We provide a general argument against value incomparability, based on a new style of impossibility result. In particular, we show that, against plausible background assumptions, value incomparability creates an incompatibility between two very plausible principles for ranking lotteries: a weak "negative dominance" principle (to the effect that Lottery 1 can be better than Lottery 2 only if some possible outcome of Lottery 1 is better than some possible outcome of Lottery 2) and a weak form of ex ante Pareto (to (...) the effect that, if Lottery 1 gives an unambiguously better prospect to some individuals than Lottery 2, and equally good prospects to everyone else, then Lottery 1 is better than Lottery 2). After spelling out our results, and the arguments based on them, we consider which principle the proponent of incomparability ought to reject. (shrink)

How should you evaluate your choices when you’re unsure what their outcomes will be? One popular answer is to rank your options in terms of their expected utilities. But what should you do when you think that the value of their respective outcomes might be incommensurable? In the face of incommensurable values, it no longer makes sense to speak of ranking your options according to expected utility. Are there any general principles to guide us when facing decisions of this kind? (...) If only! This chapter develops an impossibility result: it holds that there are a handful of independently plausible constraints that no such decision theory can jointly satisfy. The result, while depressing, can be used to helpfully classify extant approaches based on which of the constraints they violate. (shrink)

If updating with E has the same result across all epistemically possible worlds, then the agent has no uncertainty as to the behavior of the update, and we may call it a transparent update. If an agent is uncertain about the behavior of an update, we may call it opaque. In order to model the uncertainty an agent has about the result of an update, the same update must behave differently across different possible worlds. In this paper, I study (...) opaque updates using a simple system of dynamic epistemic logic suitably modified for that purpose. The paper highlights the connection between opaque updates and the dynamic-epistemic principles Perfect-Recall and No-Miracles. I argue that opaque updates are central to contemporary discussions in epistemology, in particular to externalist theories of knowledge and to the related problem of epistemic bootstrapping, or easy knowledge. Opaque updates allow us to explicitly investigate a dynamic form of uncertainty, using simple and precise logical tools. (shrink)

Moral options are permissions to do less than best, impartially speaking. In this paper, we investigate the challenge of reconciling moral options with the ideal of justifiability to each individual. We examine ex-post and ex-ante views of moral options and show how they might conflict with this ideal in single-choice and sequential-choice cases, respectively. We consider some ways of avoiding this conflict in sequential-choice cases, showing that they face significant problems.

This article takes up the question of whether legal rules are reasons for action. They are commonly regarded in this way, yet are legal rules reasons for action themselves (the reflexivity thesis) or are they instead merely statements of other reasons that we may already have (the paraphrastic thesis)? I argue for a version of the paraphrastic thesis. In doing so, considerable attention is given to the neglected but important puzzle of the opaqueness of rules, which arises out of what (...) some regard as the gap between the evaluative grounds of legal rules and what makes them into reasons for action. After examining an important articulation of the puzzle in the work of Joseph Raz, I argue that the reflexivity thesis is (i) undermined by certain features of rule making and (ii) defeated by the principle of presumptive sufficiency. The result is that it is possible for legal rules to be paraphrastic statements of reasons but, conversely, impossible for them to be reasons in themselves. (shrink)

In 2020, the European Union announced the award of the contract for the biometric part of the new database for border control, the Entry Exit System, to two companies: IDEMIA and Sopra Steria. Both companies had been previously involved in the development of databases for border and migration management. While there has been a growing amount of publicly available documents that show what kind of technologies are being implemented, for how much money, and by whom, there has been limited engagement (...) with digital methods in this field. Moreover, critical border and security scholarship has largely focused on qualitative and ethnographic methods. Building on a data feminist approach, we propose a transdisciplinary methodology that goes beyond binaries of qualitative/quantitative and opacity/transparency, examines power asymmetries and makes the labour of coding visible. Empirically, we build and analyse a dataset of the contracts awarded by two European Union agencies key to its border management policies – the European Agency for Large-Scale Information Systems (eu-LISA) and the European Border and Coast Guard Agency (Frontex). We supplement the digital analysis and visualisation of networks of companies with close reading of tender documents. In so doing, we show how a transdisciplinary methodology can be a device for making datafication ‘intelligible’ at the European Union borders. (shrink)

In this paper, we outline an approach to giving extensional truth-theoretic semantics for what have traditionally been seen as opaque sentential contexts. We outline an approach to providing a compositional truth-theoretic semantics for opaque contexts which does not require quantifying over intensional entities of any kind, and meets standard objections to such accounts. The account we present aims to meet the following desiderata on a semantic theory T for opaque contexts: (D1) T can be formulated in a (...) first-order extensional language; (D2) T does not require quantification over intensional entities­i.e., meanings, propositions, properties, relations, or the like­in its treatment of opaque contexts; (D3) T captures the entailment relations that hold in virtue of form between sentences in the language for which it is a theory; (D4) T has a finite number of axioms. If the approach outlined here is correct, it resolves a longstanding complex of problems in metaphysics, the philosophy of mind and the philosophy of language. (shrink)

