Machine learning algorithms may radically improve our ability to diagnose and treat disease. For moral, legal, and scientific reasons, it is essential that doctors and patients be able to understand and explain the predictions of these models. Scalable, customisable, and ethical solutions can be achieved by working together with relevant stakeholders, including patients, data scientists, and policy makers.

One of the most common contemporary approaches for developing an ethics of artificial intelligence (AI) involves elaborating guiding principles. This essay explores the limitations of this approach, using the history of bioethics as a comparative case. The examples of bioethics and recent AI ethics suggest that principles are difficult to implement in everyday practice, fail to direct individual action, and can frequently result in a pure proceduralism. The essay encourages an additional attention to virtue, which forms the dispositions of actors, (...) directs them to valued ends, and encourages right action in particular situation to complement principles. Practices and professions like medicine or warfighting that already have rich traditions of reflection on virtue ethics will be especially fruitful fields for developing an AI ethics of virtue. (shrink)

This essay introduces this special issue on virtue ethics in relation to military AI. It describes the current situation of military AI ethics as following that of AI ethics in general, caught between consequentialism and deontology. Virtue ethics serves as an alternative that can address some of the weaknesses of these dominant forms of ethics. The essay describes how the articles in the issue exemplify the value of virtue-related approaches for these questions, before ending with thoughts for further research.

Artificial intelligence is increasingly permeating many types of high-stake societal decision-making such as the work at the criminal courts. Various types of algorithmic tools have already been introduced into sentencing. This article concerns the use of algorithms designed to deliver sentence recommendations. More precisely, it is considered how one should determine whether one type of sentencing algorithm (e.g., a model based on machine learning) would be ethically preferable to another type of sentencing algorithm (e.g., a model based on old-fashioned (...) programming). Whether the implementation of sentencing algorithms is ethically desirable obviously depends upon various questions. For instance, some of the traditional issues that have received considerable attention are algorithmic biases and lack of transparency. However, the purpose of this article is to direct attention to a further challenge that has not yet been considered in the discussion of sentencing algorithms. That is, even if is assumed that the traditional challenges concerning biases, transparency, and cost-efficiency have all been solved or proven insubstantial, there will be a further serious challenge associated with the comparison of sentencing algorithms; namely, that we do not yet possess an ethically plausible and applicable criterion for assessing how well sentencing algorithms are performing. (shrink)

In our information societies, tasks and decisions are increasingly outsourced to automated systems, machines, and artificial agents that mediate human relationships, by taking decisions and acting on the basis of algorithms. This raises a critical issue: how are algorithmic procedures and applications to be appraised and governed? This question needs to be investigated, if one wishes to avoid the traps of ICTs ending up in isolating humans behind their screens and digital delegates, or harnessing them in a passive role, (...) by curtailing their freedom and autonomy. (shrink)

The use of black box algorithms in medicine has raised scholarly concerns due to their opaqueness and lack of trustworthiness. Concerns about potential bias, accountability and responsibility, patient autonomy and compromised trust transpire with black box algorithms. These worries connect epistemic concerns with normative issues. In this paper, we outline that black box algorithms are less problematic for epistemic reasons than many scholars seem to believe. By outlining that more transparency in algorithms is not always necessary, (...) and by explaining that computational processes are indeed methodologically opaque to humans, we argue that the reliability of algorithms provides reasons for trusting the outcomes of medical artificial intelligence. To this end, we explain how computational reliabilism, which does not require transparency and supports the reliability of algorithms, justifies the belief that results of medical AI are to be trusted. We also argue that several ethical concerns remain with black box algorithms, even when the results are trustworthy. Having justified knowledge from reliable indicators is, therefore, necessary but not sufficient for normatively justifying physicians to act. This means that deliberation about the results of reliable algorithms is required to find out what is a desirable action. Thus understood, we argue that such challenges should not dismiss the use of black box algorithms altogether but should inform the way in which these algorithms are designed and implemented. When physicians are trained to acquire the necessary skills and expertise, and collaborate with medical informatics and data scientists, black box algorithms can contribute to improving medical care. (shrink)

