I characterize two different senses in which a formal theory of location can be said to be neutral: neutrality with respect to some locational principles, and neutrality with respect to some controversial cases that have been presented as counterexamples to some principles. Theories neutral in the second sense are often weaker than theories neutral in the first. To illustrate this, I consider some popular principles and some of the counterexamples that have been used to argue against them in the literature, including inexact objects, cases of Nested Multilocation, and failures of Additivity. I then show that no currently existing theory of location built on just one primitive is able to account for all the cases that I have discussed. Lastly, I present a theory of location based on just one primitive which can account for all the discussed cases.
Performatives are declarative sentences where their use by a speaker enables the speaker to perform an additional action. This paper argues that performativity requires knowledge. The speaker must know what they say to perform the additional action. The significance of this conclusion is then discussed.
Metaethicists have often claimed that expressivism either avoids or elegantly answers explanatory challenges about the metaphysics and epistemology of ethics in ways that “realist” metaethical views cannot. We argue that currently influential “conciliatory” or “quasi-realist” versions of expressivism lack distinctive means of either avoiding or answering two of the most important explanatory challenges in metaethics: the challenge to explain the supervenience of the ethical, and the challenge to explain the reliability of our ethical judgments. We offer a general framework which explains conciliatory expressivism’s inability to address these challenges. First, there are just two fundamental ways of addressing the sorts of explanatory challenges we pose: to undercut them (by showing that the relevant explanatory demand does not arise), or to answer them (by providing explanations of the relevant modal correlations). Second, the hypotheses about language that are necessary to achieve an even modestly conciliatory form of expressivism entail that the conciliatory expressivist cannot undercut these challenges in a straightforward way. Third, however, the conciliatory expressivist’s characteristic hypotheses about language and thought provide them with no distinctive additional resources that could be used to answer the challenges. If we are correct, the upshot is that conciliatory expressivism is at least as poorly off with respect to these challenges as its realist foils.
It is a commonplace of the philosophy of language that Gottlob Frege’s pragmatics, understood as a theory of communication, engages with what is commonly referred to as “transmission model.” According to this model, successful communication requires the identity between the thought expressed by the speaker when uttering a sentence and the thought grasped by the hearer. Additionally, there is widespread agreement that indexicals and demonstratives present significant challenges to Frege’s theory of sense and reference. Philosophers of language and mind have attempted to address these issues by proposing various Fregean, neo-Fregean or directly non-Fregean approaches. The paper aims to provide insights into Frege’s theory of communication by discussing the specific problems that proper names, indexicals, and demonstratives pose to (both pure or impure versions of) the transmission model, as well as the role Frege might have envisaged for a logically perfect language. We conclude that, contrary to what might appear at first glance, Frege did not support the transmission model of communication.
This article advances a novel approach to the question of who should take responsibility when no one is to blame, and shows how it can be applied to the emerging problem of faultless harm caused by artificial intelligence (AI). As AI systems increase in complexity and scale, so too does the risk of harm that cannot be neatly traced back to human intent or negligence. This trajectory challenges conventional, blame-based responsibility practices, calling instead for an account that does not presuppose wrongdoing. I argue here that AI deployers bear a strict moral duty to take responsibility for harmful outcomes, even when the deployment is, all things considered, morally justified. This duty arises from what I call the ‘accidental relationship’, a distinctive moral relationship that emerges whenever an agent, here the AI deployer, non-negligently and non-intentionally inflicts harm on another party. Through this distinct agential involvement in the harm, the deployer incurs special duties to acknowledge and redress the victim’s loss. The present account does not conceptualize the duty to take responsibility as a matter of repairing a prior wrongdoing, but as a matter of affirming the victim’s equal moral standing within the accidental relationship.
Over the past few decades, findings from the empirical sciences of memory have demonstrated that the folk conception of this mental faculty, which views it as a preservative device, is inadequate due to the constructive nature of memory processes. This discrepancy between the empirically established nature of memory and the folk conception raises an intriguing question about the reasons for this mismatch. In this paper, I argue that, rather than treating the natural language concept of ‘remembering’ as aiming at encapsulating the nature of the actual mental faculty of memory, we should regard it as part of our folk epistemology toolkit, alongside perceptual concepts and the concept of ‘knowing.’ Specifically, I argue that ‘remember’ functions as a direct evidential, a term used to indicate the source of evidence for a given statement. In this capacity, it belongs to the same category of expressions that typical perception verbs do.
