Search

For years, the number of opaque algorithmic decision-making systems (ADM systems) with a large impact on society has been increasing: e.g., systems that compute decisions about future recidivism of criminals, credit worthiness, or the many small decision computing systems within social networks that create rankings, provide recommendations, or filter content. Concerns that such a system makes biased decisions can be difficult to investigate: be it by people affected, NGOs, stakeholders, governmental testing and auditing authorities, or other external parties. Scientific testing and auditing literature rarely focuses on the specific needs for such investigations and suffers from ambiguous terminologies. With this paper, we aim to support this investigation process by collecting, explaining, and categorizing methods of testing for bias, which are applicable to black-box systems, given that inputs and respective outputs can be observed. For this purpose, we provide a taxonomy that can be used to select suitable test methods adapted to the respective situation. This taxonomy takes multiple aspects into account, for example the effort to implement a given test method, its technical requirement (such as the need of ground truth) and social constraints of the investigation, e.g., the protection of business secrets. Furthermore, we analyze which test method can be used in the context of which black box audit concept. It turns out that various factors, such as the type of black box audit or the lack of an oracle, may limit the selection of applicable tests. With the help of this paper, people or organizations who want to test an ADM system for bias can identify which test methods and auditing concepts are applicable and what implications they entail. [ABSTRACT FROM AUTHOR]

This paper provides an empirical and conceptual account on seeing machine learning models as part of a sociotechnical system to identify relevant vulnerabilities emerging in the context of use. As ML is increasingly adopted in socially sensitive and safety-critical domains, many ML applications end up not delivering on their promises, and contributing to new forms of algorithmic harm. There is still a lack of empirical insights as well as conceptual tools and frameworks to properly understand and design for the impact of ML models in their sociotechnical context. In this paper, we follow a design science research approach to work towards such insights and tools. We center our study in the financial industry, where we first empirically map recently emerging MLOps practices to govern ML applications, and corroborate our insights with recent literature. We then perform an integrative literature research to identify a long list of vulnerabilities that emerge in the sociotechnical context of ML applications, and we theorize these along eight dimensions. We then perform semi-structured interviews in two real-world use cases and across a broad set of relevant actors and organizations, to validate the conceptual dimensions and identify challenges to address sociotechnical vulnerabilities in the design and governance of ML-based systems. The paper proposes a set of guidelines to proactively and integrally address both the dimensions of sociotechnical vulnerability, as well as the challenges identified in the empirical use case research, in the organization of MLOps practices. [ABSTRACT FROM AUTHOR]

In recent years, there has been a surge in research addressing the question which properties predictive algorithms ought to satisfy in order to be considered fair. Three of the most widely discussed criteria of fairness are the criteria called equalized odds, predictive parity, and counterfactual fairness. In this paper, I will present a new impossibility result involving these three criteria of algorithmic fairness. In particular, I will argue that there are realistic circumstances under which any predictive algorithm that satisfies counterfactual fairness will violate both other fairness criteria, that is, equalized odds and predictive parity. As will be shown, this impossibility result forces us to give up one of four intuitively plausible assumptions about algorithmic fairness. I will explain and motivate each of the four assumptions and discuss which of them can plausibly be given up in order to circumvent the impossibility. [ABSTRACT FROM AUTHOR]

As scientists start to adopt machine learning (ML) as one research tool, the security of ML and the knowledge generated become a concern. In this paper, I explain how supervised ML can be improved with better data ontology, or the way we make categories and turn information into data. More specifically, we should design data ontology in such a way that is consistent with the knowledge that we have about the target phenomenon so that such ontology can help us make the inductive leap. I do so by thinking through a thought experiment, Goodman's New Riddle of Induction (Fact, fiction, and forecast, Harvard University Press, 1955). Goodman's riddle helps flesh out three problems of induction: (1) the problem of equal goodies, that there are often too many equally good inductive results given the same data; (2) the problem of diverging performance, that these equally good results can give opposite predictions in the future; and (3) the problem of mediocrity, that when averaged across all equally possible datasets and tasks, no inductive algorithm outperforms any other. I show that all these three problems are manifested as real obstacles in ML practice, namely, the Rashomon effect (Breiman in Stat Sci 16(3):199–231, 2001), the problem of underspecification (D'Amour et al. in J Mach Learn Res, 2020, https://doi.org/10.48550/arXiv.2011.03395), and the No Free Lunch theorem (Wolpert in Neural Comput 8(7):1341–90, 1996, https://doi.org/10.1162/neco.1996.8.7.1341). Lastly, I argue that proper data ontology can help mitigate these problems and I demonstrate how using concrete examples from climate science. This research highlights the links between philosophers' discussions of induction and implications in ML practice. [ABSTRACT FROM AUTHOR]

