Many AI systems that make important decisions are black boxes: how they function is opaque even to their developers. This is due to their high complexity and to the fact that they are trained rather than programmed. Efforts to alleviate the opacity of black box systems are typically discussed in terms of transparency, interpretability, and explainability. However, there is little agreement about what these key concepts mean, which makes it difficult to adjudicate the success or promise of opacity alleviation methods. I argue for a unified account of these key concepts that treats the concept of understanding as fundamental. This allows resources from the philosophy of science and the epistemology of understanding to help guide opacity alleviation efforts. A first significant benefit of this understanding account is that it defuses one of the primary, in-principle objections to post hoc explainable AI (XAI) methods. This “rationalization objection” argues that XAI methods provide mere rationalizations rather than genuine explanations. This is because XAI methods involve using a separate “explanation” system to approximate the original black box system. These explanation systems function in a completely different way than the original system, yet XAI methods make inferences about the original system based on the behavior of the explanation system. I argue that, if we conceive of XAI methods as idealized scientific models, this rationalization worry is dissolved. Idealized scientific models misrepresent their target phenomena, yet are capable of providing significant and genuine understanding of their targets.

Subjectivism about wellbeing rests on the idea that what is good for a person must ‘fit’ her, ‘resonate’ with her, not be ‘alien’ to her, etc. This idea has been called the ‘beating heart’ of subjectivism. In this article, I present the No-Beating-Heart Challenge for subjectivism, which holds that there is no satisfactory statement of this idea. I proceed by first identifying three criteria that any statement of the idea must meet if it is to provide support for subjectivism: Distinctness, Exclusiveness, and Explicitness. I then argue that no statement of this idea meets these criteria.

This paper addresses the issue of finite versus countable additivity in Bayesian probability and decision theory – in particular, Savage’s theory of subjective expected utility and personal probability. I show that Savage’s reason for not requiring countable additivity in his theory is inconclusive. The assessment leads to an analysis of various highly idealized assumptions commonly adopted in Bayesian theory, where I argue that a healthy dose of, what I call, conceptual realism is often helpful in understanding the interpretational value of sophisticated mathematical structures employed in applied sciences like decision theory. In the last part, I introduce countable additivity into Savage’s theory and explore some technical properties in relation to other axioms of the system.

Current debates about ideal theory and idealization in modern moral and political philosophy do not typically scrutinize the form of reflection itself. This is an unfortunate oversight: assumptions about the form of reflection shape the positions defended in those debates. I argue that the appropriate form of reflection on the nature and justification of standards of justice and morality is the form of practical reason. I further argue that the form of practical reason cannot support many of the idealizations typically deployed in modern moral and political philosophy.

Rawls ignited a debate in political theory when he introduced a division between the ideal and nonideal parts of a theory of justice. In the ideal part of the theory, one presents a positive conception of justice in a setting that assumes perfect compliance with the rules of justice. In the nonideal part, one addresses the question of what happens under departures from compliance. Critics of Rawls have attacked his focus on ideal theory as a form of utopianism, and have argued that political theory should be focused instead on providing solutions to the manifest injustices of the real world. In this essay, I offer a defense of the ideal/nonideal theory distinction according to which it amounts to nothing more than a division of labor, and explore some scientific analogies. Rawls’s own focus on the ideal part of the theory, I argue, stems from a felt need to clarify the foundations of justice, rather than a utopian neglect of real world problems.

It is argued that core areas of philosophy can benefit from reflection on cross-disciplinary research (CDR). We start by giving a brief account of CDR, describing its variability and some of the ways in which philosophers can interact with it. We then provide an argument in principle for the conclusion that CDR is philosophically fecund, arguing that since CDR highlights fundamental differences among disciplinary research worldviews, it can be used to motivate new philosophical problems and supply new insights into old problems. We close by providing an argument by example that uses the epistemology of peer disagreement to establish the potential of CDR for core philosophical areas. With this argument, we aim to demonstrate how the complex research contexts that CDR affords can point the way toward important avenues of epistemological research by highlighting potential limitations of key epistemological components, such as peerage and uniqueness.

Le calcul de structures mécaniques est un outil d'aide à la conception aujourd'hui largement utilisé. Quelle que soit la méthode de calcul de structures utilisée, l'adaptation du modèle de conception (géométrie, données technologiques, spécifications fonctionnelles) en un modèle de comportement spécifique au calcul de structures est une étape nécessaire nommée idéalisation. Cette étape particulièrement critique repose principalement sur l'expertise de l'analyste. Cet article propose une méthode d'aide aux choix d'idéalisation d'un modèle CAO en un modèle de comportement. La méthode proposée peut aider un analyste à effectuer ses choix d'idéalisation en respectant la qualité, le coût et les délais de simulation du système analysé. Elle est basée sur l'exploitation de résultats obtenus sur des calculs de référence, jugés similaires au calcul à traiter.