In recent years, the ethical impact of AI has been increasingly scrutinized, with public scandals emerging over biased outcomes, lack of transparency, and the misuse of data. This has led to a growing mistrust of AI and increased calls for mandated ethical audits of algorithms. Current proposals for ethical assessment of algorithms are either too high level to be put into practice without further guidance, or they focus on very specific and technical notions of fairness or transparency that do not (...) consider multiple stakeholders or the broader social context. In this article, we present an auditing framework to guide the ethical assessment of an algorithm. The audit instrument itself is comprised of three elements: a list of possible interests of stakeholders affected by the algorithm, an assessment of metrics that describe key ethically salient features of the algorithm, and a relevancy matrix that connects the assessed metrics to stakeholder interests. The proposed audit instrument yields an ethical evaluation of an algorithm that could be used by regulators and others interested in doing due diligence, while paying careful attention to the complex societal context within which the algorithm is deployed. (shrink)

In this article, we re-examine Pascal’s Mugging, and argue that it is a deeper problem than the St. Petersburg paradox. We offer a way out that is consistent with classical decision theory. Specifically, we propose a “many muggers” response analogous to the “many gods” objection to Pascal’s Wager. When a very tiny probability of a great reward becomes a salient outcome of a choice, such as in the offer of the mugger, it can be discounted on the condition that there (...) are other symmetric, non-salient rewards that one may receive if one chooses otherwise. (shrink)

The “problem of memory” in epistemology is concerned with whether and how we could have knowledge, or at least justification, for trusting our apparent memories. I defend an inductive solution—more precisely, an abductive solution—to the problem. A natural worry is that any such solution would be circular, for it would have to depend on memory. I argue that belief in the reliability of memory can be justified from the armchair, without relying on memory. The justification is, roughly, that my having (...) the sort of experience that my apparent memory should lead me to expect is best explained by the hypothesis that my memories are reliable. My solution is inspired by Harrod’s (1942) inductive solution. Coburn (1960) argued that Harrod’s solution contains a fatal flaw. I show that my solution is not vulnerable to Coburn’s objection, and respond to a number of other, recent and likely objections. (shrink)

In Justification without Awareness (2006), Michael Bergmann presents a dilemma for internalism from which he claims there is “no escape”: The awareness allegedly required for justification is either strong awareness, which involves conceiving of some justification-contributor as relevant to the truth of a belief, or weak awareness, which does not. Bergmann argues that the former leads to an infinite regress of justifiers, while the latter conflicts with the “clearest and most compelling” motivation for endorsing internalism, namely, that for a belief (...) to be justified its truth must not be an accident from the subject’s perspective. Bergmann’s dilemma might initially seem to have the force of a knock-down argument against the classical foundationalist accounts he considers, if not against all forms of internalism. I argue, however, that the weak-awareness horn of Bergmann’s dilemma is unsuccessful. Classical foundationalists can hold on to the main motivation for internalism and avoid a vicious regress of justifiers. (shrink)

In this paper, we distinguish between different sorts of assessments of algorithmic systems, describe our process of assessing such systems for ethical risk, and share some key challenges and lessons for future algorithm assessments and audits. Given the distinctive nature and function of a third-party audit, and the uncertain and shifting regulatory landscape, we suggest that second-party assessments are currently the primary mechanisms for analyzing the social impacts of systems that incorporate artificial intelligence. We then discuss two kinds of as-sessments: (...) an ethical risk assessment and a narrower, technical algo-rithmic bias assessment. We explain how the two assessments depend on each other, highlight the importance of situating the algorithm within its particular socio-technical context, and discuss a number of lessons and challenges for algorithm assessments and, potentially, for algorithm audits. The discussion builds on our team’s experience of advising and conducting ethical risk assessments for clients across dif-ferent industries in the last four years. Our main goal is to reflect on the key factors that are potentially ethically relevant in the use of algo-rithms, and draw lessons for the nascent algorithm assessment and audit industry, in the hope of helping all parties minimize the risk of harm from their use. (shrink)

I defend external world realism. I assume that the principle of inference to the best explanation is justified: roughly, a hypothesis that provides a better explanation of the total evidence is more probable than one that does not. I argue that the existence of a world of spatial objects provides a systematic explanation of the spatial contents of visual experience, and that it provides a better explanation than traditional skeptical hypotheses. This paper thus pursues the explanationist strategy of Laurence BonJour (...) and Jonathan Vogel. It is an improved, more compelling defense, for at least two reasons. First, the attention to spatial properties, and in particular to what I call perspectival projections, makes the explanatory power of the realist hypothesis much more vivid and concrete. Second, the argument preserves and elucidates much that seems correct in the explanationist arguments others have offered while avoiding significant problems and shortcomings. (shrink)