Artificial intelligence (AI) and machine learning (ML) models have undoubtedly become useful tools in science. In general, scientists and ML developers are optimistic – perhaps rightfully so – about the potential that these models have in facilitating scientific progress. The philosophy of AI literature carries a different mood. The attention of philosophers remains on potential epistemological issues that stem from the so-called “black box” features of ML models. For instance, Eamon Duede (2023) argues that opacity in deep learning models (DLMs) (...) is epistemically problematic in the context of justification, though not in the context of discovery. -/- In this paper, I aim to show that a similar epistemological concern is echoed in the epistemology of imagination literature. It is traditionally held that, given its black box features, reliance on the imagination is epistemically problematic in the context of justification, though not in the context of discovery. The constraints-based approach to the imagination answers the epistemological concern by providing an account of how we can rely on the imagination in the context of justification by way of constraints. I argue by analogy that a similar approach can be applied to the opaque DLM case. Ultimately, my goal is to explore just how far this analogy extends, and whether a constraints-based approach to opaque DLMs can answer the epistemological concern surrounding their black box features in the context of justification. -/- (Note that this paper is IN PROGRESS and UNPUBLISHED). (shrink)

Under what conditions does machine learning (ML) model opacity inhibit the possibility of explaining and understanding phenomena? In this article, I argue that nonepistemic values give shape to the ML opacity problem even if we keep researcher interests fixed. Treating ML models as an instance of doing model-based science to explain and understand phenomena reveals that there is (i) an external opacity problem, where the presence of inductive risk imposes higher standards on externally validating models, and (ii) an internal opacity (...) problem, where greater inductive risk demands a higher level of transparency regarding the inferences the model makes. (shrink)

Many artificial intelligence (AI) systems currently used for decision-making are opaque, i.e., the internal factors that determine their decisions are not fully known to people due to the systems’ computational complexity. In response to this problem, several researchers have argued that human decision-making is equally opaque and since simplifying, reason-giving explanations (rather than exhaustive causal accounts) of a decision are typically viewed as sufficient in the human case, the same should hold for algorithmic decision-making. Here, I contend that (...) this argument overlooks that human decision-making is sometimes significantly more transparent and trustworthy than algorithmic decision-making. This is because when people explain their decisions by giving reasons for them, this frequently prompts those giving the reasons to govern or regulate themselves so as to think and act in ways that confirm their reason reports. AI explanation systems lack this self-regulative feature. Overlooking it when comparing algorithmic and human decision-making can result in underestimations of the transparency of human decision-making and in the development of explainable AI that may mislead people by activating generally warranted beliefs about the regulative dimension of reason-giving. (shrink)

The increasing demand for transparency in AI has recently come under scrutiny. The question is often posted in terms of “epistemic double standards”, and whether the standards for transparency in AI ought to be higher than, or equivalent to, our standards for ordinary human reasoners. I agree that the push for increased transparency in AI deserves closer examination, and that comparing these standards to our standards of transparency for other opaque systems is an appropriate starting point. I suggest that (...) a more fruitful exploration of this question will involve a different comparison class. We routinely treat judgments made by highly trained experts in specialized fields as fair or well-grounded even though—by the nature of expert/layperson division of epistemic labor—an expert will not be able to provide an explanation of the reasoning behind these judgments that makes sense to most other people. Regardless, laypeople are thought to be acting reasonably—and ethically—in deferring to the judgments of experts that concern their areas of specialization. I suggest that we reframe our question regarding the appropriate standards of transparency in AI as one that asks when, why, and to what degree it would be ethical to accept opacity in AI. I argue that our epistemic relation to certain opaque AI technology may be relevantly similar to the layperson’s epistemic relation to the expert in certain respects, such that the successful expert/layperson division of epistemic labor can serve as a blueprint for the ethical use of opaque AI. (shrink)

Population obesity and associated morbidities pose significant public health and economic burdens in the United Kingdom, United States, and globally. As a response, public health initiatives often seek to change individuals’ unhealthy behavior, with the dual aims of improving their health and conserving health care resources. One such initiative—taxes on sugar‐sweetened beverages (SSB)—has sparked considerable ethical debate. Prominent in the debate are arguments seeking to demonstrate the supposed impermissibility of SSB taxes and similar policies on the grounds that they interfere (...) with individuals’ freedom and autonomy. Commentators have often assumed that a policy intended to restrict or change private individuals’ consumption behavior will necessarily curtail freedom and, as a corollary, will undermine individuals’ autonomy with respect to their consumption choices. Yet this assumption involves a conceptual mistake. To address the misunderstanding, it’s necessary to attend to the differences between negative liberty, freedom of options, and autonomy. Ultimately, concerns about negative liberty, freedom, and autonomy do not provide strong grounds for opposing SSB taxes. (shrink)

