goodness is at stake. Crucially, potential losses as well as gains are relevant. So long as goodness, rather than rightness, is in the driving seat, we should not be ‘bestness fetishists’. As the name suggest, this would be an irrational privileging of the best option. This argument does not apply to pure deontology: a pure deontology does not use the notion of goodness at all, and so there is nothing to compromise with. If the agent does not know what is right, there is nothing further to say. I end by arguing against a recent strategy that aims to show that although objectivism is true (the right option is the best one), we should sometimes do what is wrong (i.e. what is prospectively best). I argue that in so far as this is correct it is simply prospectivism with awkward terminology. (shrink)

How ought you to evaluate your options if you're uncertain about what's fundamentally valuable? A prominent response is Expected Value Maximisation (EVM)—the view that under axiological uncertainty, an option is better than another if and only if it has the greater expected value across axiologies. But the expected value of an option depends on quantitative probability and value facts, and in particular on value comparisons across axiologies. We need to explain what it is for such facts to hold. Also, EVM (...) is by no means self-evident. We need an argument to defend that it’s true. This book introduces an axiomatic approach to answer these worries. It provides an explication of what EVM means by use of representation theorems: intertheoretic comparisons can be understood in terms of facts about which options are better than which, and mutatis mutandis for intratheoretic comparisons and axiological probabilities. And it provides a systematic argument to the effect that EVM is true: the theory can be vindicated through simple axioms. The result is a formally cogent and philosophically compelling extension of standard decision theory, and original take on the problem of axiological or normative uncertainty. (shrink)

Recent work in applied ethics has advanced a raft of arguments regarding individual responsibilities to address collective challenges like climate change or the welfare and environmental impacts of meat production. Frequently, such arguments suggest that individual actors have a responsibility to be more conscientious with their consumption decisions, that they can and should harness the power of the market to bring about a desired outcome. A common response to these arguments, and a challenge in particular to act-consequentialist reasoning, is that (...) it “makes no difference” if one takes conscious consumption action or not – that one is “causally impotent” to change an outcome. In this paper, I break causal impotence objections into three distinct lines of argument and present causal indeterminacy as a third, unexplored variation of much more common causal impotence lines. I suggest that the causal indeterminacy argument presents additional challenges to consequentialist moral theory because it acknowledges that individual actions can have an impact on outcome, but suggests instead that the outcome can neither be known nor secured by the action itself. (shrink)

Is evil a distinct moral concept? Or are evil actions just very wrong actions? Some philosophers have argued that evil is a distinct moral concept. These philosophers argue that evil is qualitatively distinct from ordinary wrongdoing. Other philosophers have suggested that evil is only quantitatively distinct from ordinary wrongdoing. On this view, evil is just very wrong. In this paper I argue that evil is qualitatively distinct from ordinary wrongdoing. The first part of the paper is critical. I argue that (...) Luke Russell’s attempt to show that evil is only quantitatively distinct from ordinary wrongdoing fails. Russell’s argument fails because it is based on an implausible criterion for determining whether two concepts are qualitatively distinct. I offer a more plausible criterion and argue that based on this criterion evil and wrongdoing are qualitatively distinct. To help make my case, I sketch a theory of evil which makes a genuinely qualitative distinction between evil and wrongdoing. I argue that we cannot characterize evil as just very wrong on plausible conceptions of evil and wrongdoing. I focus on act-consequentialist, Kantian, and contractarian conceptions of wrongdoing. (shrink)

This paper offers some refinements to a particular objection to act consequentialism, the “causal impotence” objection. According to proponents of the objection, when we find circumstances in which severe, unnecessary harms result entirely from voluntary acts, it seems as if we should be able to indict at least one act among those acts, but act consequentialism appears to lack the resources to offer this indictment. Our aim is to show is that the most promising response on behalf of act consequentialism, (...) the threshold argument, cannot offer a fully general prescription about what to do in cases of collective action. (shrink)

We argue that philosophers ought to distinguish epistemic decision theory and epistemology, in just the way ordinary decision theory is distinguished from ethics. Once one does this, the internalist arguments that motivate much of epistemic decision theory make sense, given specific interpretations of the formalism. Making this distinction also causes trouble for the principle called Propriety, which says, roughly, that the only acceptable epistemic utility functions make probabilistically coherent credence functions immodest. We cast doubt on this requirement, but then argue (...) that epistemic decision theorists should never have wanted such a strong principle in any case. (shrink)

