With automation of routine decisions coupled with more intricate and complex information architecture operating this automation, concerns are increasing about the trustworthiness of these systems. These concerns are exacerbated by a class of artificial intelligence that uses deep learning, an algorithmic system of deep neural networks, which on the whole remain opaque or hidden from human comprehension. This situation is commonly referred to as the black box problem in AI. Without understanding how AI reaches its conclusions, it is an open (...) question to what extent we can trust these systems. The question of trust becomes more urgent as we delegate more and more decision-making to and increasingly rely on AI to safeguard significant human goods, such as security, healthcare, and safety. Models that “open the black box” by making the non-linear and complex decision process understandable by human observers are promising solutions to the black box problem in AI but are limited, at least in their current state, in their ability to make these processes less opaque to most observers. A philosophical analysis of trust will show why transparency is a necessary condition for trust and eventually for judging AI to be trustworthy. A more fruitful route for establishing trust in AI is to acknowledge that AI is situated within a socio-technical system that mediates trust, and by increasing the trustworthiness of these systems, we thereby increase trust in AI. (shrink)

We are currently witnessing a sharp rise in the use of algorithmic decision-making tools. In these instances, a new wave of policy concerns is set forth. This article strives to map out these issues, separating the wheat from the chaff. It aims to provide policy makers and scholars with a comprehensive framework for approaching these thorny issues in their various capacities. To achieve this objective, this article focuses its attention on a general analytical framework, which will be applied to a (...) specific subset of the overall discussion. The analytical framework will reduce the discussion to two dimensions, every one of which addressing two central elements. These four factors call for a distinct discussion, which is at times absent in the existing literature. The two dimensions are the specific and novel problems the process assumedly generates and the specific attributes which exacerbate them. While the problems are articulated in a variety of ways, they most likely could be reduced to two broad categories: efficiency and fairness-based concerns. In the context of this discussion, such problems are usually linked to two salient attributes the algorithmic processes feature—its opaque and automated nature. (shrink)

This article considers the issue of opacity as a problem for socially consequential mechanisms of classification and ranking, such as spam filters, credit card fraud detection, search engines, news trends, market segmentation and advertising, insurance or loan qualification, and credit scoring. These mechanisms of classification all frequently rely on computational algorithms, and in many cases on machine learning algorithms to do this work. In this article, I draw a distinction between three forms of opacity: opacity as intentional corporate or state (...) secrecy, opacity as technical illiteracy, and an opacity that arises from the characteristics of machine learning algorithms and the scale required to apply them usefully. The analysis in this article gets inside the algorithms themselves. I cite existing literatures in computer science, known industry practices, and do some testing and manipulation of code as a form of lightweight code audit. I argue that recognizing the distinct forms of opacity that may be coming into play in a given application is a key to determining which of a variety of technical and non-technical solutions could help to prevent harm. (shrink)

Privacy and surveillance scholars increasingly worry that data collectors can use the information they gather about our behaviors, preferences, interests, incomes, and so on to manipulate us. Yet what it means, exactly, to manipulate someone, and how we might systematically distinguish cases of manipulation from other forms of influence—such as persuasion and coercion—has not been thoroughly enough explored in light of the unprecedented capacities that information technologies and digital media enable. In this paper, we develop a definition of manipulation that (...) addresses these enhanced capacities, investigate how information technologies facilitate manipulative practices, and describe the harms—to individuals and to social institutions—that flow from such practices. -/- We use the term “online manipulation” to highlight the particular class of manipulative practices enabled by a broad range of information technologies. We argue that at its core, manipulation is hidden influence—the covert subversion of another person’s decision-making power. We argue that information technology, for a number of reasons, makes engaging in manipulative practices significantly easier, and it makes the effects of such practices potentially more deeply debilitating. And we argue that by subverting another person’s decision-making power, manipulation undermines his or her autonomy. Given that respect for individual autonomy is a bedrock principle of liberal democracy, the threat of online manipulation is a cause for grave concern. (shrink)

