Legislation and ethical guidelines around the globe call for effective human oversight of AI-based systems in high-risk contexts – that is oversight that reliably reduces the risks otherwise associated with the use of AI-based systems. Such risks may relate to the imperfect accuracy of systems (e.g., inaccurate classifications) or to ethical concerns (e.g., unfairness of outputs). Given the significant role that human oversight is expected to play in the operation of AI-based systems, it is crucial to better understand the conditions (...) for effective human oversight. We argue that the reliable detection of errors (as an umbrella term for inaccuracies and unfairness) is crucial for effective human oversight. We then propose that Signal Detection Theory (SDT) offers a promising framework for better understanding what affects people’s sensitivity (i.e., how well they are able to detect errors) and response bias (i.e., the tendency to report errors given a perceived evidence of an error) in detecting errors. Whereas an SDT perspective on the detection of inaccuracies is straightforward, we demonstrate its broader applicability by detailing the specifics for an SDT perspective on unfairness detection, including the need to choose a standard for (un)fairness. Additionally, we illustrate that an SDT perspective helps to better understand the conditions for effective error detection by showing examples of task-, system-, and person-related factors that may affect the sensitivity and response bias of humans tasked with detecting unfairness associated with the use of AI-based systems. Finally, we discuss future research directions for an SDT perspective on error detection. (shrink)

Recent work in explainable artificial intelligence (XAI) attempts to render opaque AI systems understandable through a divide-and-conquer strategy. However, this fails to illuminate how trained AI systems work as a whole. Precisely this kind of functional understanding is needed, though, to satisfy important societal desiderata such as safety. To remedy this situation, we argue, AI researchers should seek mechanistic interpretability, viz. apply coordinated discovery strategies familiar from the life sciences to uncover the functional organisation of complex AI systems. Additionally, theorists (...) should accommodate for the unique costs and benefits of such strategies in their portrayals of XAI research. (shrink)

Released mid-November 2022, Galactica is a set of six large language models (LLMs) of different sizes (from 125 M to 120B parameters) designed by Meta AI to achieve the ultimate ambition of “a single neural network for powering scientific tasks”, according to its accompanying whitepaper. It aims to carry out knowledge-intensive tasks, such as publication summarization, information ordering and protein annotation. However, just a few days after the release, Meta had to pull back the demo due to the strong hallucinatory (...) tendencies or underwhelming performances of the model. This article aims to study, through a critical threefold argument, the potential impacts of LLMs once deployed in the scientific value chain. Our first argument is a technical one. By examining the technicity of Galactica, it is possible to explain the descripancies between its promotional corporate discourse and abysmal outputs. Second, by going back to debates in both computer science and computational philosophy on the automation of abduction, we argue from the epistemological front that LLMs indeed cannot produce strong abductions and, therefore, claims about the automation of hypothesis generation remains chambering. Finally, our third argument is a sociological one. By conceptualizing the scientific field through Nancy Katherine Hayles’ cognitive assemblage theory, we aim to outline the potential steering of science by LLMs, mainly through information ordering. The core of our argument rests on the assertion that excessive control on information risks contravening a certain serendipitous aspect inherent to scientific discoveries. (shrink)

In this paper, I reflect on the puzzle that machine learning presents to social theory to develop an account of its distinct impact on social reality. I start by presenting how machine learning has presented a challenge to social theory as a research subject comprising both familiar and alien characteristics (1.). Taking this as an occasion for theoretical inquiry, I then propose a conceptual framework to investigate how algorithmic models of social phenomena relate to social reality and what their stochastic (...) mode of operation entails in terms of their sociality (2.). Analyzed through a theoretical lens that relies on central tenets of sociological systems theory, I find that machine learning implies a distinct epistemic transformation, based on how algorithmic modeling techniques process meaning as represented in data embedded in vector space. Building on this characterization, I introduce my conceptualization of stochastic technology as distinct from mechanistic technologies that rely on causal fixation (3.). Based on this understanding, I suggest that real-world applications of machine learning are often characterized by a constitutive tension between the stochastic properties of their outputs and the ways in which they are put to use in practice. Focussing on the large language models LaMDA and ChatGPT, I examine the epistemological implications of LLMs to account for the confusion of correlation and causality as the root of this tension. Next, I illustrate my theoretical conception by way of discussing an essay on image models by German media artist Hito Steyerl (4.). Following a critical reflection on Steyerl's characterization of Stable Diffusion as a “white box ”, I finally propose to conceive ofmachine learning-based technologies as stochastic contingency machines that transform social indeterminacy into contingent observations of social phenomena (5.) In this perspective, machine learning constitutes an epistemic technology that operates on meaning as extractable from data by means of algorithmic data modeling techniques to produce stochastic accounts of social reality. (shrink)

Bioethical consensus promotes a shared decision making model, which requires healthcare professionals to partner their knowledge with that of their patients—who, at a minimum, contribute their valu...