The topic of this thesis is axiological uncertainty – the question of how you should evaluate your options if you are uncertain about which axiology is true. As an answer, I defend Expected Value Maximisation (EVM), the view that one option is better than another if and only if it has the greater expected value across axiologies. More precisely, I explore the axiomatic foundations of this view. I employ results from state-dependent utility theory, extend them in various ways and interpret (...) them accordingly, and thus provide axiomatisations of EVM as a theory of axiological uncertainty. (shrink)

Frank Jackson has put forward a famous thought experiment of a physician who has to decide on the correct treatment for her patient. Subjective consequentialism tells the physician to do what intuitively seems to be the right action, whereas objective consequentialism fails to guide the physician’s action. I suppose that objective consequentialists want to supplement their theory so that it guides the physician’s action towards what intuitively seems to be the right treatment. Since this treatment is wrong according to objective (...) consequentialism, objective consequentialists have to license it without calling it right. I consider eight strategies to spell out the idea of licensing the intuitively right treatment and argue that objective consequentialism is on the horns of what I call the licensing dilemma : Either the physician’s action is not guided towards the intuitively right treatment. Or the guidance towards the intuitively right treatment is ad hoc in some respect or the other. (shrink)

Monists, pluralists, and particularists disagree about the structure of the best explanation of the rightness (wrongness) of actions. In this paper I argue that the availability of good moral advice gives us reason to prefer particularist theories and pluralist theories to monist theories. First, I identify two distinct roles of moral theorizing—explaining the rightness (wrongness) of actions, and providing moral advice—and I explain how these two roles are related. Next, I explain what monists, pluralists, and particularists disagree about. Finally, I (...) argue that particularists and pluralists are better situated than monists to explain why it is a good idea to think before we act, and that this gives us reason to favor particularism and pluralism over monism. (shrink)

Cost-benefit analysis is commonly understood to be intimately connected with utilitarianism and incompatible with other moral theories, particularly those that focus on deontological concepts such as rights. We reject this claim and argue that cost-benefit analysis can take moral rights as well as other non-utilitarian moral considerations into account in a systematic manner. We discuss three ways of doing this, and claim that two of them (output filters and input filters) can account for a wide range of rights-based moral theories, (...) including the absolute notions of moral rights proposed by Hayek, Mayo, Nozick, and Shue. We also discuss whether the use of output filters and input filters can be generalized to cover other non-utilitarian theories, such as Kantian duty ethics and virtue ethics. (shrink)

W. D. Ross is commonly considered to be a generalist about prima facie duty but a particularist about absolute duty. That is, many philosophers hold that Ross accepts that there are true moral principles involving prima facie duty but denies that there are any true moral principles involving absolute duty. I agree with the former claim: Ross surely accepts prima facie moral principles. However, in this paper, I challenge the latter claim. Ross, I argue, is no more a particularist about (...) absolute duty than a utilitarian or a Kantian is. While this conclusion is interesting in its own right, it is also important, I argue, because it prevents us from overlooking Ross's criterion of moral obligation and because it may have implications on the broader debate between particularists and generalists. (shrink)

Consequentialist theories determine rightness solely based on real or expected consequences. Although such theories are popular, they often have difficulty with generalizing intuitions, which demand concern for questions like “What if everybody did that?” Rule consequentialism attempts to incorporate these intuitions by shifting the locus of evaluation from the consequences of acts to those of rules. However, detailed rule-consequentialist theories seem ad hoc or arbitrary compared to act consequentialist ones. We claim that generalizing can be better incorporated into consequentialism by (...) keeping the locus of evaluation on acts but adjusting the decision theory behind act selection. Specifically, we should adjust which types of dependencies the theory takes to be decision-relevant. Using this strategy, we formulate a new theory, generalized act consequentialism, which we argue is more compelling than rule consequentialism both in modeling the actual reasoning of generalizers and in delivering correct verdicts. (shrink)

Some moral theories, such as objective forms of consequentialism, seem to fail to be practically useful: they are of little to no help in trying to decide what to do. Even if we do not think this constitutes a fatal flaw in such theories, we may nonetheless agree that being practically useful does make a moral theory a better theory, or so some have suggested. In this paper, I assess whether the uncontroversial respect in which a moral theory can be (...) claimed to be better if it is practically useful can provide a ground worth taking into account for believing one theory rather than another. I argue that this is not the case. The upshot is that if there is a sound objection to theories such as objective consequentialism that is based on considerations of practical usefulness, the objection requires that it is established that the truth about what we morally ought to do cannot be epistemically inaccessible to us. The value of practical usefulness has no bearing on the issue. (shrink)

A general introduction to consequentialism.