In Outlines of Pyrrhonism, Sextus Empiricus takes pains to differentiate the skeptical way of life from other positions with which it is often confused, and in the course of this discussion he briefly explains how skepticism differs from Cyrenaicism. Surprisingly, Sextus does not mention an important apparent difference between the two. The Cyrenaics have a positive epistemic commitment--that we can apprehend our own affections. Although we cannot know whether the honey is really sweet, we can know infallibly that right now (...) we are being sweetened. By contrast, Sextus says explicitly that, as skeptics, Pyrrhonists apprehend nothing whatsoever. A case can (and has) been made that Sextus does not mention this difference because, on this matter, there really isn't an important difference between the two: the skeptic is perfectly able to report how things appear to him, e.g., that the honey seems sweet, and it is crucial for the skeptic that he not abolish the appearances. But, I argue, what the skeptics are doing when they report how things appear to them is importantly different from the sort of immediate, infallible apprehension of one's own affections claimed by the Cyrenaics, as the latter involves theoretical commitments regarding the nature of one's affections that the skeptic eschews. (shrink)

As systems based on opaque Artificial Intelligence (AI) continue to flourish in diverse real-world applications, understanding these black box models has become paramount. In response, Explainable AI (XAI) has emerged as a field of research with practical and ethical benefits across various domains. This paper not only highlights the advancements in XAI and its application in real-world scenarios but also addresses the ongoing challenges within XAI, emphasizing the need for broader perspectives and collaborative efforts. We bring together experts from (...) diverse fields to identify open problems, striving to synchronize research agendas and accelerate XAI in practical applications. By fostering collaborative discussion and interdisciplinary cooperation, we aim to propel XAI forward, contributing to its continued success. Our goal is to put forward a comprehensive proposal for advancing XAI. To achieve this goal, we present a manifesto of 27 open problems categorized into nine categories. These challenges encapsulate the complexities and nuances of XAI and offer a road map for future research. For each problem, we provide promising research directions in the hope of harnessing the collective intelligence of interested stakeholders. (shrink)

Three classic distinctions specify that truths can be necessary versus contingent,analytic versus synthetic, and a priori versus a posteriori. The philosopher reading this article knows very well both how useful and ordinary such distinctions are in our conceptual work and that they have been subject to many and detailed debates, especially the last two. In the following pages, I do not wish to discuss how far they may be tenable. I shall assume that, if they are reasonable and non problematic (...) in some ordinary cases, then they can be used in order to understand what kind of knowledge the maker’s knowledge is. By this I mean the sort of knowledge that Alice enjoys when she holds the information that Bob’s coffee is sweetened because she just put two spoons of sugar in it herself. The maker’s knowledge tradition is quite important but it is not mainstream in modern and analytic epistemology and lacks grounding in terms of exactly what sort of knowledge one is talking about. My suggestion is that this grounding can be provided by a minimalist approach, based on an information-theoretical analysis. In the article, I argue that we need to decouple a fourth distinction, namely informative versus uninformative, from the previous three and, in particular, from its implicit association with analytic versus synthetic and a priori versus a posteriori; such a decoupling facilitates, and is facilitated by, moving from a monoagent to a multiagent approach: the distinctions qualify a proposition, a message, or some information not just in themselves but relationally, with respect to an informational agent; the decoupling and the multiagent approach enable a re-mapping of currently available positions in epistemology on these four dichotomies; within such a re-mapping, two positions, capturing the nature of a witness’ knowledge and of a maker’s knowledge, can best be described as contingent, synthetic, a posteriori, and uninformative and as contingent, synthetic, weakly a priori, and uninformative respectively. In the conclusion, I indicate why the analysis of the maker’s knowledge has important consequences in all those cases in which the poietic intervention on a system determines the truth of the model of that system. (shrink)

One way to interpret the work of Joshua Greene (2001; 2008; 2014) is that the wave of empirical research into moral decision-making is a way for us to become more confident in our ability to gain moral knowledge. We argue that empirical research into moral judgment has shown (both survey-based and brain-based) that the grounds of moral judgment are opaque on several dimensions. We argue that we cannot firmly grasp what the morally relevant/irrelevant features of a decision context are, (...) understand which cognitive system responds to said features, and further know that we are actually basing our judgment upon what we take to be the morally relevant features. That is, we argue that moral judgment is very much like other kinds of judgment: Deeply influenced by facts about the decision-context and the decision-maker. However, if moral judgment is merely the exercise of our general judgment making faculty, we shouldn’t be surprised by this result. We conclude by considering what this means for our practice of making moral judgments. (shrink)

Simple idealized models seem to provide more understanding than opaque, complex, and hyper-realistic models. However, an increasing number of scientists are going in the opposite direction by utilizing opaque machine learning models to make predictions and draw inferences, suggesting that scientists are opting for models that have less potential for understanding. Are scientists trading understanding for some other epistemic or pragmatic good when they choose a machine learning model? Or are the assumptions behind why minimal models provide understanding (...) misguided? In this paper, using the case of deep neural networks, I argue that it is not the complexity or black box nature of a model that limits how much understanding the model provides. Instead, it is a lack of scientific and empirical evidence supporting the link that connects a model to the target phenomenon that primarily prohibits understanding. (shrink)