A central goal of research in explainable artificial intelligence (XAI) is to facilitate human understanding. However, understanding is an elusive concept that is difficult to target. In this paper, we argue that a useful way to conceptualize understanding within the realm of XAI is via certain human abilities. We present four criteria for a useful conceptualization of understanding in XAI and show that these are fulfilled by an abilities-based approach: First, thinking about understanding in terms of specific abilities is motivated (...) by research from numerous disciplines involved in XAI. Second, an abilities-based approach is highly versatile and can capture different forms of understanding important in XAI application contexts. Third, abilities can be operationalized for empirical studies. Fourth, abilities can be used to clarify the link between explainability, understanding, and societal desiderata concerning AI, like fairness and trustworthiness. Conceptualizing understanding as abilities can therefore support interdisciplinary collaboration among XAI researchers, provide practical benefit across diverse XAI application contexts, facilitate the development and evaluation of explainability approaches, and contribute to satisfying the societal desiderata of different stakeholders concerning AI systems. (shrink)

This paper delves into the use of machine learning-based systems in decision-making processes and its implications for moral responsibility as traditionally defined. It focuses on the emergence of responsibility gaps and examines proposed strategies to address them. The paper aims to provide an introductory and comprehensive overview of the ongoing debate surrounding moral responsibility in automated decision-making. By thoroughly examining these issues, we seek to contribute to a deeper understanding of the implications of AI integration in society.

This paper analyses the phenomenology and epistemology of chatbots such as ChatGPT and Bard. The computational architecture underpinning these chatbots are large language models (LLMs), which are generative artificial intelligence (AI) systems trained on a massive dataset of text extracted from the Web. We conceptualise these LLMs as multifunctional computational cognitive artifacts, used for various cognitive tasks such as translating, summarizing, answering questions, information-seeking, and much more. Phenomenologically, LLMs can be experienced as a “quasi-other”; when that happens, users anthropomorphise them. (...) For most users, current LLMs are black boxes, i.e., for the most part, they lack data transparency and algorithmic transparency. They can, however, be phenomenologically and informationally transparent, in which case there is an interactional flow. Anthropomorphising and interactional flow can, in some users, create an attitude of (unwarranted) trust towards the output LLMs generate. We conclude this paper by drawing on the epistemology of trust and testimony to examine the epistemic implications of these dimensions. Whilst LLMs generally generate accurate responses, we observe two epistemic pitfalls. Ideally, users should be able to match the level of trust that they place in LLMs to the degree that LLMs are trustworthy. However, both their data and algorithmic opacity and their phenomenological and informational transparency can make it difficult for users to calibrate their trust correctly. The effects of these limitations are twofold: users may adopt unwarranted attitudes of trust towards the outputs of LLMs (which is particularly problematic when LLMs hallucinate), and the trustworthiness of LLMs may be undermined. (shrink)

The use of artificial intelligence in healthcare contexts is highly controversial for the (bio)ethical conundrums it creates. One of the main problems arising from its implementation is the lack of transparency of machine learning algorithms, which is thought to impede the patient’s autonomous choice regarding their medical decisions. If the patient is unable to clearly understand why and how an AI algorithm reached certain medical decision, their autonomy is being hovered. However, there are alternatives to prevent the negative impact of (...) AI opacity in shared (healthcare professional–patient) decision-making processes, and benefit from the high accuracy of such systems. Despite these ethical issues, this paper advocates for its usage and also defends the positive role that it could have in enhancing patients’ autonomy. I advocate for a system where the implementation of AI is only partial, and humans will remain in control of oversighting algorithm interpretation of medical images and patients’ data. (shrink)

Definition of the problem In medicine, as well as in other disciplines, computer science expertise is becoming increasingly important. This requires a culture of interdisciplinary assessment, for which medical ethics committees are not well prepared. The use of big data and artificial intelligence (AI) methods (whether developed in-house or in the form of “tools”) pose further challenges for research ethics reviews. Arguments This paper describes the problems and suggests solving them through procedural changes. Conclusion An assessment that is interdisciplinary from (...) the outset appears to be more suitable than having two commissions with different expertise. However, this would require that the composition of medical ethics committees be altered. In addition, the article recommends initial measures to be taken during research ethics reviews of big data and AI projects in order to consolidate the review process and ensure standardization of the criteria. (shrink)

To analyze potential negative consequences of adopting generative AI solutions in the digital afterlife industry (DAI), in this paper we present three speculative design scenarios for AI-enabled simulation of the deceased. We highlight the perspectives of the data donor, data recipient, and service interactant – terms we employ to denote those whose data is used to create ‘deadbots,’ those in possession of the donor’s data after their death, and those who are meant to interact with the end product. We draw (...) on the scenarios to map out several key ethical concerns posed by ‘re-creation services’ and to put forward recommendations on the ethical development of AI systems in this specific area of application. The recommendations, targeted at providers of AI-enabled re-creation services, include suggestions for developing sensitive procedures for retiring deadbots, ensuring meaningful transparency, restricting access to such services to adult users only, and adhering to the principle of mutual consent of both data donors and service interactants. While we suggest practical solutions to the socio-ethical challenges posed by the emergence of re-creation services, we also emphasize the importance of ongoing interdisciplinary research at the intersection of the ethics of AI and the ethics of the DAI. (shrink)