Over the years, consequentialism has been subjected to numerous serious objections. Its adherents, however, have been remarkably successful in fending them off. As I argue in this paper, the reason why the case against consequentialism has not been more successful lies, at least partly, in the methodological approach that critics have commonly used. Their arguments have usually proceeded in two steps. First, a definition of consequentialism is given. Then, objections are put forward based on that definition. This procedure runs into (...) one of two problems. Substantive criticisms of consequentialism can only be formulated, if the posited definition is sufficiently concrete and narrow. In that case, however, consequentialists can defend themselves using a strategy that I call “interpretive divergence”. They can simply point out that the critic's definition does not accord with their understanding of consequentialism to which criticisms do not apply. If, on the other hand, an all-encompassing definition is used, it is so abstract that it is doubtful whether any substantive criticisms can be formulated. To escape this dilemma, I sketch a methodological approach which drops the assumption that consequentialism should be defined. It assumes, rather, that the term “consequentialism” should be interpreted as a Wittgensteinian family resemblance term. (shrink)

This paper faces the problem of how to determinate the framework in which, according to Aristotle, the deliberation and his results can be assessed: the decision and the consequent action. The problem emerges mainly because of what Aristotle calls “indetermination of deliberation,” since, in situations whose options about what to deliberate are blurred, it is difficult to determinate which is the framework that can be used to determinate if the decision resulting from the deliberation is accurate or not. To face (...) this problem, in the first place, we distinguish between technical and moral deliberation. The distinction is framed in terms of the ends because in technical deliberation the end is extrinsic to the action itself, and for this reason it is possible to offer different assessments of the actions achieved. In moral deliberation, the ends are only intrinsic to the action, and they are determined only in behalf of their connection with good deliberation. So, it is only possible to establish an adequate normative framework to assess each deliberation based upon a “good deliberation model”, established by practical wisdom, and extended also for the community from which the subject takes part. (shrink)

Some consequentialists argue that ordinary individuals are obligated to act in specific, concrete ways to address large-scale harms. For example, they argue that we should each refrain from meat-eating and avoid buying sweatshop-made clothing. The case they advance for such prescriptions can seem intuitive and compelling: by acting in those ways, a person might help prevent serious harms from being produced at little or no personal cost, and so one should act in those ways. But I argue that such reasoning (...) often relies on an overly simplistic assessment of the costs and benefits of those prescriptions, one that misconstrues or neglects important issues. Indeed, a closer look at those costs and benefits reveals just how little we often know about a number of real-world matters that bear on the expected consequences of the prescribed individual actions. Our predicament is one of radical uncertainty: we currently lack a sound basis for concluding that the relevant actions are more likely to do good than to backfire. I distinguish this empirically grounded objection from others, which are not convincing—including one based on the mere conceivability of the actions in question backfiring and another based on the supposition that such actions cannot make a difference in addressing the massive problems at issue. The upshot is that, at least for now, consequentialist arguments for many specific individual actions aimed at addressing large-scale harms are inconclusive. (shrink)

The objective of the paper is to explicate and critically appreciate two forms of consequentialism, namely objective and subjective consequentialism. Consequentialism is a substantive moral theory according to which moral value or good is to produce/promote best consequences (in a sense welfare); and morally right consists in acting so as to promote maximum good (in case of utilitarianism) or to promote best or most good. However, the paper considers important questions, replies to which give us two forms of consequentialism, namely (...) subjective and objective consequentialism, each having their unique features, nuances and limits. Moot problems are as follows: Whether consequentialism as a theory of good is to promote actual good, that is good as matter-of-fact or expected good, that is good expected by the moral agent; and whether as a theory of right, it urges to make decisions for the promotion of good based on some heuristic devices or subjective motives, commitments and so on. Objective consequentialism, as a theory of good prefers promotion of actual good than expected good, and as a theory of right decision-making urges the moral agents to rely on heuristic devices than subjective factors. The contest has been absorbing and fierce, each having so-called strong arguments. The paper reconsiders these contesting arguments and critically examines those to conclude in favour of objective consequentialism on important counts notwithstanding, the value of motives, motivations and commitments of moral agents. (shrink)

I claim that objective consequentialism faces a problem stemming from the existence in some situations of a plurality of chances relevant to the outcomes of an agent’s acts. I suggest that this phenomenon bears structural resemblance to the well-known Reference Class problem. I outline a few ways in which one could attempt to deal with the issue, suggesting that it is the higher-level chance that should be employed by OC.