This paper argues that fairness-oriented interventions in AI ethics have failed to produce meaningful improvements in the carceral outcomes of Black men. I contend that this failure reveals that AI systems are not the significant causal drivers of these disparities. I develop the deflationary argument, a causal test based on interventionist reasoning: if a component of a causal system exerts real influence, then interventions on that component should yield measurable and morally significant changes in downstream outcomes. Applied to AI, this means that if predictive policing or risk assessment algorithms were substantial causes of adverse outcomes, then technical adjustments—such as data reweighting, fairness constraints, or threshold recalibration—should materially lessen the burdens borne by the most harmed demographic group. I claim they have not. I conclude that algorithmic bias operates downstream of more fundamental causal structures—namely, the racial caste system and carceral logics that organize American institutions—and that AI ethics must shift its methodological orientation from purely correlation-based prediction to predicting the effects of interventions and discovering the causal mechanisms that generate observed data distributions.
A popular reason for denying resultant moral luck has to do with its alleged unfairness. In this paper, I argue that although two existing objections against the unfairness argument are unsuccessful, it ultimately fails. First, I criticize the objection that the unfairness argument commits a category mistake. Second, I criticize the objection that because morality is unfair, the unfairness of resultant moral luck provides no evidence against its existence. Nevertheless, the unfairness argument fails because it cannot be fleshed out in a dialectically forceful way. The argument fails, in other words, because it cannot tip the scales in favor of denying resultant moral luck.
I offer a new position in debates about the Purity Principle, roughly, whether the fundamental truths/facts may have nonfundamental constituents. Purity is intuitively compelling and plays an important role in many debates. Nevertheless, there are also reasons to doubt Purity. In this paper I attempt to retain the grain of truth in the Purity Principle by proposing a different, less-demanding Neutrality Principle, roughly, that fundamental truths or facts turn only their fundamental constituents.
Soft-line compatibilism has come to occupy a privileged position in the debate over manipulation arguments against compatibilism. By identifying a responsibility-relevant difference between manipulated and ordinary agents, soft-liners promise to accommodate the intuition that manipulated agents are not responsible while preserving ordinary responsibility judgments. I argue that this promise cannot be kept. Soft-liners face a dilemma. On the one hand, if the conditions they identify as responsibility-undermining fully excuse manipulated agents, they commit themselves to a form of responsibility skepticism, since structurally similar forms of influence pervade ordinary agency as well. If, on the other hand, soft-liners maintain that manipulation merely mitigates rather than eliminates responsibility, they abandon the no responsibility intuition that motivates their view in the first place, and collapse into a hard-line position. Either way, soft-lining loses its dialectical advantage. The upshot is that the most promising compatibilist response to manipulation arguments must engage directly with the hard-line strategy, explaining why the no-responsibility intuition is ultimately to be resisted.
Causation is grounded in explanation; or so I argue here. According to the view which I propose, the fact that event c causes event e is grounded in two facts: the occurrence of c explains the occurrence of e, and c and e are wholly distinct. After formulating and explicating this view, I give three arguments for it. The first is based on a complete reduction of one relation to another which this view affords. The second is based on plausible claims about causal facts being identical to certain sorts of conjunctions. And the third is based on answers to explanatory challenges which this view supports.
On the margins of normative competence, matters of responsibility become ambiguous. Can a 6-year-old be considered responsible for stealing? Is a 14-year-old blameworthy for plagiarizing on a test? Parents, educators, and philosophers provide different and seemingly conflicting answers to these types of questions. The ages at which children can be considered legally culpable also vary widely between democratic states. One reason for this is that, for evolving agents, it is difficult to assess whether they are insufficiently able to do what is expected of them, or whether they are able but unwilling to do what is expected of them. This gives rise to what I will refer to as a ‘can’t-or-won’t dilemma’. The two options are to either exculpate the person on the basis of insufficient ability, or to consider them responsible for wrongdoing. But neither option seems entirely satisfactory or appropriate in the light of uncertainty about a person’s level of normative competence. Recent alternative approaches to responsibility aim to address this problem by providing accounts of responsibility which do not rely on capacity criteria. On these accounts so called ‘marginal agents’ are in some way considered responsible for what they do. I discuss these solutions and argue that they do not resolve or dissolve the dilemma. Instead, their real contribution lies in how they can alleviate the dilemma. This insight helps move the debate forward, and has implications for how we think about adults too.