A novel advantage of the use of machine learning (ML) systems in medicine is their potential to continue learning from new data after implementation in clinical practice. To date, considerations of the ethical questions raised by the design and use of adaptive machine learning systems in medicine have, for the most part, been confined to discussion of the so-called “update problem,” which concerns how regulators should approach systems whose performance and parameters continue to change even after they have received regulatory (...) approval. In this paper, we draw attention to a prior ethical question: whether the continuous learning that will occur in such systems after their initial deployment should be classified, and regulated, as medical research? We argue that there is a strong prima facie case that the use of continuous learning in medical ML systems should be categorized, and regulated, as research and that individuals whose treatment involves such systems should be treated as research subjects. (shrink)

Intelligibility and interpretability related to artificial intelligence (AI) are crucial for enabling explicability, which is vital for establishing constructive communication and agreement among various stakeholders, including users and designers of AI. It is essential to overcome the challenges of sharing an understanding of the details of the various structures of diverse AI systems, to facilitate effective communication and collaboration. In this paper, we propose four fundamental terms: “I/O,” “Constraints,” “Objectives,” and “Architecture.” These terms help mitigate the challenges associated with intelligibility (...) and interpretability in AI by providing appropriate levels of abstraction to describe structure of AI systems generally, thereby facilitating the sharing of understanding among various stakeholders. The relationship between the Objective of AI designers and the Purpose of AI users is linked to the issues of AI alignment. (shrink)

The ability to interact in a natural language profoundly changes devices’ interfaces and potential applications of speaking technologies. Concurrently, this phenomenon challenges our mainstream theories of knowledge, such as how to analyze linguistic outputs of devices under existing anthropocentric theoretical assumptions. In section 1, I present the topic of machines that speak, connecting between Descartes and Generative AI. In section 2, I argue that accepted testimonial theories of knowledge and justification commonly reject the possibility that a speaking technological artifact can (...) give testimony. In section 3, I identify three assumptions underlying the view that rejects conversational AIs – AI-based technologies that converse, as testifiers: conversational AIs (1) lack intentions, (2) cannot be normatively assessed, and (3) cannot constitute an object in trust relations, while humans can. In section 4, I propose the concept ‘AI-testimony’ for analyzing outputs of conversational AIs, suggesting three conditions for technologies to deliver AI-testimony: (1) content is propositional, (2) generated and delivered with no other human directly involved, (3) the output is perceived as phenomenologically similar to that of a human. I conclude that this concept overcomes the limitations of the anthropocentric concept of testimony, opening future directions of research without associating conversational AIs with human-like agency. (shrink)

For synergistic interactions between humans and artificial intelligence (AI) systems, AI outputs often need to be explainable to people. Explainable AI (XAI) systems are commonly tested in human user studies. However, whether XAI researchers consider potential cultural differences in human explanatory needs remains unexplored. We highlight psychological research that found significant differences in human explanations between many people from Western, commonly individualist countries and people from non-Western, often collectivist countries. We argue that XAI research currently overlooks these variations and that (...) many popular XAI designs implicitly and problematically assume that Western explanatory needs are shared cross-culturally. Additionally, we systematically reviewed over 200 XAI user studies and found that most studies did not consider relevant cultural variations, sampled only Western populations, but drew conclusions about human-XAI interactions more generally. We also analyzed over 30 literature reviews of XAI studies. Most reviews did not mention cultural differences in explanatory needs or flag overly broad cross-cultural extrapolations of XAI user study results. Combined, our analyses provide evidence of a cultural bias toward Western populations in XAI research, highlighting an important knowledge gap regarding how culturally diverse users may respond to widely used XAI systems that future work can and should address. (shrink)

This paper is about the opposite of judgement and calculation. This opposition has been a traditional anchor of critiques concerned with the rise of AI decision making over human judgement. Contrary to these approaches, it is argued that human judgement is not and cannot be replaced by calculation, but that it is human judgement that contextualises computational structures and gives them meaning and purpose. The article focuses on the epistemic structure of algorithms and artificial neural networks to find that they (...) always depend on human judgement to be related to real life objects or purposes. By introducing the philosophical concept of judgement, it becomes clear that the property of judgement to provide meaning and purposiveness is based on the temporality of human life and the ambiguity of language, which quantitative processes lack. A juxtaposition shows that calculations and clustering can be used and referred to in more or less prejudiced and reflecting as well as opaque and transparent ways, but thereby always depend on human judgement. The paper clearly asserts that the transparency of AI is necessary for their autonomous use. This transparency requires the explicitness of the judgements that constitute these computational structures, thereby creating an awareness of the conditionality of such epistemic entities. (shrink)

Artificial intelligent (AI) systems that perform image classification tasks are being used to great success in many application contexts. However, many of these systems are opaque, even to experts. This lack of understanding can be problematic for ethical, legal, or practical reasons. The research field Explainable AI (XAI) has therefore developed several approaches to explain image classifiers. The hope is to bring about understanding, e.g., regarding why certain images are classified as belonging to a particular target class. Most of these (...) approaches use visual explanations. Drawing on Elgin’s work (True enough. MIT Press, Cambridge, 2017), I argue that analyzing what those explanations exemplify can help to assess their suitability for producing understanding. More specifically, I suggest to distinguish between two forms of examples according to their suitability for producing understanding. I call these forms samples and exemplars, respectively. Samples are prone to misinterpretation and thus carry the risk of leading to misunderstanding. Exemplars, by contrast, are intentionally designed or chosen to meet contextual requirements and to mitigate the risk of misinterpretation. They are thus preferable for bringing about understanding. By reviewing several XAI approaches directed at image classifiers, I show that most of them explain with samples. If my analysis is correct, it will be beneficial if such explainability methods use explanations that qualify as exemplars. (shrink)