We are sceptical of concerns over the opacity of algorithmic decision tools. While transparency and explainability are certainly important desiderata in algorithmic governance, we worry that automated decision-making is being held to an unrealistically high standard, possibly owing to an unrealistically high estimate of the degree of transparency attainable from human decision-makers. In this paper, we review evidence demonstrating that much human decision-making is fraught with transparency problems, show in what respects AI fares little worse or better and argue that (...) at least some regulatory proposals for explainable AI could end up setting the bar higher than is necessary or indeed helpful. The demands of practical reason require the justification of action to be pitched at the level of practical reason. Decision tools that support or supplant practical reasoning should not be expected to aim higher than this. We cast this desideratum in terms of Daniel Dennett’s theory of the “intentional stance” and argue that since the justification of action for human purposes takes the form of intentional stance explanation, the justification of algorithmic decisions should take the same form. In practice, this means that the sorts of explanations for algorithmic decisions that are analogous to intentional stance explanations should be preferred over ones that aim at the architectural innards of a decision tool. (shrink)

For several years, scholars have (for good reason) been largely preoccupied with worries about the use of artificial intelligence and machine learning (AI/ML) tools to make decisions about us. Only recently has significant attention turned to a potentially more alarming problem: the use of AI/ML to influence our decision-making. The contexts in which we make decisions—what behavioral economists call our choice architectures—are increasingly technologically-laden. Which is to say: algorithms increasingly determine, in a wide variety of contexts, both the sets of (...) options we choose from and the way those options are framed. Moreover, artificial intelligence and machine learning (AI/ML) makes it possible for those options and their framings—the choice architectures—to be tailored to the individual chooser. They are constructed based on information collected about our individual preferences, interests, aspirations, and vulnerabilities, with the goal of influencing our decisions. At the same time, because we are habituated to these technologies we pay them little notice. They are, as philosophers of technology put it, transparent to us—effectively invisible. I argue that this invisible layer of technological mediation, which structures and influences our decision-making, renders us deeply susceptible to manipulation. Absent a guarantee that these technologies are not being used to manipulate and exploit, individuals will have little reason to trust them. (shrink)

Decision-making assisted by algorithms developed by machine learning is increasingly determining our lives. Unfortunately, full opacity about the process is the norm. Would transparency contribute to restoring accountability for such systems as is often maintained? Several objections to full transparency are examined: the loss of privacy when datasets become public, the perverse effects of disclosure of the very algorithms themselves, the potential loss of companies’ competitive edge, and the limited gains in answerability to be expected since sophisticated algorithms usually are (...) inherently opaque. It is concluded that, at least presently, full transparency for oversight bodies alone is the only feasible option; extending it to the public at large is normally not advisable. Moreover, it is argued that algorithmic decisions preferably should become more understandable; to that effect, the models of machine learning to be employed should either be interpreted ex post or be interpretable by design ex ante. (shrink)

The ever-increasing application of algorithms to decision-making in a range of social contexts has prompted demands for algorithmic accountability. Accountable decision-makers must provide their decision-subjects with justifications for their automated system’s outputs, but what kinds of broader principles should we expect such justifications to appeal to? Drawing from political philosophy, I present an account of algorithmic accountability in terms of the democratic ideal of ‘public reason’. I argue that situating demands for algorithmic accountability within this justificatory framework enables us to (...) better articulate their purpose and assess the adequacy of efforts toward them. (shrink)

One of the most noticeable trends in recent years has been the increasing reliance of public decision-making processes on algorithms, i.e. computer-programmed step-by-step instructions for taking a given set of inputs and producing an output. The question raised by this article is whether the rise of such algorithmic governance creates problems for the moral or political legitimacy of our public decision-making processes. Ignoring common concerns with data protection and privacy, it is argued that algorithmic governance does pose a significant threat (...) to the legitimacy of such processes. Modelling my argument on Estlund’s threat of epistocracy, I call this the ‘threat of algocracy’. The article clarifies the nature of this threat and addresses two possible solutions. It is argued that neither solution is likely to be successful, at least not without risking many other things we value about social decision-making. The result is a somewhat pessimistic conclusion in which we confront the possibility that we are creating decision-making processes that constrain and limit opportunities for human participation. (shrink)

