This paper looks at philosophical questions that arise in the context of AI alignment. It defends three propositions. First, normative and technical aspects of the AI alignment problem are interrelated, creating space for productive engagement between people working in both domains. Second, it is important to be clear about the goal of alignment. There are significant differences between AI that aligns with instructions, intentions, revealed preferences, ideal preferences, interests and values. A principle-based approach to AI alignment, which combines these elements (...) in a systematic way, has considerable advantages in this context. Third, the central challenge for theorists is not to identify ‘true’ moral principles for AI; rather, it is to identify fair principles for alignment that receive reflective endorsement despite widespread variation in people’s moral beliefs. The final part of the paper explores three ways in which fair principles for AI alignment could potentially be identified. (shrink)

The Doctrine of Double Effect (DDE) is an influential non-consequentialist principle positing a role for intention in affecting the moral permissibility of some actions. In particular, the DDE focuses on the intend/foresee distinction, the core claim being that it is sometimes permissible to bring about as a foreseen but unintended side-effect of one’s action some harm it would have been impermissible to aim at as a means or as an end, all else being equal. This article explores the meaning and (...) application of the DDE along with current debates over the nature of the intend/foresee distinction and its moral significance. How is the line between intended effects and merely foreseen but unintended side-effects to be drawn, and how are problem cases best handled? What is an appropriate methodology for debating the tenability of the DDE? How might the DDE interact with other non-consequentialist principles, and how might it be modified to capture other related factors (resulting perhaps in a Doctrine of Triple Effect) or to avoid certain problem cases? Does the DDE make permissibility turn on the actual intentions of particular agents, as critics such as Thomson and Scanlon assume, or is the role of intention more abstract? (shrink)

The practical, normative dimension of planning is a plausible source of the ‘family resemblances’ noted by a number of legal theorists between Scott Shapiro’s Planning Theory and natural law jurisprudence. Foremost among these resemblances is Shapiro’s contention that the law, necessarily, has a moral aim. The moral aim thesis is at first glance surprising given Shapiro’s intention to defend exclusive legal positivism and unequivocal rejection of what he takes to be the core commitments of natural law theory. Shapiro’s claim, however, (...) is that although the law necessarily has a moral aim, this does not entail that it is successful in satisfying that aim. In order to assess this thesis, it is helpful to compare the Planning Theory with contemporary natural law approaches. Bringing Shapiro’s Planning Theory into dialogue with contemporary natural law theories can demonstrate some of the Planning Theory’s weaknesses as an alternative explanation of the ultimate grounds of the authoritativeness of legal norms. Some of these weaknesses, moreover, are instructive beyond the specific contours of the Planning Theory insofar as they generalise to other legal positivist approaches. In section one I consider Shapiro’s treatment of the so-called ‘Possibility Puzzle’ regarding the grounding relation between authoritative norms and legal authority. Shapiro’s denial of the capacity of earlier jurisprudential theories to resolve this puzzle overlooks what is – I suggest – a plausible solution developed by John Finnis on the basis of Joseph Raz’s theory of practical reason and norms. Section two then demonstrates why Shapiro’s attempt to combine a robust construal of the social facts thesis with a commitment to the thesis that law necessarily has a moral aim is ultimately unsuccessful. (shrink)

Desires matter. What are desires? Many believe that desire is a motivational state: desiring is being disposed to act. This conception aligns with the functionalist approach to desire and the standard account of desire's role in explaining action. According to a second influential approach, however, desire is first and foremost an evaluation: desiring is representing something as good. After all, we seem to desire things under the guise of the good. Which understanding of desire is more accurate? Is the guise (...) of the good even right to assume? Should we adopt an alternative picture that emphasizes desire's deontic nature? What do neuroscientific studies suggest? -/- Essays in the first section of the volume are devoted to these questions, and to the puzzle of desire's essence. In the second part of the volume, essays investigate some implications that the various conceptions of desire have on a number of fundamental issues. For example, why are inconsistent desires problematic? What is desire's role in practical deliberation? How do we know what we want? -/- This volume will contribute to the emergence of a fruitful debate on a neglected, albeit crucial, dimension of the mind. (shrink)

Should we cognitively alter animals in ways that might change their moral status? There has been some discussion of this question. For example, Chan (2009) and Chan and Harris (2001) consider whether we should radically enhance the cognitive capacities of animals, while Thompson (2008) and Shriver (2009) argue that we should in fact substantially disenhance some animals to protect them from suffering. More controversially, some have countenanced radical and possibly moral status-altering transformations of human persons. ... One question relevant to (...) all of these discussions is whether it is good for a being to have moral status; whether a being’s moral status contributes to its wellbeing or (as I take to be equivalent) has prudential value for that being. (shrink)