According to foundationalism, some beliefs are justified but do not depend for their justification on any other beliefs. According to access internalism, a subject is justified in believing some proposition only if that subject is aware of or has access to some reason to think that the proposition is true or probable. In this paper I discusses a fundamental challenge to internalist foundationalism often referred to as the Sellarsian dilemma. I consider three attempts to respond to the dilemma – phenomenal (...) conservatism, BonJour’s classical foundationalism, and Fumerton’s classical foundationalism. I argue that, of these three, only the last seems to avoid getting impaled on one or the other horn of the dilemma. I end by responding to some concerns with Fumerton’s account. (shrink)

An increasing number of regulations propose the notion of ‘AI audits’ as an enforcement mechanism for achieving transparency and accountability for artificial intelligence (AI) systems. Despite some converging norms around various forms of AI auditing, auditing for the purpose of compliance and assurance currently have little to no agreed upon practices, procedures, taxonomies, and standards. We propose the ‘criterion audit’ as an operationalizable compliance and assurance external audit framework. We model elements of this approach after financial auditing practices, and argue (...) that AI audits should similarly provide assurance to their stakeholders about AI organizations’ ability to govern their algorithms in ways that mitigate harms and uphold human values. We discuss the necessary conditions for the criterion audit, and provide a procedural blueprint for performing an audit engagement in practice. We illustrate how this framework can be adapted to current regulations by deriving the criteria on which ‘bias audits’ for hiring algorithms can be performed, as required by the recently effective New York City Local Law 144 of 2021. We conclude by offering critical discussion on the benefits, inherent limitations, and implementation challenges of applying practices of the more mature financial auditing industry to AI auditing where robust guardrails against quality assurance issues are only starting to emerge. Our discussion as informed by experiences in performing these audits in practice highlights the critical role that an audit ecosystem plays in ensuring the effectiveness of such methodology. (shrink)

In this paper I argue that, given the way that AI models work and the way that ordinary human rationality works, it is very likely that people are anthropomorphizing AI, with potentially serious consequences. I start with the core idea, recently defended by Thomas Kelly (2022) among others, that bias involves a systematic departure from a genuine standard or norm. I briefly discuss how bias can take on different explicit, implicit, and “truly implicit” (Johnson 2021) forms such as bias by (...) proxy. I then discuss biased anthropomorphism of technology, focusing on the case of Large Language Models (LLMs) like chatGPT. As with other kinds of bias, there are importantly different kinds of anthropomorphism, some of which can persist without others, and some of which can encourage others. Anthropomorphism can take rather subtle, implicit forms, that can be difficult to detect, resist, and dislodge. Attention to these kinds of anthropomorphism can help us avoid confusing importantly different kinds of evaluation, better assess the risks, and inform strategies for bias prevention and mitigation. (shrink)

Al-Ghazali (1058-1111) was concerned that early Islamic philosophers were leaning too heavily and uncritically on Aristotelian and Neoplatonic ideas in developing their models of God and His relation to the world. He argued that their views were not only irreligious, but philosophically problematic, and he defended an alternative view aimed at staying closer to the Qur’an and the beliefs of the ordinary Muslim. Ibn Rushd (1126-1198) responded to al-Ghazali’s critique and developed a sophisticated Aristotelian view. The present chapter explores their (...) views in light of a problem facing any philosophical model of God in Islam or classical theism more generally, the problem of conceiving of God’s nature and relation to the world in a way that places an appropriate distance between God and humans. On the one hand, we want a notion of God that is not overly anthropomorphic, or that does not make him to be too much like us. On the other hand, we want to be able to say something positive and substantive about God. And we want to do this while preserving the harmony of reason and revelation, of philosophy and religion, as much as possible. (shrink)

It is commonly thought that we depend fundamentally on the “evidence of the senses” for our empicial beliefs, including and most directly, our beliefs about our local environment, the spatial world around us. The ultimate evidence we have for our perceptual beliefs is provided in some way by perception or perceptual experience. But what is this evidence? There seem to be three main options: external factualism allows that the evidence include facts about the external world; internal factualism takes facts that (...) involve only the internal, mental world—like facts about one’s perceptual experiences or appearances—to count as one’s perceptual evidence; and non-factualism takes propositions, including false ones, to count as evidence—for example, propositional contents of appearances or seemings. We shall see that all these options face significant challenges. Some might conclude that perceptual beliefs can be justified without depending on evidence. Others might choose to accept skepticism: our perceptual beliefs are not justified. I shall end, however, by tentatively suggesting that there is a sort of internal fact that is a plausible candidate for perceptual evidence that has been overlooked. (shrink)

This chapter centers around the question of whether theism is rational. We begin by discussing different theories of rationality, and introducing some importantly related epistemic concepts and controversies. We then consider the possible sources of rational belief in God and argue that even if these provide some positive support, the fact of religious disagreement defeats the rationality of theism.