Much philosophy presupposes a picture of reality as stratified into levels or layers. Yet a precise articulation of such a picture has been missing. It is tempting to think that we now have the tools to remedy this, notably due to ground-theoretic accounts of relative fundamentality. One might expect levels of facts, and their hierarchical arrangement, to be definable from relative fundamentality among facts. I show that such a project can at best partially succeed, by establishing certain limitative results. Moreover, I argue that the limitations should not be surprising, in light of an analogy between levels of reality and instants of time. Facts may exist at more than one level, just like persisting objects exist at more than one time. Finally, I develop a new positive proposal, on which levels are not made up of facts, but of what I call “fact realisations”—roughly, facts together with a complete genealogy of their grounds.
The phenomenology of heroic action seems to be at odds with commonsense morality: while the latter tells us that heroic acts of self-sacrifice are morally optional, heroes reliably believe that such acts are morally required of them. Do heroes systematically exhibit false moral beliefs, or does our commonsense commitment to the possibility of supererogation need revising? This paper argues, contrary to both supererogationists and anti-supererogationists, that we need not decide: as what we are morally required to do depends on how virtuous we are, heroes are right to believe that heroic actions are morally required of them, while ordinary folk morality is right that, for ordinary folk, these very same actions are morally optional. I arrive at this conclusion by supplementing the foremost contemporary account of supererogation with two naturally plausible claims about the virtue of selflessness.
I discuss a novel strategy for blocking Bradleyan regresses that utilizes grounding claims for claims of the form aRb that include themselves within the claimed grounds. Such grounding claims act as regress blockers in the sense that raising the question why aRb, given the claimed grounds that include the grounding claim itself, entails either outright inconsistency – if the grounding claim concerns full grounds for aRb – or pragmatic infelicity – if the grounding claim concerns only partial grounds for aRb.
This paper addresses a new skeptical challenge called “meta-skepticism.” While traditional skepticism is concerned with whether our beliefs are justified or amount to knowledge, meta-skepticism is concerned with whether the epistemic concepts we use to evaluate our beliefs are significant. The worry is that there are many knowledge-like and justification-like concepts, and no principled reason to prefer our epistemic concepts over rivals. Thus, meeting our epistemic standards might be insignificant. In response, I argue that our epistemic concepts are significant because of the role they play in identifying reliable believers, regulating inquiry, and coordinating information exchange. By comparing our concept of knowledge with alternatives, we can appreciate that our ordinary notion of knowledge is better suited to these roles than its rivals. I conclude that meta-skepticism mistakenly assumes that we must evaluate epistemic importance independently of the functions our concepts serve.
I argue in favour of a novel conception of ontological parsimony. According to this novel conception, ontological parsimony qua theoretical virtue should not be measured solely by the number of entities (or kinds of entities, or fundamental kinds of entities) posited, but also by the internal complexity of those entities. I argue for this claim via the familiar methodology of appealing to intuitions about particular cases and to observations about successful scientific methodology. I furthermore point out some implications of my argument for other debates, and suggest a method for rendering parsimony qua degree of internal complexity commensurable with traditional conceptions of parsimony.
Social interactions are often interpreted differently depending on how the individuals involved are categorized.This paper argues that the categories that guide interpretation are intersectional categories, such as “Black woman” or “White man.” Drawing on empirical work in categorization and vision science, I show that person perception typically relies on categories that are highly informative. Highly informative categories are those that guide interaction. I first argue that race or gender used alone are not particularly informative as they do not reliably predict interactions. I then show that intersectional categories are informative, as they activate distinct stereotypes that directly guide interpretation. In advancing this argument, I aim to initiate a philosophical conversation about the structure of social perception itself. To understand and address how social perception shapes our everyday lives, we must acknowledge the intersectionality of perception. Until we do, we will fail to capture the nuance of intergroup dynamics and the means by which they might be improved.