The concept of rationality, predominantly in the guise of rational choice theory, plays a key role in the social sciences. Yet, whilst rational choice theory is usually understood as part of positive political science, it is also widely employed within normative political theories. In this paper, we examine how allegedly positive rational choice arguments can find application within normative political theories. To this effect, we distinguish between two interpretations of rationality ascriptions, one empirical, the other normative. Since, as we demonstrate, (...) empirical readings of the rationality assumption cannot convincingly explain the role of rational choice arguments in normative theories, we argue that the rationality assumption should be given a normative interpretation. We conclude by considering what this result implies for the use of rational choice arguments in normative and positive political science. (shrink)

Desire has not been at the center of recent preoccupations in the philosophy of mind. Consequently, the literature settled into several dogmas. The first part of this introduction presents these dogmas and invites readers to scrutinize them. The main dogma is that desires are motivational states. This approach contrasts with the other dominant conception: desires are positive evaluations. But there are at least four other dogmas: the world should conform to our desires (world-to-mind direction of fit), desires involve a positive (...) evaluation (the “guise of the good”), we cannot desire what we think is actual (the “death of desire” principle), and, in neuroscience, the idea that the reward system is the key to understanding desire. The second part of the introduction summarizes the contributions to this volume. The hope is to contribute to the emergence of a fruitful debate on this neglected, albeit crucial, aspect of the mind. (shrink)

Desires matter. How are we to understand the intentionality of desire? According to the two classical views, desire is either a positive evaluation or a disposition to act: to desire a state is to positively evaluate it or to be disposed to act to realize it. This Ph.D. Dissertation examines these conceptions of desire and proposes a deontic alternative inspired by Meinong. On this view, desiring is representing a state of affairs as what ought to be or, if one prefers, (...) as what should be. Desire involves a deontic manner of representing: a norm of the ought-to-be type features in desire’s intentional mode, as opposed to content. The dissertation is structured in three parts. In order to defend this conception, I formulate three main desiderata for a promising theory of the intentionality of desire in the introduction (§0). The first concerns desire’s direction of fit, i.e. the intuition that the world should conform to our desires. The second concerns the death of desire principle, i.e. the intuition that one cannot desire what one represents as actual. The last pertains to desire’s role in psychological explanations, i.e. the intuition that desires can explain some mental states and be explained by other mental states. The first part examines the main conceptions of desire in light of these desiderata. I argue that the classical pictures of desire do not adequately meet our desiderata. The first chapter is devoted to the evaluative conception (§1), while the second examines the motivational approach (§2). Following these criticisms, I then present the deontic view of desire (§3). In the second part, I defend this conception with the help of three arguments. The main idea is that appealing to norms of the ought-to-be type can satisfy our chief desiderata: the world should conform to norms (world-to-mind direction of fit, §4), norms are grounded on values and in turn ground obligations (explanation, §5), and norms are about non-actual states of affairs (death of desire principle, §6). In the last part, I develop the deontic view to draw a cartography of the various types of desire. Some desires are correct, while others are inappropriate. This distinction is explained by the deontic conception, as it matches that between states of affairs that ought to obtain and states that should not obtain (§7). Two study cases are examined: caprice and the impermissibility of desire aggregation. Intuitively, hopes, wishes, or urges are types of desire. The next chapter presents a typology inspired by the deontic view and the type of norms there are (§8). The last chapter discusses the main objections to the deontic approach (§9). In conclusion, I show the relevance of the deontic view for several debates in philosophy of mind and ethics. Desires are crucial because they are the ‘eye’ of what should be. (shrink)

Against non-analytic naturalism and quietist realism, I defend a robust form of non-naturalism. The argument proceeds as follows: In the face of extensional underdetermination, quietist realism cannot non-question-beggingly respond to alternative accounts that offer formally identical but substantively different interpretations of what reasons are. They face what we might call the reasons appropriation problem. In light of this problem, quietists ought to abandon their view in favor of robust realism. By permitting substantive metaphysical claims we can then argue, based on (...) reasonhood being a relation, that reasonhood is an abstract universal. If reasonhood is an abstract universal, then we can non-question-beggingly assert that the counting in favor of relation is a general kind or genus. This poses a dilemma for non-analytic naturalists: either they need to give the same reductive analysis for epistemic and practical reasons, or they can give a different analyses by treating epistemic and practical reasons as a species of the larger genus. However, the former looks extensionally implausible epistemic reasons are not desire-based and the latter entails that the reduction, via a violation of irreflexivity, fails to ground reasonhood. Naturalistic reductions of the normative, accordingly, face a damning dilemma. We should, in the face of this argument, thus accept a robust form of non-naturalist realism. (shrink)

This essay contends that the constitutive elements of well-being are plural, partly objective, and separable. The essay argues that these elements are pleasure, friendship, significant achievement, important knowledge, and autonomy, but not either the appreciation of beauty or the living of a morally good life. The essay goes on to attack the view that elements of well-being must be combined in order for well-being to be enhanced. The final section argues against the view that, because anything important to say about (...) well-being could be reduced to assertions about these separable elements, the concept of well-being or personal good is ultimately unimportant. (shrink)