We explore aversion to the use of algorithms in moral decision-making. So far, this aversion has been explained mainly by the fear of opaque decisions that are potentially biased. Using incentivized experiments, we study which role the desire for human discretion in moral decision-making plays. This seems justified in light of evidence suggesting that people might not doubt the quality of algorithmic decisions, but still reject them. In our first study, we found that people prefer humans with decision-making discretion to (...) algorithms that rigidly apply exogenously given human-created fairness principles to specific cases. In the second study, we found that people do not prefer humans to algorithms because they appreciate flesh-and-blood decision-makers per se, but because they appreciate humans’ freedom to transcend fairness principles at will. Our results contribute to a deeper understanding of algorithm aversion. They indicate that emphasizing the transparency of algorithms that clearly follow fairness principles might not be the only element for fostering societal algorithm acceptance and suggest reconsidering certain features of the decision-making process. (shrink)

In this article, I show why it is necessary to abolish the use of predictive algorithms in the US criminal justice system at sentencing. After presenting the functioning of these algorithms in their context of emergence, I offer three arguments to demonstrate why their abolition is imperative. First, I show that sentencing based on predictive algorithms induces a process of rewriting the temporality of the judged individual, flattening their life into a present inescapably doomed by its past. Second, I demonstrate (...) that recursive processes, comprising predictive algorithms and the decisions based on their predictions, systematically suppress outliers and progressively transform reality to match predictions. In my third and final argument, I show that decisions made on the basis of predictive algorithms actively perform a biopolitical understanding of justice as management and modulation of risks. In such a framework, justice becomes a means to maintain a perverse social homeostasis that systematically exposes disenfranchised Black and Brown populations to risk. (shrink)

Decision-making assisted by algorithms developed by machine learning is increasingly determining our lives. Unfortunately, full opacity about the process is the norm. Would transparency contribute to restoring accountability for such systems as is often maintained? Several objections to full transparency are examined: the loss of privacy when datasets become public, the perverse effects of disclosure of the very algorithms themselves, the potential loss of companies’ competitive edge, and the limited gains in answerability to be expected since sophisticated algorithms usually are (...) inherently opaque. It is concluded that, at least presently, full transparency for oversight bodies alone is the only feasible option; extending it to the public at large is normally not advisable. Moreover, it is argued that algorithmic decisions preferably should become more understandable; to that effect, the models of machine learning to be employed should either be interpreted ex post or be interpretable by design ex ante. (shrink)

Recent work on interpretability in machine learning and AI has focused on the building of simplified models that approximate the true criteria used to make decisions. These models are a useful pedagogical device for teaching trained professionals how to predict what decisions will be made by the complex system, and most importantly how the system might break. However, when considering any such model it’s important to remember Box’s maxim that "All models are wrong but some are useful." We focus on (...) the distinction between these models and explanations in philosophy and sociology. These models can be understood as a "do it yourself kit" for explanations, allowing a practitioner to directly answer "what if questions" or generate contrastive explanations without external assistance. Although a valuable ability, giving these models as explanations appears more difficult than necessary, and other forms of explanation may not have the same trade-offs. We contrast the different schools of thought on what makes an explanation, and suggest that machine learning might benefit from viewing the problem more broadly. (shrink)