According to what is perhaps the dominant picture of reference, what a referential term refers to in a context is determined by what the speaker intends for her audience to identify as the referent. I argue that this sort of broadly Gricean view entails, counterintuitively, that it is impossible to knowingly use referential terms in ways that one expects or intends to be misunderstood. Then I sketch an alternative which can better account for such opaque uses of language, or (...) what I call “sneaky reference.” I close by reflecting on the ramifications of these arguments for the theory of meaning more broadly. (shrink)

Recent metaphysics has turned its focus to two notions that are—as well as having a common Aristotelian pedigree—widely thought to be intimately related: grounding and essence. Yet how, exactly, the two are related remains opaque. We develop a unified and uniform account of grounding and essence, one which understands them both in terms of a generalized notion of identity examined in recent work by Fabrice Correia, Cian Dorr, Agustín Rayo, and others. We argue that the account comports with antecedently (...) plausible principles governing grounding, essence, and identity taken individually, and illuminates how the three interact. We also argue that the account compares favorably to an alternative unification of grounding and essence recently proposed by Kit Fine. (shrink)

Scientists depend on complex computational systems that are often ineliminably opaque, to the detriment of our ability to give scientific explanations and detect artifacts. Some philosophers have s...

Sometimes a proposition is ‘opaque’ to an agent: he doesn't know it, but he does know something about how coming to know it should affect his or her credence function. It is tempting to assume that a rational agent's credence function coheres in a certain way with his or her knowledge of these opaque propositions, and I call this the ‘Opaque Proposition Principle’. The principle is compelling but demonstrably false. I explain this incongruity by showing that the (...) principle is ambiguous: the term ‘know’ as it appears in the principle can be interpreted in two different ways, as either basic-know or super-know. I use this distinction to construct a plausible version of the principle, and then to similarly construct plausible versions of the Reflection Principle and the Sure-Thing Principle. (shrink)

Computation is central to contemporary mathematics. Many accept that we can acquire genuine mathematical knowledge of the Four Color Theorem from Appel and Haken's program insofar as it is simply a repetitive application of human forms of mathematical reasoning. Modern LLMs / DNNs are, by contrast, opaque to us in significant ways, and this creates obstacles in obtaining mathematical knowledge from them. We argue, however, that if a proof-checker automating human forms of proof-checking is attached to such machines, then (...) we can obtain apriori mathematical knowledge from them, even though the original machines are entirely opaque to us and the proofs they output are not human-surveyable. (shrink)

Attempts to elucidate grounding are often made by connecting grounding to metaphysical explanation, but the notion of metaphysical explanation is itself opaque, and has received little attention in the literature. We can appeal to theories of explanation in the philosophy of science to give us a characterization of metaphysical explanation, but this reveals a tension between three theses: that grounding relations are objective and mind-independent; that there are pragmatic elements to metaphysical explanation; and that grounding and metaphysical explanation share (...) a close connection. Holding fixed the mind-independence of grounding, I show that neither horn of the resultant dilemma can be blunted. Consequently, we should reject the assumption that grounding relations are mind-independent. (shrink)

People increasingly form beliefs based on information gained from automatically filtered Internet ‎sources such as search engines. However, the workings of such sources are often opaque, preventing ‎subjects from knowing whether the information provided is biased or incomplete. Users’ reliance on ‎Internet technologies whose modes of operation are concealed from them raises serious concerns about ‎the justificatory status of the beliefs they end up forming. Yet it is unclear how to address these concerns ‎within standard theories of knowledge and (...) justification. To shed light on the problem, we introduce a ‎novel conceptual framework that clarifies the relations between justified belief, epistemic responsibility, ‎action, and the technological resources available to a subject. We argue that justified belief is subject to ‎certain epistemic responsibilities that accompany the subject’s particular decision-taking circumstances, ‎and that one typical responsibility is to ascertain, so far as one can, whether the information upon which ‎the judgment will rest is biased or incomplete. What this responsibility comprises is partly determined by ‎the inquiry-enabling technologies available to the subject. We argue that a subject’s beliefs that are ‎formed based on Internet-filtered information are less justified than they would be if she either knew how ‎filtering worked or relied on additional sources, and that the subject may have the epistemic ‎responsibility to take measures to enhance the justificatory status of such beliefs.‎. (shrink)

We might worry that our seamless reliance on AI systems makes us prone to adopting the strange errors that these systems commit. One proposed solution is to design AI systems so that they are not phenomenally transparent to their users. This stops cognitive extension and the automatic uptake of errors. Although we acknowledge that some aspects of AI extension are concerning, we can address these concerns without discouraging transparent employment altogether. First, we believe that the potential danger should be put (...) into perspective – many unreliable technologies are unlikely to be used transparently precisely because they are unreliable. Second, an agent who transparently employs a resource may also reflect (opaquely) on its reliability. Finally, agents can rely on a process transparently and be yanked out of their transparent use when it turns unreliable. When an agent is responsive to the reliability of their process in this way, they have epistemically integrated it, and the beliefs they form with it are formed responsibly. This prevents the agent from automatically incorporating problematic beliefs. Responsible (and transparent) use of AI resources – and consequently responsible AI extension – is hence possible. We end the paper with several design and policy recommendations that encourage epistemic integration of AI-involving belief-forming processes. (shrink)