Recommender systems are artificial intelligence technologies, deployed by online platforms, that model our individual preferences and direct our attention to content we’re likely to engage with. As the digital world has become increasingly saturated with information, we’ve become ever more reliant on these tools to efficiently allocate our attention. And our reliance on algorithmic recommendation may, in turn, reshape us as moral agents. While recommender systems could in principle enhance our moral agency by enabling us to cut through the information (...) saturation of the internet and focus on things that matter, as they’re currently designed and implemented they’re apt to interfere with our ability to attend appropriately to morally relevant factors. In order to analyze the distinctive moral problems algorithmic recommendation poses, we develop a framework for the ethics of attention and an account of judicious attention allocation as a moral skill. We then discuss empirical evidence suggesting that attentional moral skill can be thwarted and undermined in various ways by algorithmic recommendation and related affordances of online platforms, as well as economic and technical considerations that support this concern. Finally, we consider how emerging technologies might overcome the problems we identify. (shrink)

This paper addresses a question about the moral status of Artificial Intelligence (AI): will AIs ever become moral patients? I argue that, while it is in principle possible for an intelligent machine to be a moral patient, there is no good reason to believe this will in fact happen. I start from the plausible assumption that traditional artifacts do not meet a minimal necessary condition of moral patiency: having a good of one's own. I then argue that intelligent machines are (...) no different from traditional artifacts in this respect. To make this argument, I examine the feature of AIs that enables them to improve their intelligence, i.e., machine learning. I argue that there is no reason to believe that future advances in machine learning will take AIs closer to having a good of their own. I thus argue that concerns about the moral status of future AIs are unwarranted. Nothing about the nature of intelligent machines makes them a better candidate for acquiring moral patiency than the traditional artifacts whose moral status does not concern us. (shrink)

The difficulty of explaining the outputs of artificial intelligence (AI) models and what has led to them is a notorious ethical problem wherever these technologies are applied, including in the medical domain, and one that has no obvious solution. This paper examines the proposal, made by Luciano Floridi and colleagues, to include a new ‘principle of explicability’ alongside the traditional four principles of bioethics that make up the theory of ‘principlism’. It specifically responds to a recent set of criticisms that (...) challenge the supposed need for such a principle to perform an enabling role in relation to the traditional four principles and therefore suggest that these four are sufficient without the addition of explicability. The paper challenges the critics’ premise that explicability cannot be an ethical principle like the classic four because it is explicitly subordinate to them. It argues instead that principlism in its original formulation locates the justification for ethical principles in a midlevel position such that they mediate between the most general moral norms and the contextual requirements of medicine. This conception of an ethical principle then provides a mold for an approach to explicability on which it functions as an enabling principle that unifies technical/epistemic demands on AI and the requirements of high-level ethical theories. The paper finishes by anticipating an objection that decision-making by clinicians and AI fall equally, but implausibly, under the principle of explicability’s scope, which it rejects on the grounds that human decisions, unlike AI’s, can be explained by their social environments. (shrink)

In this paper I argue that Artificial Intelligence and the many data science methods associated with it, such as machine learning and large language models, are first and foremost epistemic technologies. In order to establish this claim, I first argue that epistemic technologies can be conceptually and practically distinguished from other technologies in virtue of what they are designed for, what they do and how they do it. I then proceed to show that unlike other kinds of technology (_including_ other (...) epistemic technologies) AI can be uniquely positioned as an epistemic technology in that it is primarily designed, developed and deployed to be used in _epistemic contexts_ such as inquiry, it is specifically designed, developed and deployed to manipulate _epistemic content_ such as data, _and_ it is designed, developed and deployed to do so particularly through _epistemic operations_ such as prediction and analysis. As has been shown in recent work in the philosophy and ethics of AI (Alvarado, AI and Ethics, 2022a), understanding AI as an epistemic technology will also have significant implications for important debates regarding our relationship to AI technologies. This paper includes a brief overview of such implications, particularly those pertaining to explainability, opacity, trust and even epistemic harms related to AI technologies. (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)

This paper addresses the black-box problem in artificial intelligence (AI), and the related problem of explainability of AI in the legal context. We argue, first, that the black box problem is, in fact, a superficial one as it results from an overlap of four different – albeit interconnected – issues: the opacity problem, the strangeness problem, the unpredictability problem, and the justification problem. Thus, we propose a framework for discussing both the black box problem and the explainability of AI. We (...) argue further that contrary to often defended claims the opacity issue is not a genuine problem. We also dismiss the justification problem. Further, we describe the tensions involved in the strangeness and unpredictability problems and suggest some ways to alleviate them. (shrink)