This essay asks how we might articulate a political theory of algorithms. To do so, I propose a political ontology of the algorithm dispositif that elaborates how algorithms arrange the movement of energies in space and time, and how they do so automatically. This force of arrangement is what I refer to as the dispositional power of algorithms that I identify as a political physics of vital processes. The essay is divided into three sections. The first provides (...) readers of Political Theory with a discussion of three notable works in the field of critical algorithm studies relevant to a political theory of algorithms. The subsequent sections of the essay elaborate an understanding of the political ontology of the algorithm dispositif by focusing (first) on the difference that a virtual ontology introduces to our political reflections and (second) on the cybernetic operation of negative feedback that I identify as foundational to understanding an algorithm’s political physics of vital processes. I conclude that any political theory engagement with technical media can’t simply rest on an epistemic analysis of the normative effects of media but must also pursue an investigation into a medium’s modes of existence in the world. (shrink)

Shane Harwood and Richard Scheines. Learning Linear Causal Structure Equation Models with Genetic Algorithms.

Many actors mobilize the cognitive, legal and technical tool-box of data protection when they discuss and address controversial issues such as digital mass surveillance. Yet, critical approaches to the digital only barely explore the politics of data protection in relation to data-driven governance. Building on governmentality studies and Actor-Network-Theory, this article analyses the potential and limits of using data protection to critique the ‘digital age’. Using the conceptual tool of dispositifs, it sketches an analytics of data protection and the emergence (...) of its configuration as ‘data protection by design and by default’. This exploration reminds us that governing through data implies, first and foremost, governing digital data. (shrink)

Over the years, companies have adopted hiring algorithms because they promise wider job candidate pools, lower recruitment costs and less human bias. Despite these promises, they also bring perils. Using them can inflict unintentional harms on individual human rights. These include the five human rights to work, equality and nondiscrimination, privacy, free expression and free association. Despite the human rights harms of hiring algorithms, the AI ethics literature has predominantly focused on abstract ethical principles. This is problematic for (...) two reasons. First, AI principles have been criticized for being vague and not actionable. Second, the use of vague ethical principles to discuss algorithmic risks does not provide any accountability. This lack of accountability creates an algorithmic accountability gap. Closing this gap is crucial because, without accountability, the use of hiring algorithms can lead to discrimination and unequal access to employment opportunities. This paper makes two contributions to the AI ethics literature. First, it frames the ethical risks of hiring algorithms using international human rights law as a universal standard for determining algorithmic accountability. Second, it evaluates four types of algorithmic impact assessments in terms of how effectively they address the five human rights of job applicants implicated in hiring algorithms. It determines which of the assessments can help companies audit their hiring algorithms and close the algorithmic accountability gap. (shrink)

A number of artificial intelligence systems have been proposed to assist users in identifying the issues of algorithmic fairness and transparency. These AI systems use diverse bias detection methods from various perspectives, including exploratory cues, interpretable tools, and revealing algorithms. This study explains the design of AI systems by probing how users make sense of fairness and transparency as they are hypothetical in nature, with no specific ways for evaluation. Focusing on individual perceptions of fairness and transparency, this study (...) examines the roles of normative values in over-the-top platforms by empirically testing their effects on sensemaking processes. A mixed-method design incorporating both qualitative and quantitative approaches was used to discover user heuristics and to test the effects of such normative values on user acceptance. Collectively, a composite concept of transparent fairness emerged around user sensemaking processes and its formative roles regarding their underlying relations to perceived quality and credibility. From a sensemaking perspective, this study discusses the implications of transparent fairness in algorithmic media platforms by clarifying how and what should be done to make algorithmic media more trustable and reliable platforms. Based on the findings, a theoretical model is developed to define transparent fairness as an essential algorithmic attribute in the context of OTT platforms. (shrink)