Despite the recent proliferation of scientific, clinical, and narrative accounts of auditory verbal hallucinations, the phenomenology of voice hearing remains opaque and undertheorized. In this article, we outline an interdisciplinary approach to understanding hallucinatory experiences which seeks to demonstrate the value of the humanities and social sciences to advancing knowledge in clinical research and practice. We argue that an interdisciplinary approach to the phenomenology of AVH utilizes rigorous and context-appropriate methodologies to analyze a wider range of first-person accounts of (...) AVH at 3 contextual levels: cultural, social, and historical; experiential; and biographical. We go on to show that there are significant potential benefits for voice hearers, clinicians, and researchers. These include informing the development and refinement of subtypes of hallucinations within and across diagnostic categories; “front-loading” research in cognitive neuroscience; and suggesting new possibilities for therapeutic intervention. In conclusion, we argue that an interdisciplinary approach to the phenomenology of AVH can nourish the ethical core of scientific enquiry by challenging its interpretive paradigms, and offer voice hearers richer, potentially more empowering ways to make sense of their experiences. (shrink)

The ongoing explosion of interest in artificial intelligence is fueled in part by recently developed techniques in machine learning. Those techniques allow automated systems to process huge amounts of data, utilizing mathematical methods that depart from traditional statistical approaches, and resulting in impressive advancements in our ability to make predictions and uncover correlations across a host of interesting domains. But as is now widely discussed, the way that those systems arrive at their outputs is often opaque, even to the (...) experts who design and deploy them. Is it morally problematic to make use of opaque automated methods when making high-stakes decisions, like whether to issue a loan to an applicant, or whether to approve a parole request? Many scholars answer in the affirmative. However, there is no widely accepted explanation for why transparent systems are morally preferable to opaque systems. We argue that the use of automated decision-making systems sometimes violates duties of consideration that are owed by decision-makers to decision-subjects, duties that are both epistemic and practical in character. Violations of that kind generate a weighty consideration against the use of opaque decision systems. In the course of defending our approach, we show that it is able to address three major challenges sometimes leveled against attempts to defend the moral import of transparency in automated decision-making. (shrink)

In this paper, I identify the hard problem of access for epistemological disjunctivism (ED): given that perceptual experience E is opaque with respect to its own epistemic properties, subject S is not in a position to know epistemic proposition (i) (that E is factive with respect to empirical proposition p) just by having E and/or reflecting on E. This is the case even if (i) is true. I first motivate the hard problem of access (Section 2) and then reconstruct (...) and analyze three of the ways in which EDists have argued for the internal accessibility of the factive character of perceptual experience. These arguments explain internal access in terms of the unity of perceptual and rational capacities (Section 3), favoring support (Section 4), and the outward-looking model of self-knowledge (Section 5). My conclusion (Section 6) is that none of these responses works. I then suggest how ED might be modified to succeed as an access internalist epistemology. (shrink)

We demonstrate that many philosophers accept the following claim: When an aesthetic object is apprehended correctly, taking pleasure in said object is a reliable sign that the object is aesthetically successful. We undermine this position by showing that what grounds our pleasurable experience is opaque: In many cases, the experienced pleasure is attributable to factors that have little to do with the aesthetic object. The evidence appealed to is a form of Higher-Order Evidence (HOE) and we consider attempts to (...) overcome said evidence. We argue they are unsuccessful. We conclude by considering what this means for our practice of making aesthetic judgements. (shrink)

Since the 1990s the so-called transparency of experience has played a crucial role in core debates in philosophy of mind. However, recent developments in the literature have made transparency itself quite opaque. The very idea of transparent experience has become quite fuzzy, due to the articulation of many different notions of transparency and transparency theses. Absent a unified logical space where these notions and theses can be mapped and confronted, we are left with an overall impression of conceptual chaos. (...) This is a problem, given the constant and ubiquitous references to transparency in the literature and its prominent position in the contemporary philosophy of mind. My goal in this paper is to restore clarity through proper analysis of the mutual relations between the different transparency theses. This allows me to uncover a unitary multidimensional logical space where existing (as well as possible) views can be properly singled out and located. (shrink)

Advancements in machine learning have fuelled the popularity of using AI decision algorithms in procedures such as bail hearings, medical diagnoses and recruitment. Academic articles, policy texts, and popularizing books alike warn that such algorithms tend to be opaque: they do not provide explanations for their outcomes. Building on a causal account of transparency and opacity as well as recent work on the value of causal explanation, I formulate a moral concern for opaque algorithms that is yet to (...) receive a systematic treatment in the literature: when such algorithms are used in life-changing decisions, they can obstruct us from effectively shaping our lives according to our goals and preferences, thus undermining our autonomy. I argue that this concern deserves closer attention as it furnishes the call for transparency in algorithmic decision-making with both new tools and new challenges. (shrink)