Definition of the problem The umbrella term “explicability” refers to the reduction of opacity of artificial intelligence (AI) systems. These efforts are challenging for medical AI applications because higher accuracy often comes at the cost of increased opacity. This entails ethical tensions because physicians and patients desire to trace how results are produced without compromising the performance of AI systems. The centrality of explicability within the informed consent process for medical AI systems compels an ethical reflection on the trade-offs. Which (...) levels of explicability are needed to obtain informed consent when utilizing medical AI? Arguments We proceed in five steps: First, we map the terms commonly associated with explicability as described in the ethics and computer science literature, i.e., disclosure, intelligibility, interpretability, and explainability. Second, we conduct a conceptual analysis of the ethical requirements for explicability when it comes to informed consent. Third, we distinguish hurdles for explicability in terms of epistemic and explanatory opacity. Fourth, this then allows to conclude the level of explicability physicians must reach and what patients can expect. In a final step, we show how the identified levels of explicability can technically be met from the perspective of computer science. Throughout our work, we take diagnostic AI systems in radiology as an example. Conclusion We determined four levels of explicability that need to be distinguished for ethically defensible informed consent processes and showed how developers of medical AI can technically meet these requirements. (shrink)

This paper investigates how the introduction of AI to decision making increases moral distance and recommends the ethics of care to augment the ethical examination of AI decision making. With AI decision making, face-to-face interactions are minimized, and decisions are part of a more opaque process that humans do not always understand. Within decision-making research, the concept of moral distance is used to explain why individuals behave unethically towards those who are not seen. Moral distance abstracts those who are impacted (...) by the decision and leads to less ethical decisions. The goal of this paper is to identify and analyze the moral distance created by AI through both proximity distance (in space, time, and culture) and bureaucratic distance (derived from hierarchy, complex processes, and principlism). We then propose the ethics of care as a moral framework to analyze the moral implications of AI. The ethics of care brings to the forefront circumstances and context, interdependence, and vulnerability in analyzing algorithmic decision making. (shrink)

The need for fair and just AI is often related to the possibility of understanding AI itself, in other words, of turning an opaque box into a glass box, as inspectable as possible. Transparency and explainability, however, pertain to the technical domain and to philosophy of science, thus leaving the ethics and epistemology of AI largely disconnected. To remedy this, we propose an integrated approach premised on the idea that a glass-box epistemology should explicitly consider how to incorporate values and (...) other normative considerations, such as intersectoral vulnerabilities, at critical stages of the whole process from design and implementation to use and assessment. To connect ethics and epistemology of AI, we perform a double shift of focus. First, we move from trusting the output of an AI system to trusting the process that leads to the outcome. Second, we move from expert assessment to more inclusive assessment strategies, aiming to facilitate expert and non-expert assessment. Together, these two moves yield a framework usable for experts and non-experts when they inquire into relevant epistemological and ethical aspects of AI systems. We dub our framework ‘epistemology-cum-ethics’ to signal the equal importance of both aspects. We develop it from the vantage point of the designers: how to create the conditions to internalize values into the whole process of design, implementation, use, and assessment of an AI system, in which values (epistemic and non-epistemic) are explicitly considered at each stage and inspectable by every salient actor involved at any moment. (shrink)

This paper examines the paradoxical transparency involved in training machine-learning models. Existing literature typically critiques the opacity of machine-learning models such as neural networks or collaborative filtering, a type of critique that parallels the black-box critique in technology studies. Accordingly, people in power may leverage the models’ opacity to justify a biased result without subjecting the technical operations to public scrutiny, in what Dan McQuillan metaphorically depicts as an “algorithmic state of exception”. This paper attempts to differentiate the black-box abstraction (...) that wraps around complex computational systems from the opacity of machine-learning models. It contends that the degree of asymmetry in knowledge is greater in the former than the latter. In the case of software systems, software codes are difficult to understand as only software experts with sufficient domain knowledge are equipped to formulate a sound critique. In contrast, the meanings of trained parameters in a machine-learning model are obscure even to the data scientists who configure and train the model. Hence, the asymmetry of knowledge lies only in how data examples are collected, the choice and configuration of machine-learning models, and the specification of features in model design. Under the trend of algorithmic decision-making proliferating with machine-learning heuristics, the paper contends that the more symmetric distribution of knowledge in machine learning could lead to a more transparent production process if proper policies are in place. (shrink)

”Meaningful human control” is a term invented in the political and legal debate on autonomous weapons system, but it is nowadays also used in many other contexts. It is supposed to specify conditions under which an artificial system is under the right kind of control to avoid responsibility gaps: that is, situations in which no moral agent is responsible. Santoni de Sio and Van den Hoven have recently suggested a framework that can be used by system designers to operationalize this (...) kind of control. It is the purpose of this paper to facilitate further operationalization of ”meaningful human control”. -/- This paper consists of two parts. In the first part I resolve an ambiguity that plagues current operationalizations of MHC. One of the design conditions says that the system should track the reasons of the relevant agents. This condition is ambiguous between the kind of reasons involved. On one interpretation it says that a system should track motivating reasons, while it is concerned with normative reasons on the other. Current participants in the debate interpret the framework as being concerned with (something in the vicinity of) motivating reasons. I argue against this interpretation by showing that meaningful human control requires that a system tracks normative reasons. Moreover, I maintain that an operationalization of meaningful human control that fails to track the right kind of reasons is morally problematic.When this is properly understood, it can be shown that the framework of MHC is committed to the agent-relativity of reasons. More precisely, I argue in the second part of this paper that if the tracking condition of MHC plays an important role in responsibility attribution (as the proponents of the view maintain), then the framework is incompatible with first-order normative theories that hold that normative reasons are agent-neutral (such as many versions of consequentialism). In the final section I present three ways forward for the proponent of MHC as reason-responsiveness. (shrink)