The value-ladenness of computer algorithms is typically framed around issues of epistemic risk. In this article, I examine a deeper sense of value-ladenness: algorithmic methods are not only themselves value-laden but also introduce value into how we reason about their domain of application. I call this domain distortion. In particular, using insights from jurisprudence, I show that the use of recidivism risk assessment algorithms presupposes legal formalism and blurs the distinction between liability assessment and sentencing, which distorts how (...) the domain of criminal punishment is conceived and provides a distinctive avenue for values to enter the legal process. (shrink)

We examine the performance of some standard causal discovery algorithms, both constraint-based and score-based, from the perspective of how robust they are against failures of the Causal Faithfulness Assumption. For this purpose, we make only the so-called Triangle-Faithfulness assumption, which is a fairly weak consequence of the Faithfulness assumption, and otherwise allows unfaithful distributions. In particular, we allow violations of Adjacency-Faithfulness and Orientation-Faithfulness. We show that the PC algorithm, a representative constraint-based method, can be made more robust against unfaithfulness (...) by incorporating elements of the GES algorithm, a representative score-based method; similarly, the GES algorithm can be made less error-prone by incorporating elements of the conservative PC algorithm. As our simulations demonstrate, the increased robustness seems to matter even when faithfulness is not exactly violated, for with only finite sample, distributions that are not exactly unfaithful may be sufficiently close to being unfaithful to make trouble. (shrink)

In the digital age, the use of advanced technology is becoming a new paradigm in police work, criminal justice, and the penal system. Algorithms promise to predict delinquent behaviour, identify potentially dangerous persons, and support crime investigation. Algorithm-based applications are often deployed in this context, laying the groundwork for a ‘smart criminal justice’. In this qualitative study based on 32 interviews with criminal justice and police officials, we explore the reasons why and extent to which such a smart criminal (...) justice system has already been established in Switzerland, and the benefits perceived by users. Drawing upon this research, we address the spread, application, technical background, institutional implementation, and psychological aspects of the use of algorithms in the criminal justice system. We find that the Swiss criminal justice system is already significantly shaped by algorithms, a change motivated by political expectations and demands for efficiency. Until now, algorithms have only been used at a low level of automation and technical complexity and the levels of benefit perceived vary. This study also identifies the need for critical evaluation and research-based optimization of the implementation of advanced technology. Societal implications, as well as the legal foundations of the use of algorithms, are often insufficiently taken into account. By discussing the main challenges to and issues with algorithm use in this field, this work lays the foundation for further research and debate regarding how to guarantee that ‘smart’ criminal justice is actually carried out smartly. (shrink)

Mind and environment evolve in tandem—almost a platitude. Much of judgment and decision making research, however, has compared cognition to standard statistical models, rather than to how well it is adapted to its environment. The author argues two points. First, cognitive algorithms are tuned to certain information formats, most likely to those that humans have encountered during their evolutionary history. In par ticular, Bayesian computations are simpler when the information is in a frequency format than when it is in (...) a probability format. The author investigates whether fre quency formats can make physicians reason more often the Bayesian way. Second, cognitive algorithms need to operate under constraints of limited time, knowledge, and computational power, and they need to exploit the structures of their environments. The author describes a fast and frugal algorithm, Take The Best, that violates standard principles of rational inference but can be as accurate as sophisticated "optimal" mod els for diagnostic inference. Key words: Bayes' theorem; bounded rationality; infor mation format; probabilistic reasoning; satisficing; training; medical education. (shrink)

In a legendary technical report, the Google founders sketched a wisdom-of-crowds justification for PageRank arguing that the algorithm, by aggregating incoming links to webpages in a sophisticated way, tracks importance on the web. On this reading of the report, webpages that have a high impact as measured by PageRank are supposed to be important webpages in a sense of importance that is not reducible to mere impact or popularity. In this paper, we look at the state of the art regarding (...) the more precise statement of the thesis that PageRank and other similar in-link-based ranking algorithms can be justified by reference to the wisdom of crowds. We argue that neither the influential preferential attachment models due to Barabási and Albert in nor the recent model introduced by Masterton et al. in allows for a satisfactory wisdom-of-crowds justification of PageRank. As a remedy, we suggest that future work should explore “dual models” of linking on the web, i.e., models that combine the two previous approaches. Dual models view links as being attracted to both popularity and importance. (shrink)