This paper aims to provide an overview of surveillance theories and concepts that can help to understand and debate surveillance in its many forms. As scholars from an increasingly wide range of disciplines are discussing surveillance, this literature review can offer much-needed common ground for the debate. We structure surveillance theory in three roughly chronological/thematic phases. The first two conceptualise surveillance through comprehensive theoretical frameworks which are elaborated in the third phase. The first phase, featuring Bentham and Foucault, offers architectural (...) theories of surveillance, where surveillance is often physical and spatial, involving centralised mechanisms of watching over subjects. Panoptic structures function as architectures of power, not only directly but also through disciplining of the watched subjects. The second phase offers infrastructural theories of surveillance, where surveillance is networked and relies primarily on digital rather than physical technologies. It involves distributed forms of watching over people, with increasing distance to the watched and often dealing with data doubles rather than physical persons. Deleuze, Haggerty and Ericson, and Zuboff develop different theoretical frameworks than panopticism to conceptualise the power play involved in networked surveillance. The third phase of scholarship refines, combines or extends the main conceptual frameworks developed earlier. Surveillance theory branches out to conceptualise surveillance through concepts such as dataveillance, access control, social sorting, peer-to-peer surveillance and resistance. With the datafication of society, surveillance combines the physical with the digital, government with corporate surveillance and top-down with self-surveillance. (shrink)

Digital platforms and application software have changed how people work in a range of industries. Empirical studies of the gig economy have raised concerns about new systems of algorithmic management exercised over workers and how these alter the structural conditions of their work. Drawing on the republican literature, we offer a theoretical account of algorithmic domination and a framework for understanding how it can be applied to ride hail and food delivery services in the on-demand economy. We argue that certain (...) algorithms can facilitate new relationships of domination by sustaining a socio-technical system in which the owners and managers of a company dominate workers. This analysis has implications for the growing use of algorithms throughout the gig economy and broader labor market. (shrink)

We argue that artificial networks are explainable and offer a novel theory of interpretability. Two sets of conceptual questions are prominent in theoretical engagements with artificial neural networks, especially in the context of medical artificial intelligence: Are networks explainable, and if so, what does it mean to explain the output of a network? And what does it mean for a network to be interpretable? We argue that accounts of “explanation” tailored specifically to neural networks have ineffectively reinvented the wheel. In (...) response to, we show how four familiar accounts of explanation apply to neural networks as they would to any scientific phenomenon. We diagnose the confusion about explaining neural networks within the machine learning literature as an equivocation on “explainability,” “understandability” and “interpretability.” To remedy this, we distinguish between these notions, and answer by offering a theory and typology of interpretation in machine learning. Interpretation is something one does to an explanation with the aim of producing another, more understandable, explanation. As with explanation, there are various concepts and methods involved in interpretation: Total or Partial, Global or Local, and Approximative or Isomorphic. Our account of “interpretability” is consistent with uses in the machine learning literature, in keeping with the philosophy of explanation and understanding, and pays special attention to medical artificial intelligence systems. (shrink)

Christopher Hood: Transparency in Historical Perspective David Heald: Varieties of Transparency Patrick Birkinshaw: Transparency as a Human Right David Heald: Transparency as an Instrumental Value Onora O'Neill: Transparency and the Ethics of Communication Andrea Prat: The More Closely We Are Watched, the Better We Behave? Alasdair Roberts: Dashed Expectations: Governmental Adaptation to Transparency Rules Andrew McDonald: What Hope Freedom of Information in th UK James Savage: Member State Bedgetary Transparency in the Economic and Monetary Union David Stasavage: Does Transparency Make (...) a Difference? The Example of European Council of Ministers Jean Camp: Varieties of Software and their Implications for Effective Democratic Government Helen Margetts: Transparency and Digital Government Christopher Hood: Conclusion. (shrink)