One of the primary, if not most critical, difficulties in the design and implementation of autonomous systems is the black-boxed nature of the decision-making structures and logical pathways. How human values are embodied and actualised in situ may ultimately prove to be harmful if not outright recalcitrant. For this reason, the values of stakeholders become of particular significance given the risks posed by opaque structures of intelligent agents (IAs). This paper explores how decision matrix algorithms, via the belief-desire-intention model (...) for autonomous vehicles, can be designed to minimize the risks of opaque architectures. Primarily through an explicit orientation towards designing for the values of explainability and verifiability. In doing so, this research adopts the Value Sensitive Design (VSD) approach as a principled framework for the incorporation of such values within design. VSD is recognized as a potential starting point that offers a systematic way for engineering teams to formally incorporate existing technical solutions within ethical design, while simultaneously remaining pliable to emerging issues and needs. It is concluded that the VSD methodology offers at least a strong enough foundation from which designers can begin to anticipate design needs and formulate salient design flows that can be adapted to the changing ethical landscapes required for utilisation in autonomous vehicles. (shrink)

A satisfactory theory of linguistic communication must explain how it is that, through the interpersonal exchange of auditory, visual, and tactile stimuli, the communicative preconditions for the acquisition of testimonial knowledge regularly come to be satisfied. Without an account of knowledge-yielding communication this success condition for linguistic theorizing is left opaque, and we are left with an incomplete understanding of testimony, and communication more generally, as a source of knowledge. This paper argues that knowledge-yielding communication should be modelled on (...) knowledge itself. It is argued that knowledge-yielding communication occurs iff interlocutors coordinate on truth values in a non-lucky and non-deviant way. This account is able to do significant explanatory work: it sheds light on the nature of referential communication, and it allows us to capture, in an informative way, the sense in which interlocutors must entertain similar propositions in order to communicate successfully. (shrink)

The increasing application of artificial intelligence (AI) to healthcare raises both hope and ethical concerns. Some advanced machine learning methods provide accurate clinical predictions at the expense of a significant lack of explainability. Alex John London has defended that accuracy is a more important value than explainability in AI medicine. In this article, we locate the trade-off between accurate performance and explainable algorithms in the context of distributive justice. We acknowledge that accuracy is cardinal from outcome-oriented justice because it helps (...) to maximize patients’ benefits and optimizes limited resources. However, we claim that the opaqueness of the algorithmic black box and its absence of explainability threatens core commitments of procedural fairness such as accountability, avoidance of bias, and transparency. To illustrate this, we discuss liver transplantation as a case of critical medical resources in which the lack of explainability in AI-based allocation algorithms is procedurally unfair. Finally, we provide a number of ethical recommendations for when considering the use of unexplainable algorithms in the distribution of health-related resources. (shrink)

“Experimental jurisprudence” draws on empirical data to inform questions typically associated with jurisprudence or legal theory. Scholars in this flourishing movement conduct empirical studies about a variety of legal language and concepts. Despite the movement’s growth, its justification is still opaque. Jurisprudence is the study of deep and longstanding theoretical questions about law’s nature, but “experimental jurisprudence,” it might seem, simply surveys laypeople. This Article elaborates and defends experimental jurisprudence. Experimental jurisprudence, appropriately understood, is not only consistent with traditional (...) jurisprudence; it is an essential branch of it. (shrink)

A common criticism of the use of algorithms in criminal justice is that algorithms and their determinations are in some sense ‘opaque’—that is, difficult or impossible to understand, whether because of their complexity or because of intellectual property protections. Scholars have noted some key problems with opacity, including that opacity can mask unfair treatment and threaten public accountability. In this paper, we explore a different but related concern with algorithmic opacity, which centers on the role of public trust in (...) grounding the legitimacy of criminal justice institutions. We argue that algorithmic opacity threatens the trustworthiness of criminal justice institutions, which in turn threatens their legitimacy. We first offer an account of institutional trustworthiness before showing how opacity threatens to undermine an institution’s trustworthiness. We then explore how threats to trustworthiness affect institutional legitimacy. Finally, we offer some policy recommendations to mitigate the threat to trustworthiness posed by the opacity problem. (shrink)

Since its emergence in the 1960s, Artifcial Intelligence (AI) has grown to conquer many technology products and their felds of application. Machine learning, as a major part of the current AI solutions, can learn from the data and through experience to reach high performance on various tasks. This growing success of AI algorithms has led to a need for interpretability to understand opaque models such as deep neural networks. Various requirements have been raised from diferent domains, together with numerous (...) tools to debug, justify outcomes, and establish the safety, fairness and reliability of the models. This variety of tasks has led to inconsistencies in the terminology with, for instance, terms such as interpretable, explainable and transparent being often used interchangeably in methodology papers. These words, however, convey diferent meanings and are “weighted" diferently across domains, for example in the technical and social sciences. In this paper, we propose an overarching terminology of interpretability of AI systems that can be referred to by the technical developers as much as by the social sciences community to pursue clarity and efciency in the defnition of regulations for ethical and reliable AI development. We show how our taxonomy and defnition of interpretable AI difer from the ones in previous research and how they apply with high versatility to several domains and use cases, proposing a—highly needed—standard for the communication among interdisciplinary areas of AI. (shrink)