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. (shrink)

Artificial intelligence (AI) systems have demonstrated impressive performance across a variety of clinical tasks. However, notoriously, sometimes these systems are 'black boxes'. The initial response in the literature was a demand for 'explainable AI'. However, recently, several authors have suggested that making AI more explainable or 'interpretable' is likely to be at the cost of the accuracy of these systems and that prioritising interpretability in medical AI may constitute a 'lethal prejudice'. In this paper, we defend the value of interpretability (...) in the context of the use of AI in medicine. Clinicians may prefer interpretable systems over more accurate black boxes, which in turn is sufficient to give designers of AI reason to prefer more interpretable systems in order to ensure that AI is adopted and its benefits realised. Moreover, clinicians may be justified in this preference. Achieving the downstream benefits from AI is critically dependent on how the outputs of these systems are interpreted by physicians and patients. A preference for the use of highly accurate black box AI systems, over less accurate but more interpretable systems , may itself constitute a form of lethal prejudice that may diminish the benefits of AI to - and perhaps even harm - patients. (shrink)

The ‘black box problem’ is a long-standing talking point in debates about artificial intelligence. This is a significant point of tension between ethicists, programmers, clinicians and anyone else working on developing AI for healthcare applications. However, the precise definition of these systems are often left undefined, vague, unclear or are assumed to be standardised within AI circles. This leads to situations where individuals working on AI talk over each other and has been invoked in numerous debates between opaque and explainable (...) systems. This paper proposes a coherent and clear definition for the black box problem to assist in future discussions about AI in healthcare. This is accomplished by synthesising various definitions in the literature and examining several criteria that can be extrapolated from these definitions. There are no data in this work. (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)

In recent years, scholars in the social sciences and humanities have turned their attention to how the rise of digital technologies is reshaping political life in contemporary society. Here, we analyze this issue by distinguishing between two classification technologies typical of pre-digital and digital eras that differently constitute the relationship between individuals and groups. In class-based systems, characteristic of the pre-digital era, one’s status as an individual is gained through membership in a group in which salient social identities are shared (...) in common with other group members. In attribute-based systems, characteristic of the digital era, one’s status as an individual is determined by virtue of possession of a set of attributes that need not be shared with others. We argue that differences between these two types of classification technologies have important implications for how persons attach (or fail to attach) to groups, and therefore what kinds of political mobilization are possible. We illustrate this argument by examining contention over the use of gender as a variable in the pricing of risk in insurance and credit – two markets in which individuals directly encounter class-based and attribute-based systems of classification, respectively. (shrink)

This article presents Epistemic Spatialization as a new framework for investigating the interconnected patterns of biases when identifying objects with convolutional neural networks (convnets). It draws upon Foucault’s notion of spatialized knowledge to guide its method of enquiry. We argue that decisions involved in the creation of algorithms, alongside the labeling, ordering, presentation, and commercial prioritization of objects, together create a distorted “nomination of the visible”: they harden the visibility of some objects, make other objects excessively visible, and consign yet (...) others to permanent or haphazard invisibility. Our approach differs from those who focus on high-stakes misidentifications, such as errors tied to structural racism. Examining the far more dominant series of low-stakes mistakes shows the scope of errors, destabilizing the goal of image content identification with considerable societal impact. We explore these issues by closely examining the demonstration video of a popular convnet. This examination reveals an interlocking series of biases undermining the content identification process. The picture we paint is crucial for a better understanding of the errors that result as these convnets become further embedded in everyday products. The framework is valuable for critical work on computer vision, AI studies, and large-scale visual analysis. (shrink)

Machine learning algorithms are increasingly used to support decision-making in the exercise of public authority. Here, we argue that an important consideration has been overlooked in previous discussions: whether the use of ML undermines the democratic legitimacy of public institutions. From the perspective of democratic legitimacy, it is not enough that ML contributes to efficiency and accuracy in the exercise of public authority, which has so far been the focus in the scholarly literature engaging with these developments. According to one (...) influential theory, exercises of administrative and judicial authority are democratically legitimate if and only if administrative and judicial decisions serve the ends of the democratic law maker, are based on reasons that align with these ends and are accessible to the public. These requirements are not satisfied by decisions determined through ML since such decisions are determined by statistical operations that are opaque in several respects. However, not all ML-based decision support systems pose the same risk, and we argue that a considered judgment on the democratic legitimacy of ML in exercises of public authority need take the complexity of the issue into account. This paper outlines considerations that help guide the assessment of whether a ML undermines democratic legitimacy when used to support public decisions. We argue that two main considerations are pertinent to such normative assessment. The first is the extent to which ML is practiced as intended and the extent to which it replaces decisions that were previously accessible and based on reasons. The second is that uses of ML in exercises of public authority should be embedded in an institutional infrastructure that secures reason giving and accessibility. (shrink)