Stochasticity and ambiguity are two aspects of uncertainty in economic problems. In the case of investments in risky assets, this uncertainty is manifested in the uncertainty of future returns. On the contrary, the complexity of the economic phenomenon itself and the ambiguity inherent in human thinking and judgment are characterized by indistinct boundaries. For the same problem, research from different perspectives can often provide us with more comprehensive and systematic information. Currently, the expected value of return or the variance representing (...) risk is still used as a rational investment criterion for both single-stage portfolios and multistage portfolios. However, in general, the greater the expected return of an investor, the greater the risk he should take. Different investors have different requirements for profitability, but regardless of their expected return, they always hope to find a set of portfolios that maximize the probability of achieving the expected rate of return. In this paper, after analyzing the development of portfolio investment theory research, we take fuzzy information processing as the entry point and systematically discuss the theory and methods of fuzzy modeling of portfolio investment decision-making from the perspective of fuzziness around the portfolio investment decision-making process. The results of the empirical analysis show that the existence of basis constraints affects investors’ investment strategies as well as their final returns, but there is a limit to the influence of basis constraints on portfolio performance, and investors can obtain optimal investment returns by selecting a reasonable number of securities to form a portfolio based on the characteristics of different securities. (shrink)

The article examines the algorithms for JPEG and JPEG-2000 compression of various graphic images. The main steps of the operation of both algorithms are given, their advantages and disadvantages are noted. The main differences between JPEG and JPEG-2000 are analyzed. It is noted that the JPEG-2000 algorithm allows re-moving visually unpleasant effects. This makes it possible to highlight important areas of the image and improve the quality of their compression. The features of each step of the algorithms (...) are considered and the difficulties of their implementation are compared. The effectiveness of each algorithm is demonstrated by the example of a full-color image of the BSU emblem. The obtained compression ratios were obtained and shown in the corresponding tables using both algorithms. Compression ratios are obtained for a wide range of quality values from 1 to ten. We studied various types of images: black and white, business graphics, indexed and full color. A modified LZW-Lempel-Ziv-Welch algorithm is presented, which is applicable to compress a variety of information from text to images. The modification is based on limiting the graphic file to 256 colors. This made it possible to index the color with one byte instead of three. The efficiency of this modification grows with increasing image sizes. The modified LZW-algorithm can be adapted to any image from single-color to full-color. The prepared tests were indexed to the required number of colors in the images using the FastStone Image Viewer program. For each image, seven copies were obtained, containing 4, 8, 16, 32, 64, 128 and 256 colors, respectively. Testing results showed that the modified version of the LZW algorithm allows for an average of twice the compression ratio. However, in a class of full-color images, both algorithms showed the same results. The developed modification of the LZW algorithm can be successfully applied in the field of site design, especially in the case of so-called flat design. The comparative characteristics of the basic and modified methods are presented. (shrink)

Search engines are important contemporary sources of information and contribute to shaping our beliefs about the world. Each time they are consulted, various algorithms filter and order content to show us relevant results for the inputted search query. Because these search engines are frequently and widely consulted, it is necessary to have a clear understanding of the distinctively epistemic role that these algorithms play in the background of our online experiences. To aid in such understanding, this paper argues (...) that search engine algorithms are providers of “bent testimony”—that, within certain contexts of interactions, users act as if these algorithms provide us with testimony—and acquire or alter beliefs on that basis. Specifically, we treat search engine algorithms as if they were asserting as true the content ordered at the top of a search results page—which has interesting parallels with how we might treat an ordinary testifier. As such, existing discussions in the philosophy of testimony can help us better understand and, in turn, improve our interactions with search engines. By explicating the mechanisms by which we come to accept this “bent testimony,” our paper discusses methods to help us control our epistemic reliance on search engine algorithms and clarifies the normative expectations one ought to place on the search engines that deploy these algorithms. (shrink)