Models developed using machine learning are increasingly prevalent in scientific research. At the same time, these models are notoriously opaque. Explainable AI aims to mitigate the impact of opacity by rendering opaque models transparent. More than being just the solution to a problem, however, Explainable AI can also play an invaluable role in scientific exploration. This paper describes how post-hoc analytic techniques from Explainable AI can be used to refine target phenomena in medical science, to identify starting points for future investigations of (potentially) causal relationships, and to generate possible explanations of target phenomena in cognitive science. In this way, this paper describes how Explainable AI—over and above machine learning itself—contributes to the efficiency and scope of data-driven scientific research. [ABSTRACT FROM AUTHOR]

In this paper we make the case for the emergence of novel kind of bias with the use of algorithmic decision-making systems. We argue that the distinctive generative process of feature creation, characteristic of machine learning (ML), contorts feature parameters in ways that can lead to emerging feature spaces that encode novel algorithmic bias involving already marginalized groups. We term this bias assembled bias. Moreover, assembled biases are distinct from the much-discussed algorithmic bias, both in source (training data versus feature creation) and in content (mimics of extant societal bias versus reconfigured categories). As such, this problem is distinct from issues arising from bias-encoding training feature sets or proxy features. Assembled bias is not epistemically transparent in source or content. Hence, when these ML models are used as a basis for decision-making in social contexts, algorithmic fairness concerns are compounded. [ABSTRACT FROM AUTHOR]

This paper looks at philosophical questions that arise in the context of AI alignment. It defends three propositions. First, normative and technical aspects of the AI alignment problem are interrelated, creating space for productive engagement between people working in both domains. Second, it is important to be clear about the goal of alignment. There are significant differences between AI that aligns with instructions, intentions, revealed preferences, ideal preferences, interests and values. A principle-based approach to AI alignment, which combines these elements in a systematic way, has considerable advantages in this context. Third, the central challenge for theorists is not to identify 'true' moral principles for AI; rather, it is to identify fair principles for alignment that receive reflective endorsement despite widespread variation in people's moral beliefs. The final part of the paper explores three ways in which fair principles for AI alignment could potentially be identified. [ABSTRACT FROM AUTHOR]

Here, as in other ML applications, there is a well-known problem of balancing the fit of the ML model to the training data with a possible overfitting to that very data; something which is generally tackled by so called I regularization i methods. Machine Learning (ML) techniques are currently revolutionizing many areas of science. The authors thus compare ML models to standard statistical modeling and highlight what is missing in the case of ML models. As Lam argues in this paper, the implicit use of high-level concepts by ML systems can be seen as an example of tacit knowledge, and hence he argues for the existence of propositional content in ML black box systems. [Extracted from the article]

Necessity and sufficiency are the building blocks of all successful explanations. Yet despite their importance, these notions have been conceptually underdeveloped and inconsistently applied in explainable artificial intelligence (XAI), a fast-growing research area that is so far lacking in firm theoretical foundations. In this article, an expanded version of a paper originally presented at the 37th Conference on Uncertainty in Artificial Intelligence (Watson et al., 2021), we attempt to fill this gap. Building on work in logic, probability, and causality, we establish the central role of necessity and sufficiency in XAI, unifying seemingly disparate methods in a single formal framework. We propose a novel formulation of these concepts, and demonstrate its advantages over leading alternatives. We present a sound and complete algorithm for computing explanatory factors with respect to a given context and set of agentive preferences, allowing users to identify necessary and sufficient conditions for desired outcomes at minimal cost. Experiments on real and simulated data confirm our method's competitive performance against state of the art XAI tools on a diverse array of tasks. [ABSTRACT FROM AUTHOR]

Machine behavior that is based on learning algorithms can be significantly influenced by the exposure to data of different qualities. Up to now, those qualities are solely measured in technical terms, but not in ethical ones, despite the significant role of training and annotation data in supervised machine learning. This is the first study to fill this gap by describing new dimensions of data quality for supervised machine learning applications. Based on the rationale that different social and psychological backgrounds of individuals correlate in practice with different modes of human–computer-interaction, the paper describes from an ethical perspective how varying qualities of behavioral data that individuals leave behind while using digital technologies have socially relevant ramification for the development of machine learning applications. The specific objective of this study is to describe how training data can be selected according to ethical assessments of the behavior it originates from, establishing an innovative filter regime to transition from the big data rationale n = all to a more selective way of processing data for training sets in machine learning. The overarching aim of this research is to promote methods for achieving beneficial machine learning applications that could be widely useful for industry as well as academia. [ABSTRACT FROM AUTHOR]