This essay presents the Aestheticism of the 19th century as the foundational movement of modernist-formalist aesthetics of the 20th century. The main principle of this movement is what I denominate “productive opacity”. Aestheticism has not been recognized as a philosophical aesthetic theory. However, its definition of artwork as an exclusive kind of form—a deep, opaque form—is among the most precise ever given in the discipline. This essay offers an interpretation of aestheticism as a formalist theory, referred to here as (...) “deep formalism”, focusing on the thinking of leading aestheticists, Walter Pater, Oscar Wilde, and James Whistler. These three thinkers defined artwork as a form saturated with an inextricable content, viz. opaque form. (shrink)

Abstract: An indirect report typically takes the form of a speaker using the locution “said that” to report an earlier utterance. In what follows, I introduce the principal philosophical and pragmatic points of interest in the study of indirect reports, including the extent to which context sensitivity affects the content of an indirect report, the constraints on the substitution of co-referential terms in reports, the extent of felicitous paraphrase and translation, the way in which indirect reports are opaque, and (...) the use of indirect reports as pragmatic vehicles for other speech acts such as humor, insult, or irony. Throughout I develop several positions: (i) that a semantic analysis of indirect reports is insufficient, (ii) that the distinction between direct and indirect reports is not clear and that indirect reports are the predominate way of reporting while direct reports may be a para-linguistic variation on them, (iii) that most questions about the semantics and pragmatics of indirect reports will rely on a full understanding of the nature of what is reported and how it gets reported, (iv) that an analysis of reporting requires the pragmatic tools of metarepresentation and a social, inter-personal understanding of relevance and shared knowledge. (shrink)

Humean Supervenience is a metaphysical model of the world according to which all truths hold in virtue of nothing but the total spatiotemporal distribution of perfectly natural, intrinsic properties. David Lewis and others have worked out many aspects of HS in great detail. A larger motivational question, however, remains unanswered: As Lewis admits, there is strong evidence from fundamental physics that HS is false. What then is the purpose of defending HS? In this paper, we argue that the philosophical merit (...) of HS is largely independent of whether it correctly represents the world’s fundamental structure. In particular, we show that insofar as HS is an apt model of the world’s higher-level structure, it thereby provides a powerful argument for reductive physicalism and explains otherwise opaque inferential relations. Recent criticism of HS on the grounds that it misrepresents fundamental physical reality is, therefore, beside the point. (shrink)

Artificial intelligence (AI) systems are increasingly adopted to make decisions in domains such as business, education, health care, and criminal justice. However, such algorithmic decision systems can have prevalent biases against marginalized social groups and undermine social justice. Explainable artificial intelligence (XAI) is a recent development aiming to make an AI system’s decision processes less opaque and to expose its problematic biases. This paper argues against technical XAI, according to which the detection and interpretation of algorithmic bias can be (...) handled more or less independently by technical experts who specialize in XAI methods. Drawing on resources from feminist epistemology, we show why technical XAI is mistaken. Specifically, we demonstrate that the proper detection of algorithmic bias requires relevant interpretive resources, which can only be made available, in practice, by actively involving a diverse group of stakeholders. Finally, we suggest how feminist theories can help shape integrated XAI: an inclusive social-epistemic process that facilitates the amelioration of algorithmic bias. (shrink)

Can AI developers be held epistemically responsible for the processing of their AI systems when these systems are epistemically opaque? And can explainable AI (XAI) provide public justificatory reasons for opaque AI systems’ outputs? Koskinen (2024) gives negative answers to both questions. Here, I respond to her and argue for affirmative answers. More generally, I suggest that when considering people’s uncertainty about the factors causally determining an opaque AI’s output, it might be worth keeping in mind that (...) a degree of uncertainty about conclusions is inevitable even in entirely human-based empirical science because in induction there’s always a risk of getting it wrong. Keeping this in mind may help appreciate that requiring full transparency from AI systems before epistemically trusting their outputs might be unusually (and potentially overly) demanding. (shrink)

In the natural and social sciences, it is common to use toy models—extremely simple and highly idealized representations—to understand complex phenomena. Some of the simple surrogate models used to understand opaque machine learning (ML) models, such as rule lists and sparse decision trees, bear some resemblance to scientific toy models. They allow non-experts to understand how an opaque ML model works globally via a much simpler model that highlights the most relevant features of the input space and their (...) effect on the output. The obvious difference is that the common target of a toy and a full-scale model in the sciences is some phenomenon in the world, while the target of a surrogate model is another model. This essential difference makes toy surrogate models (TSMs) a new object of study for theories of understanding, one that is not easily accommodated under current analyses. This paper provides an account of what it means to understand an opaque ML model globally with the aid of such simple models. (shrink)