The debtor-creditor relation, which is at the heart of this book, sharpens mechanisms of exploitation and domination indiscriminately, since, in it, there is no distinction between workers and the unemployed, consumers and producers, working and non-working populations, between retirees and welfare recipients. They are all "debtors," guilty and responsible in the eyes of capital, which has become the Great, the Universal, Creditor.--from The Making of the Indebted Man Debt -- both public debt and private debt Has become a major concern (...) of economic and political leaders. In The Making of the Indebted Man, Maurizio Lazzarato shows that, far from being a threat to the capitalist economy, debt lies at the very core of the neoliberal project. Through a reading of Karl Marx's lesser-known youthful writings on John Mill, and a rereading of writings by Friedrich Nietzsche, Gilles Deleuze, Felix Guattari, and Michel Foucault, Lazzarato demonstrates that debt is above all a political construction, and that the creditor/debtor relation is the fundamental social relation of Western societies.Debt cannot be reduced to a simple economic mechanism, for it is also a technique of "public safety" through which individual and collective subjectivities are governed and controlled. Its aim is to minimize the uncertainty of the time and behavior of the governed. We are forever sinking further into debt to the State, to private insurance, and, on a more general level, to corporations. To insure that we honor our debts, we are at once encouraged and compelled to become the "entrepreneurs" of our lives, of our "human capital." In this way, our entire material, psychological, and affective horizon is upended and reconfigured. How do we extricate ourselves from this impossible situation? How do we escape the neoliberal condition of the indebted man? Lazzarato argues that we will have to recognize that there is no simple technical, economic, or financial solution. We must instead radically challenge the fundamental social relation structuring capitalism: the system of debt. (shrink)

Transparency is an increasingly prominent area of research that offers valuable insights for organizational studies. However, conceptualizations of transparency are rarely subject to critical scrutiny and thus their relevance remains unclear. In most accounts, transparency is associated with the sharing of information and the perceived quality of the information shared. This narrow focus on information and quality, however, overlooks the dynamics of organizational transparency. To provide a more structured conceptualization of organizational transparency, this article unpacks the assumptions that shape the (...) extant literature, with a focus on three dimensions: conceptualizations, conditions, and consequences. The contribution of the study is twofold: On a conceptual level, we provide a framework that articulates two paradigmatic positions underpinning discussions of transparency, verifiability approaches and performativity approaches; on an analytical level, we suggest a novel future research agenda for studying organizational transparency that pays attention to its dynamics, paradoxes, and performative characteristics. (shrink)

This article illustrates ways in which the concepts of the norm and normativity are implicated in relations of power. Specifically, I argue that these concepts have come to function in a normalizing manner. I outline Michel Foucault’s thinking on the norm and normalization and then provide an overview of Jürgen Habermas’s thinking on the norm and normativity in order to show that Habermas’s conceptualizations of the norm and normativity are not, as he posits, necessary foundations for ethics and politics, but (...) in fact simply one philosophical approach among many. Uncritically accepting a Habermasian framework therefore produces normalizing effects and inhibits alternative and potentially emancipatory thinking about ethics and politics. Having problematized the requirement of normative foundations as it is currently articulated, I conclude by examining the emancipatory potential of a particular aspect of Foucault’s work for the practice of philosophy. (shrink)

Mobile health applications (‘health apps’) that promise the user to help her with some aspect of her health are very popular: for-profit apps such as MyFitnessPal, Fitbit, or Headspace have tens of millions of users each. For-profit health apps are designed and run as optimization systems. One would expect that these health apps aim to optimize the health of the user, but in reality they aim to optimize user engagement and, in effect, conversion. This is problematic, I argue, because digital (...) health environments that aim to optimize user engagement risk being manipulative. To develop this argument, I first provide a brief analysis of the underlying business models and the resulting designs of the digital environments provided by popular for-profit health apps. In a second step, I present a concept of manipulation that can help analyze digital environments such as health apps. In the last part of the article, I use my concept of manipulation to analyze the manipulative potential of for-profit health apps. Although for-profit health can certainly empower their users, the conditions for empowerment also largely overlap with the conditions for manipulation. As a result, we should be cautious when embracing the empowerment discourse surrounding health apps. An additional aim of this article is to contribute to the rapidly growing literature on digital choice architectures and the ethics of influencing behavior through such choice architectures. I take health apps to be a paradigmatic example of digital choice architectures that give rise to ethical questions, so my analysis of the manipulative potential of health apps can also inform the larger literature on digital choice architectures. (shrink)