In the ethics of algorithms, a specifically epistemological analysis is rarely undertaken in order to gain a critique (or a defense) of the handling of or trust in medical black box algorithms (BBAs). This article aims to begin to fill this research gap. Specifically, the thesis is examined according to which such algorithms are regarded as epistemic authorities (EAs) and that the results of a medical algorithm must completely replace other convictions that patients have (preemptionism). If this (...) were true, it would be a reason to distrust medical BBAs. First, the author describes what EAs are and why BBAs can be considered EAs. Then, preemptionism will be outlined and criticized as an answer to the question of how to deal with an EA. The discussion leads to some requirements for dealing with a BBA as an EA. (shrink)

Episodic memory is memory for past events. It’s characteristically associated with an experience of ‘mentally replaying’ one’s experiences in the mind’s eye. This biological phenomenon has inspired the development of several ‘experience replay’ algorithms in AI. In this chapter, I ask whether experience replay algorithms might shed light on a puzzle about episodic memory’s function: what does episodic memory contribute to the cognitive systems in which it is found? I argue that experience replay algorithms can serve as (...) idealized models of episodic memory for the purposes of addressing this question. Taking the DQN algorithm as a case study, I suggest that these algorithms provide some support for mnemonic accounts, on which episodic memory’s function lies in the storage, encoding and retrieval of information. By extending and adapting experience replay algorithms, we might gain further insight into episodic memory’s operations and contributions to cognition. (shrink)

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

Purpose Existing algorithms for predicting suicide risk rely solely on data from electronic health records, but such models could be improved through the incorporation of publicly available socioeconomic data – such as financial, legal, life event and sociodemographic data. The purpose of this study is to understand the complex ethical and privacy implications of incorporating sociodemographic data within the health context. This paper presents results from a survey exploring what the general public’s knowledge and concerns are about such publicly (...) available data and the appropriateness of using it in suicide risk prediction algorithms. Design/methodology/approach A survey was developed to measure public opinion about privacy concerns with using socioeconomic data across different contexts. This paper presented respondents with multiple vignettes that described scenarios situated in medical, private business and social media contexts, and asked participants to rate their level of concern over the context and what factor contributed most to their level of concern. Specific to suicide prediction, this paper presented respondents with various data attributes that could potentially be used in the context of a suicide risk algorithm and asked participants to rate how concerned they would be if each attribute was used for this purpose. Findings The authors found considerable concern across the various contexts represented in their vignettes, with greatest concern in vignettes that focused on the use of personal information within the medical context. Specific to the question of incorporating socioeconomic data within suicide risk prediction models, the results of this study show a clear concern from all participants in data attributes related to income, crime and court records, and assets. Data about one’s household were also particularly concerns for the respondents, suggesting that even if one might be comfortable with their own being used for risk modeling, data about other household members is more problematic. Originality/value Previous studies on the privacy concerns that arise when integrating data pertaining to various contexts of people’s lives into algorithmic and related computational models have approached these questions from individual contexts. This study differs in that it captured the variation in privacy concerns across multiple contexts. Also, this study specifically assessed the ethical concerns related to a suicide prediction model and determining people’s awareness of the publicness of select data attributes, as well as which of these data attributes generated the most concern in such a context. To the best of the authors’ knowledge, this is the first study to pursue this question. (shrink)

We introduce a fairness criterion that we call Spanning. Spanning i) is implied by Calibration, ii) retains interesting properties of Calibration that some other ways of relaxing that criterion do not, and iii) unlike Calibration and other prominent ways of weakening it, is consistent with Equalized Odds outside of trivial cases.