The core idea of libertarianism, considered as a basic moral theory, is that people have certain negative rights and that those rights determine morally right action. Libertarianism is supposed to provide robust explanations to some of our intuitions, such as that it is wrong to steal, kill, rape or enslave other people. However, its exclusive focus on negative rights (i.e., rights to non-interference) makes it incapable of explaining some other intuitions, such as that the utterly rich should help the utterly (...) poor. Although libertarianism can explain why we should never do bad to others, it cannot explain why we should sometimes do good to others. For this reason, libertarianism is not satisfactory as it stands. A natural suggestion, therefore, is that we should either abandon libertarianism in favor of some of its better faring rivals, or revise the theory in order to get rid of the features that make it unsatisfactory. This paper proposes a new libertarian theory of morality: a theory that endorses a utilitarian proviso for use of external resources. I call this theory mid-libertarianism. The basic idea of mid-libertarianism is that individuals are free to do as they want as long as they do not violate the rights of others, given that they maximize utility whenever they use external resources. The paper is divided into four main sections. In this frst section, I introduce mid-libertarianism as a normative ethical theory. In the second main section, I put forward the key arguments for mid-libertarianism, which are, roughly, that it maintains the main explanatory powers of existing versions of libertarianism, while it avoids some of the most severe problems that these theories face. In the third section, I answer some potential objections to midlibertarianism. In the fourth section, I conclude that mid-libertarianism deserves to be taken seriously as a new contender in the normative ethics debate. (shrink)

The mismatch problem for consequentialism arises whenever the theory delivers mismatched verdicts between a group act and the individual acts that compose it. A natural thought is that moving to expected utility versions of consequentialism will solve this problem. I explain why the move to expected utility is not successful.

This thesis takes up a rights-based perspective to discuss a number of issues related to the problem of permissible harm. It appeals to a person’s capacity to shape her life in accordance with her own ideas of the good to explain why her death can be bad for her, and why each of us should have primary say over what may be done to her. The thesis begins with an investigation of the badness of death for the person who dies. (...) If death is bad for us, this helps explain the wrongness of killing. The thesis defends the deprivation account—i.e. the idea that death is bad for us when and because it deprives us of good life—against two Epicurean challenges. It adds that death is also bad when and because it thwarts our agency. Next, the thesis deals with the logic of our moral rights to non-interference. It proposes a conception of rights according to which the stringency of our rights derives from and is justified by the rational aspect of our human nature. It argues that this conception of moral rights solves the paradox of deontology. While our rights to non-interference are stringent, they are not absolute. The thesis considers two possible exceptions to the general rule that it is impermissible to harm an innocent person against her consent. First, using an actual case from WWII, it investigates the circumstances under which a government may expose some parts of its population to an increased risk of harm in order to decrease the risk to others. Second, it considers the permissibility of self-defence against an innocent threat. It argues that the potential victim of an innocent threat has a justice-based reason to treat her own interests as on a par with those of the threat. (shrink)

Cost-benefit analysis is commonly understood to be intimately connected with utilitarianism and incompatible with other moral theories, particularly those that focus on deontological concepts such as rights. We reject this claim and argue that cost-benefit analysis can take moral rights as well as other non-utilitarian moral considerations into account in a systematic manner. We discuss three ways of doing this, and claim that two of them (output filters and input filters) can account for a wide range of rights-based moral theories, (...) including the absolute notions of moral rights proposed by Hayek, Mayo, Nozick, and Shue. We also discuss whether the use of output filters and input filters can be generalized to cover other non-utilitarian theories, such as Kantian duty ethics and virtue ethics. (shrink)

Consequentialism is often charged with being self-defeating, for if a person attempts to apply it, she may quite predictably produce worse outcomes than if she applied some other moral theory. Many consequentialists have replied that this criticism rests on a false assumption, confusing consequentialism’s criterion of the rightness of an act with its position on decision procedures. Consequentialism, on this view, does not dictate that we should be always calculating which of the available acts leads to the most good, but (...) instead advises us to decide what to do in whichever manner it is that will lead to the best outcome. Whilst it is typically afforded only a small note in any text on consequentialism, this reply has deep implications for the practical application of consequentialism, perhaps entailing that a consequentialist should eschew calculation altogether. (shrink)

Utilitarianism would allow any degree of inequality whatsoever productive of the greatest happiness of the greatest number. But it does not guide political action, because determining what level of inequality would produce the greatest happiness of the greatest number is opaque due to well-known psychological coordination problems. Does Rawlsian liberalism, as is generally assumed, have some superiority to Utilitarianism in this regard? This paper argues not; for Rawls’s ‘difference principle’ would allow any degree of inequality whatsoever that best raises up (...) the worst off, and similar psychological coordination problems apply. It concludes that Rawlsian liberalism, designed to solve the problem that Utilitarianism will not give us stable rules and will counsel their violation, and giving us in the process a contract and rights within a semi-consequentialist framework, repeats (in the difference principle) that very problem of Utilitarianism. It fails substantively to guide the level of inequalities permitted in such a framework. (shrink)