Studies investigating the relationship between artificial intelligence and science tend to adopt a partial view. There is no broad and holistic view that synthesizes the channels through which this interaction occurs. Our goal is to systematically map the influence of the latest AI techniques on science. We draw on the work of Nathan Rosenberg to develop a taxonomy of the effects of technology on science. The proposed framework comprises four categories of technology effects on science: intellectual, economic, experimental and instrumental. (...) The application of the framework in the relationship between ML/DL and science allowed the identification of multiple triggers activated by the new techniques in the scientific field. Visualizing these different channels of influence allows us to identify two pressing, emerging issues. The first is the concentration of experimental effects in a few companies, which indicates a reinforcement effect between more data on the phenomenon and more capacity to commercialize the technique. The second is the diffusion of new techniques lacking in explanation throughout the fabric of science. The value of this article is twofold. First, it provides a simple framework to assess the relations between technology and science. Second, it provides this broad and holistic view of the influence of new AI techniques on science. More specifically, the article details the channels through which this relationship occurs, the nature of these channels and the loci in which the potential effects on science unfolds. (shrink)

Assistive systems based on Artificial Intelligence (AI) are bound to reshape decision-making in all areas of society. One of the most intricate challenges arising from their implementation in high-stakes environments such as medicine concerns their frequently unsatisfying levels of explainability, especially in the guise of the so-called black-box problem: highly successful models based on deep learning seem to be inherently opaque, resisting comprehensive explanations. This may explain why some scholars claim that research should focus on rendering AI systems understandable, rather (...) than explainable. Yet, there is a grave lack of agreement concerning these terms in much of the literature on AI. We argue that the seminal distinction made by the philosopher and physician Karl Jaspers between different types of explaining and understanding in psychopathology can be used to promote greater conceptual clarity in the context of Machine Learning (ML). Following Jaspers, we claim that explaining and understanding constitute multi-faceted epistemic approaches that should not be seen as mutually exclusive, but rather as complementary ones as in and of themselves they are necessarily limited. Drawing on the famous example of Watson for Oncology we highlight how Jaspers’ methodology translates to the case of medical AI. Classical considerations from the philosophy of psychiatry can therefore inform a debate at the centre of current AI ethics, which in turn may be crucial for a successful implementation of ethically and legally sound AI in medicine. (shrink)

In this 60th anniversary of Business & Society essay, we seek to make three main contributions at the intersection of governance and artificial intelligence. First, we aim to illuminate some of the deeper social, legal, organizational, and democratic challenges of rising AI adoption and resulting algorithmic power by reviewing AI research through a governance lens. Second, we propose an AI governance framework that aims to better assess AI challenges as well as how different governance modalities can support AI. At the (...) heart of our framework lies the governance forces that apply to institutions, organizations, and individuals, who ultimately provide, regulate, and use AI decision-making. We discuss how businesses may harness AI’s economic power through governance solutions without creating or amplifying societal biases and inequalities. Third, as part of our section on future research, we identify a set of governance trade-offs in AI adoption, suggest future research avenues to conceptually strengthen research on the governance of AI, and lay out key policy recommendations. (shrink)

The increasing implementation of and reliance on machine-learning (ML) algorithms to perform tasks, deliver services and make decisions in health and healthcare have made the need for fairness in ML, and more specifically in healthcare ML algorithms (HMLA), a very important and urgent task. However, while the debate on fairness in the ethics of artificial intelligence (AI) and in HMLA has grown significantly over the last decade, the very concept of fairness as an ethical value has not yet been sufficiently (...) explored. Our paper aims to fill this gap and address the AI ethics principle of fairness from a conceptual standpoint, drawing insights from accounts of fairness elaborated in moral philosophy and using them to conceptualise fairness as an ethical value and to redefine fairness in HMLA accordingly. To achieve our goal, following a first section aimed at clarifying the background, methodology and structure of the paper, in the second section, we provide an overview of the discussion of the AI ethics principle of fairness in HMLA and show that the concept of fairness underlying this debate is framed in purely distributive terms and overlaps with non-discrimination, which is defined in turn as the absence of biases. After showing that this framing is inadequate, in the third section, we pursue an ethical inquiry into the concept of fairness and argue that fairness ought to be conceived of as an ethical value. Following a clarification of the relationship between fairness and non-discrimination, we show that the two do not overlap and that fairness requires much more than just non-discrimination. Moreover, we highlight that fairness not only has a distributive but also a socio-relational dimension. Finally, we pinpoint the constitutive components of fairness. In doing so, we base our arguments on a renewed reflection on the concept of respect, which goes beyond the idea of equal respect to include respect for individual persons. In the fourth section, we analyse the implications of our conceptual redefinition of fairness as an ethical value in the discussion of fairness in HMLA. Here, we claim that fairness requires more than non-discrimination and the absence of biases as well as more than just distribution; it needs to ensure that HMLA respects persons both as persons and as particular individuals. Finally, in the fifth section, we sketch some broader implications and show how our inquiry can contribute to making HMLA and, more generally, AI promote the social good and a fairer society. (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)

The increasing use of artificial intelligence (AI) by public actors has led to a push for more transparency. Previous research has conceptualized AI transparency as knowledge that empowers citizens and experts to make informed choices about the use and governance of AI. Conversely, in this paper, we critically examine if transparency-as-knowledge is an appropriate concept for a public realm where private interests intersect with democratic concerns. We conduct a practice-based design research study in which we prototype and evaluate a transparent (...) smart electric vehicle charge point, and investigate experts’ and citizens’ understanding of AI transparency. We find that citizens experience transparency as burdensome; experts hope transparency ensures acceptance, while citizens are mostly indifferent to AI; and with absent means of control, citizens question transparency’s relevance. The tensions we identify suggest transparency cannot be reduced to a product feature, but should be seen as a mediator of debate between experts and citizens. (shrink)