Recent advances in artificial intelligence offer many beneficial prospects. However, concerns have been raised about the opacity of decisions made by these systems, some of which have turned out to be biased in various ways. This article makes a contribution to a growing body of literature on how to make systems for automated decision-making more transparent, explainable, and fair by drawing attention to and further elaborating a distinction first made by Nozick between first-level bias in the application of standards and (...) second-level bias in the choice of standards, as well as a second distinction between discrimination and arbitrariness. Applying the typology developed, a number of illuminating observations are made. First, it is observed that some reported bias in automated decision-making is first-level arbitrariness, which can be alleviated by explainability techniques. However, such techniques have only a limited potential to alleviate first-level discrimination. Second, it is argued that second-level arbitrariness is probably quite common in automated decision-making. In contrast to first-level arbitrariness, however, second-level arbitrariness is not straightforward to detect automatically. Third, the prospects for alleviating arbitrariness are discussed. It is argued that detecting and alleviating second-level arbitrariness is a profound problem because there are many contrasting and sometimes conflicting standards from which to choose, and even when we make intentional efforts to choose standards for good reasons, some second-level arbitrariness remains. (shrink)

Decision-making assisted by algorithms developed by machine learning is increasingly determining our lives. Unfortunately, full opacity about the process is the norm. Would transparency contribute to restoring accountability for such systems as is often maintained? Several objections to full transparency are examined: the loss of privacy when datasets become public, the perverse effects of disclosure of the very algorithms themselves (“gaming the system” in particular), the potential loss of companies’ competitive edge, and the limited gains in answerability to be expected (...) since sophisticated algorithms usually are inherently opaque. It is concluded that, at least presently, full transparency for oversight bodies alone is the only feasible option; extending it to the public at large is normally not advisable. Moreover, it is argued that algorithmic decisions preferably should become more understandable; to that effect, the models of machine learning to be employed should either be interpreted ex post or be interpretable by design ex ante. (shrink)

Big Data are increasingly used in machine learning in order to create predictive models. How are predictive practices that use such models to be situated? In the field of surveillance studies many of its practitioners assert that “governance by discipline” has given way to “governance by risk”. The individual is dissolved into his/her constituent data and no longer addressed. I argue that, on the contrary, in most of the contexts where predictive modelling is used, it constitutes Foucauldian discipline. Compliance to (...) a norm occupies centre stage; suspected deviants are subjected to close attention—as the precursor of possible sanctions. The predictive modelling involved uses personal data from both the focal institution and elsewhere. As a result, the individual re-emerges as the focus of scrutiny. Subsequently, small excursions into Foucauldian texts discuss his discourses on the creation of the “delinquent”, and on the governmental approach to smallpox epidemics. It is shown that his insights only mildly resemble prediction as based on machine learning; several conceptual steps had to be taken for modern machine learning to evolve. Finally, the options available to those subjected to predictive disciplining are discussed: to what extent can they comply, question, or resist? Through a discussion of the concepts of transparency and “gaming the system” I conclude that our predicament is gloomy, in a Kafkaesque fashion. (shrink)

Discipline and Punish analyzes the role of collecting, managing, and operationalizing data in disciplinary institutions. Foucault’s discussion is compared to contemporary forms of surveillance and security practices using algorithmic data processing. The article highlights important similarities and differences regarding the way data processing plays a part in subjectivation. This is also compared to Deleuzian accounts and Foucault’s later discussion in Security, Territory, Population. Using these results, the article argues that the prevailing focus on transparency and accountability in the discussion of (...) algorithmic applications needs to be amended with a perspective on different forms of subjectivity and ensuing power relations. (shrink)