Social media is a pervasive part of everyday life. That is, new media occupies more and more spaces in individuals’ lives both in intimate and work sphere. In addition, due to convergence, new media brought together interpersonal and mass communications in the same environment. This fact has caused a wide range of changes in cultural industries. One of the main changes brought about by social media in relation to the mass media is the construction of a flow of content, advertising (...) and propaganda customized for each individual, and constructed from surveillance and control of individuals’ interactions in digital networks. For so doing, one element is central: Algorithms. It is not only by means of contents produced by cultural workers or amateurs that social media guarantees the life world colonization by the system, but mainly through the tracking of user interactions. That is possible thanks to the data gathering performed by algorithms. Therefore, social networks colonize the life world in a more constant and pervasive way than mass media, facilitating surveillance and social control that are vital for both, the digital conglomerates economic power and the state capacity to watch individuals, either to strengthen the sales effort or to ensure citizen vigilance. (shrink)

This article draws on contributions from the Sociology of Science and Technology and Science and Technology Studies, the Sociology of Risk and Uncertainty, and the Sociology of Work, focusing on the transformations of employment regarding expanded automation, robotization and informatization. The new work patterns emerging due to the introduction of software and hardware technologies, which are based on artificial intelligence, algorithms, big data gathering and robotic systems are examined closely. This article attempts to “open the black boxes” of the (...) “black carpet” (robotic sorting system) and examine the reorganization of Greek postal services through the introduction of software and hardware technologies, highlighting a) the high risk of flexible, pluralistic, decentralized (under)employment and aspects of the sub-politics of automation, b) the new forms of work organization and c) aspects of labor precariousness as a result of de-specialization or “flexible expertise” interconnected with new forms of inequalities based on AI, algorithms, big data and robots as reflected in the lived experiences of workers. The above are investigated through primary empirical social research in postal service companies in Greece. (shrink)

Abstract This chapter is interested in the epistemology of algorithms. As I intend to approach the topic, this is an issue about epistemic justification. Current approaches to justification emphasize the transparency of algorithms, which entails elucidating their internal mechanisms –such as functions and variables– and demonstrating how (or that) these produce outputs. Thus, the mode of justification through transparency is contingent on what can be shown about the algorithm and, in this sense, is internal to the algorithm. In (...) contrast, I advocate for an externalist epistemology of algorithms that I term computational reliabilism (CR). While I have previously introduced and examined CR in the field of computer simulations ([42, 53, 4]), this chapter extends this reliabilist epistemology to encompass a broader spectrum of algorithms utilized in various scientific disciplines, with a particular emphasis on machine learning applications. At its core, CR posits that an algorithm’s output is justified if it is produced by a reliable algorithm. A reliable algorithm is one that has been specified, coded, used, and maintained utilizing reliability indicators. These reliability indicators stem from formal methods, algorithmic metrics, expert competencies, cultures of research, and other scientific endeavors. The primary aim of this chapter is to delineate the foundations of CR, explicate its operational mechanisms, and outline its potential as an externalist epistemology of algorithms. (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)

Overview • Alloy peculiarity • Alloy utilities • Assignments and pre- and postconditions in Alloy • Alloy for automated logical reasoning • Alloy specifications of algorithms • On your to do list: – Look through the example code in these slides, – make sure you understand what is happening. Note: Alloy Peculiarity..

Proceedings of the Steklov Institute of Mathematics is a cover-to-cover translation of the Trudy Matematicheskogo Instituta imeni V.A. Steklova of the Russian Academy of Sciences. Each issue ordinarily contains either one book-length article or a collection of articles pertaining to the same topic.

This is the first of two papers that use off-training set (OTS) error to investigate the assumption-free relationship between learning algorithms. This first paper discusses the senses in which there are no a priori distinctions between learning algorithms. (The second paper discusses the senses in which there are such distinctions.) In this first paper it is shown, loosely speaking, that for any two algorithms A and B, there are “as many” targets (or priors over targets) for which (...) A has lower expected OTS error than B as vice versa, for loss functions like zero-one loss. In particular, this is true if A is cross-validation and B is “anti-cross-validation” (choose the learning algorithm with largest cross-validation error). This paper ends with a discussion of the implications of these results for computational learning theory. It is shown that one cannot say: if empirical misclassification rate is low, the Vapnik-Chervonenkis dimension of your generalizer is small, and the training set is large, then with high probability your OTS error is small. Other implications for “membership queries” algorithms and “punting” algorithms are also discussed. (shrink)