When Sartre arrives late to meet Pierre at a local establishment, he discovers not merely that Pierre is absent, but Pierre’s absence, where this depends, or so Sartre notoriously supposes, on a frustrated expectation that Pierre would be seen at that place. Many philosophers have railed against this view, taking it to entail a treatment of the ontology of absence that Richard Gale describes as ‘attitudinal’ – one whereby absences are thought to ontologically depend on psychological attitudes. In this paper, (...) I aim to make Sartre’s intuition respectable. What Sartre perceives is an absential location, only the boundaries of which are circumscribed by what Sartre is doing at that place: meeting Pierre. I explain how this Sartrean view, though not specifically attributable to Sartre, nonetheless honours some of the phenomenological data described, if a little opaquely, in Being and Nothingness. (shrink)

According to some philosophers, when introspectively attending to experience, we seem to see right through it to the objects outside, including their properties. This is called the transparency of experience. This paper examines whether, and in what sense, emotions are transparent. It argues that emotional experiences are opaque in a distinctive way: introspective attention to them does not principally reveal non-intentional somatic qualia but rather felt valenced intentional attitudes. As such, emotional experience is attitudinally opaque.

It has been argued that our currently most satisfactory social epistemology of science can’t account for science that is based on artificial intelligence (AI) because this social epistemology requires trust between scientists that can take full responsibility for the research tools they use, and scientists can’t take full responsibility for the AI tools they use since these systems are epistemically opaque. I think this argument overlooks that much AI-based science can be done without opaque models, and that agents (...) can take full responsibility for the systems they use even if these systems are opaque. Requiring that an agent fully understand how a system works is an untenably strong condition for that agent to take full responsibility for the system and risks absolving AI developers from responsibility for their products. AI-based science need not create trust-related social epistemological problems if we keep in mind that what makes both individual scientists and their use of AI systems trustworthy isn’t full transparency of the internal processing but their adherence to social and institutional norms that ensure that scientific claims can be trusted. (shrink)

Semanticists face substitution challenges even outside of contexts commonly recognized as opaque. Jennifer M. Saul has drawn attention to pairs of simple sentences - her term for sentences lacking a that-clause operator - of which the following are typical: -/- (1) Clark Kent went into the phone booth, and Superman came out. (1*) Clark Kent went into the phone booth, and Clark Kent came out. -/- (2) Superman is more successful with women than Clark Kent. (2*) Superman is more (...) successful with women than Superman. -/- She challenges us to explain why the upper and lower sentences in each pair differ, or at least appear to differ, in their truth-values and hence truth-conditions. This appearance of substitution failure is inherently puzzling. Moreover, it is taken by Saul to generate a dilemma for anyone hostile to direct reference accounts of that-clause constructions. Direct reference theorists regard the appearance of substitution failure in that-clause contexts as mere appearance, to be dealt with pragmatically rather than semantically. -/- Critics of such accounts need to say something about simple-sentence cases. If they choose to allow that intuitions of substitution failure can be over-ridden and explained away pragmatically in simple-sentence cases but not in that-clause cases, they lay themselves open to the charge of operating a double standard. But if they do not choose this option, they must offer a semantic explanation of apparent substitution failure in simple-sentence cases - no easy task, it turns out. Other respondents to Saul's challenge have sought to provide elaborate semantic treatments. In contrast, this paper proposes a far simpler pragmatic explanation of intuitions of substitution failure in simple sentences, an explanation that deploys no more resources than are to be found in Grice's 'Logic and Conversation'. Ironically, this proposal turns out to be incompatible with a direct reference perspective. So if it is, as I maintain, the most plausible treatment of simple-sentence cases available, Saul's initial thought gets turned around 180 degrees: the phenomenon she has drawn attention to ends up representing a challenge to supporters of direct reference theories. (shrink)

The Argument from Inductive Risk (AIR) is taken to show that values are inevitably involved in making judgements or forming beliefs. After reviewing this conclusion, I pose cases which are prima facie counterexamples: the unreflective application of conventions, use of black-boxed instruments, reliance on opaque algorithms, and unskilled observation reports. These cases are counterexamples to the AIR posed in ethical terms as a matter of personal values. Nevertheless, it need not be understood in those terms. The values which load (...) a theory choice may be those of institutions or past actors. This means that the challenge of responsibly handling inductive risk is not merely an ethical issue, but is also social, political, and historical. (shrink)

Here is my thesis (and the outline of this paper). Increasingly secret, complex and inscrutable computational systems are being used to intensify existing power relations, and to create new ones (Section II). To be all-things-considered morally permissible, new, or newly intense, power relations must in general meet standards of procedural legitimacy and proper authority (Section III). Legitimacy and authority constitutively depend, in turn, on a publicity requirement: reasonably competent members of the political community in which power is being exercised must (...) be able to determine that power is being exercised legitimately and with proper authority (Section IV). The publicity requirement can be satisfied only if the powerful can explain their decision-making—including the computational tools that they use to support it—to members of their political community. Section V applies these ideas to opaque computational systems. Section VI addresses objections; Section VII concludes. (shrink)