Some artificial intelligence systems can display algorithmic bias, i.e. they may produce outputs that unfairly discriminate against people based on their social identity. Much research on this topic focuses on algorithmic bias that disadvantages people based on their gender or racial identity. The related ethical problems are significant and well known. Algorithmic bias against other aspects of people’s social identity, for instance, their political orientation, remains largely unexplored. This paper argues that algorithmic bias against people’s political orientation can arise in (...) some of the same ways in which algorithmic gender and racial biases emerge. However, it differs importantly from them because there are strong social norms against gender and racial biases. This does not hold to the same extent for political biases. Political biases can thus more powerfully influence people, which increases the chances that these biases become embedded in algorithms and makes algorithmic political biases harder to detect and eradicate than gender and racial biases even though they all can produce similar harm. Since some algorithms can now also easily identify people’s political orientations against their will, these problems are exacerbated. Algorithmic political bias thus raises substantial and distinctive risks that the AI community should be aware of and examine. (shrink)

Technologies that deploy data science methods are liable to result in epistemic harms involving the diminution of individuals with respect to their standing as knowers or their credibility as sources of testimony. Not all harms of this kind are unjust but when they are we ought to try to prevent or correct them. Epistemically unjust harms will typically intersect with other more familiar and well-studied kinds of harm that result from the design, development, and use of data science technologies. However, (...) we argue that epistemic injustices can be distinguished conceptually from more familiar kinds of harm. We argue that epistemic harms are morally relevant even in cases where those who suffer them are unharmed in other ways. Via a series of examples from the criminal justice system, workplace hierarchies, and educational contexts we explain the kinds of epistemic injustice that can result from common uses of data science technologies. (shrink)

Explainable artificial intelligence (XAI) is an emerging, multidisciplinary field of research that seeks to develop methods and tools for making AI systems more explainable or interpretable. XAI researchers increasingly recognise explainability as a context-, audience- and purpose-sensitive phenomenon, rather than a single well-defined property that can be directly measured and optimised. However, since there is currently no overarching definition of explainability, this poses a risk of miscommunication between the many different researchers within this multidisciplinary space. This is the problem we (...) seek to address in this paper. We outline a framework, called Explanatory Pragmatism, which we argue has two attractive features. First, it allows us to conceptualise explainability in explicitly context-, audience- and purpose-relative terms, while retaining a unified underlying definition of explainability. Second, it makes visible any normative disagreements that may underpin conflicting claims about explainability regarding the purposes for which explanations are sought. Third, it allows us to distinguish several dimensions of AI explainability. We illustrate this framework by applying it to a case study involving a machine learning model for predicting whether patients suffering disorders of consciousness were likely to recover consciousness. (shrink)

Journal of Political Philosophy, Volume 30, Issue 2, Page 209-229, June 2022.

Research on the ethics of algorithms has grown substantially over the past decade. Alongside the exponential development and application of machine learning algorithms, new ethical problems and solutions relating to their ubiquitous use in society have been proposed. This article builds on a review of the ethics of algorithms published in 2016, 2016). The goals are to contribute to the debate on the identification and analysis of the ethical implications of algorithms, to provide an updated analysis of epistemic and normative (...) concerns, and to offer actionable guidance for the governance of the design, development and deployment of algorithms. (shrink)

Different people have different perceptions about artificial intelligence (AI). It is extremely important to bring together all the alternative frames of thinking—from the various communities of developers, researchers, business leaders, policymakers, and citizens—to properly start acknowledging AI. This article highlights the ‘fruitful collaboration’ that sociology and AI could develop in both social and technical terms. We discuss how biases and unfairness are among the major challenges to be addressed in such a sociotechnical perspective. First, as intelligent machines reveal their nature (...) of ‘magnifying glasses’ in the automation of existing inequalities, we show how the AI technical community is calling for transparency and explainability, accountability and contestability. Not to be considered as panaceas, they all contribute to ensuring human control in novel practices that include requirement, design and development methodologies for a fairer AI. Second, we elaborate on the mounting attention for technological narratives as technology is recognized as a social practice within a specific institutional context. Not only do narratives reflect organizing visions for society, but they also are a tangible sign of the traditional lines of social, economic, and political inequalities. We conclude with a call for a diverse approach within the AI community and a richer knowledge about narratives as they help in better addressing future technical developments, public debate, and policy. AI practice is interdisciplinary by nature and it will benefit from a socio-technical perspective. (shrink)

The modern abundance and prominence of data has led to the development of “data science” as a new field of enquiry, along with a body of epistemological reflections upon its foundations, methods, and consequences. This article provides a systematic analysis and critical review of significant open problems and debates in the epistemology of data science. We propose a partition of the epistemology of data science into the following five domains: (i) the constitution of data science; (ii) the kind of enquiry (...) that it identifies; (iii) the kinds of knowledge that data science generates; (iv) the nature and epistemological significance of “black box” problems; and (v) the relationship between data science and the philosophy of science more generally. (shrink)