AI-based algorithms are used extensively by public institutions. Thus, for instance, AI algorithms have been used in making decisions concerning punishment providing welfare payments, making decisions concerning parole, and many other tasks which have traditionally been assigned to public officials and/or public entities. We develop a novel argument against the use of AI algorithms, in particular with respect to decisions made by public officials and public entities. We argue that decisions made by AI algorithms cannot count (...) as public decisions, namely decisions that are made in the name of citizens and that this fact should be taken into consideration when utilizing AI to replace public officials. (shrink)

The paper deals with the quandary of the neutrality and transparency of technologies. First, I show how this problem is connected with the image of the opening of 'black boxes' that is pivotal to much of science and technology studies. Second, methodological and socio-political dimensions of the 'black box' metaphor are discussed. Third, I analyze three typical solutions to the problem of the neutrality of technologies outside and inside constructivist technology studies. It is demonstrated that despite their apparent differences, these (...) solutions are similar in their logic of conceptualizing technology as a neutral intermediary. Forth, I look for an alternative to this logic in the actor-network theory of Bruno Latour. Here technologies are conceived in terms of an eventful association of heterogeneous entities irreducible to its conditions of possibility. The construction of technologies is understood as mediation, or as a 'making-do' process where creators are surprised by their creations and vice versa. In Latour's actor-network, technologies are interpreted as opaque and non-neutral entities. Finally, I turn to some object-lessons from smart technologies powered by neural networks to demonstrate that these are empirical vindications of Latour's conception of technical mediation. Particular attention is paid to the opacity and (non)interpretability of machine learning algorithms. (shrink)

The original design of a software system is rarely prepared for every new requirement. Software systems should be updated frequently, which is usually accompanied by the decline in software modularity and quality. Although many approaches have been proposed to improve the quality of software, a majority of them are guided by metrics defined on the local properties of software. In this article, we propose to use a global metric borrowed from the network science to detect the moving method refactoring. First, (...) our approach uses a bipartite network to represent classes, features, and their dependencies. Second, a new metric is introduced to quantify the modularity of a software system as a whole. Finally, a crossover-only evolutionary algorithm that uses the metric as its fitness function is introduced to optimize the class structure of a software system and detect the methods that should be moved. Empirical results on the benchmark Java projects show that our approach can find meaningful methods that should be moved with a high stability. The advantages of our approach are illustrated in comparison with some other approaches, specifically one refactoring approach, namely search-based refactoring approach, and two community detection algorithms, namely a graph theoretic clustering algorithm and a fast algorithm for community detection. Our approach provides a new way to do refactoring from the perspective of software structure. (shrink)

In contrast to traditional research on base-rate neglect, an ecologically-oriented research program would analyze the correspondence between cognitive algorithms and the nature of information in the environment. Bayesian computations turn out to be simpler when information is represented in frequency formats as opposed to the probability formats used in previous research. Frequency formats often enable even uninstructed subjects to perform Bayesian reasoning.

In order to improve the efficiency of the English translation, machine translation is gradually and widely used. This study briefly introduces the neural network algorithm for speech recognition. Long short-term memory (LSTM), instead of traditional recurrent neural network (RNN), was used as the encoding algorithm for the encoder, and RNN as the decoding algorithm for the decoder. Then, simulation experiments were carried out on the machine translation algorithm, and it was compared with two other machine translation algorithms. The results (...) showed that the back-propagation (BP) neural network had a lower word error rate and spent less recognition time than artificial recognition in recognizing the speech; the LSTM–RNN algorithm had a lower word error rate than BP–RNN and RNN–RNN algorithms in recognizing the test samples. In the actual speech translation test, as the length of speech increased, the LSTM–RNN algorithm had the least changes in the translation score and word error rate, and it had the highest translation score and the lowest word error rate under the same speech length. (shrink)