Machine learning-based AI algorithms lack transparency. In this article, I offer an interpretation of AI's explainability problem and highlight its ethical saliency. I try to make the case for the legal enforcement of a strong explainability requirement: human organizations which decide to automate decision-making should be legally obliged to demonstrate the capacity to explain and justify the algorithmic decisions that have an impact on the wellbeing, rights, and opportunities of those affected by the decisions. This legal duty can be derived from the demands of Rawlsian public reason. In the second part of the paper, I try to show that the argument from the limitations of human cognition fails to get AI off the hook of public reason. Against a growing trend in AI ethics, my main argument is that the analogy between human minds and artificial neural networks fails because it suffers from an atomistic bias which makes it blind to the social and institutional dimension of human reasoning processes. I suggest that developing interpretive AI algorithms is not the only possible answer to the explainability problem; social and institutional answers are also available and in many cases more trustworthy than techno-scientific ones. [ABSTRACT FROM AUTHOR]

This paper attempts to describe and address a specific puzzle related to compositionality in artificial networks such as Deep Neural Networks and machine learning in general. The puzzle identified here touches on a larger debate in Artificial Intelligence related to epistemic opacity but specifically focuses on computational applications of human level linguistic abilities or properties and a special difficulty with relation to these. Thus, the resulting issue is both general and unique. A partial solution is suggested. [ABSTRACT FROM AUTHOR]

Current advances in research, development and application of artificial intelligence (AI) systems have yielded a far-reaching discourse on AI ethics. In consequence, a number of ethics guidelines have been released in recent years. These guidelines comprise normative principles and recommendations aimed to harness the "disruptive" potentials of new AI technologies. Designed as a semi-systematic evaluation, this paper analyzes and compares 22 guidelines, highlighting overlaps but also omissions. As a result, I give a detailed overview of the field of AI ethics. Finally, I also examine to what extent the respective ethical principles and values are implemented in the practice of research, development and application of AI systems—and how the effectiveness in the demands of AI ethics can be improved. [ABSTRACT FROM AUTHOR]

The danger of human operators devolving responsibility to machines and failing to detect cases where they fail has been recognised for many years by industrial psychologists and engineers studying the human operators of complex machines. We call it "the control problem", understood as the tendency of the human within a human–machine control loop to become complacent, over-reliant or unduly diffident when faced with the outputs of a reliable autonomous system. While the control problem has been investigated for some time, up to this point its manifestation in machine learning contexts has not received serious attention. This paper aims to fill that gap. We argue that, except in certain special circumstances, algorithmic decision tools should not be used in high-stakes or safety-critical decisions unless the systems concerned are significantly "better than human" in the relevant domain or subdomain of decision-making. More concretely, we recommend three strategies to address the control problem, the most promising of which involves a complementary (and potentially dynamic) coupling between highly proficient algorithmic tools and human agents working alongside one another. We also identify six key principles which all such human–machine systems should reflect in their design. These can serve as a framework both for assessing the viability of any such human–machine system as well as guiding the design and implementation of such systems generally. [ABSTRACT FROM AUTHOR]

In this paper I argue that the search for explainable models and interpretable decisions in AI must be reformulated in terms of the broader project of offering a pragmatic and naturalistic account of understanding in AI. Intuitively, the purpose of providing an explanation of a model or a decision is to make it understandable to its stakeholders. But without a previous grasp of what it means to say that an agent understands a model or a decision, the explanatory strategies will lack a well-defined goal. Aside from providing a clearer objective for XAI, focusing on understanding also allows us to relax the factivity condition on explanation, which is impossible to fulfill in many machine learning models, and to focus instead on the pragmatic conditions that determine the best fit between a model and the methods and devices deployed to understand it. After an examination of the different types of understanding discussed in the philosophical and psychological literature, I conclude that interpretative or approximation models not only provide the best way to achieve the objectual understanding of a machine learning model, but are also a necessary condition to achieve post hoc interpretability. This conclusion is partly based on the shortcomings of the purely functionalist approach to post hoc interpretability that seems to be predominant in most recent literature. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) has historically been conceptualized in anthropomorphic terms. Some algorithms deploy biomimetic designs in a deliberate attempt to effect a sort of digital isomorphism of the human brain. Others leverage more general learning strategies that happen to coincide with popular theories of cognitive science and social epistemology. In this paper, I challenge the anthropomorphic credentials of the neural network algorithm, whose similarities to human cognition I argue are vastly overstated and narrowly construed. I submit that three alternative supervised learning methods—namely lasso penalties, bagging, and boosting—offer subtler, more interesting analogies to human reasoning as both an individual and a social phenomenon. Despite the temptation to fall back on anthropomorphic tropes when discussing AI, however, I conclude that such rhetoric is at best misleading and at worst downright dangerous. The impulse to humanize algorithms is an obstacle to properly conceptualizing the ethical challenges posed by emerging technologies. [ABSTRACT FROM AUTHOR]

The Internet is an important focus of attention for the philosophy of mind and cognitive science communities. This is partly because the Internet serves as an important part of the material environment in which a broad array of human cognitive and epistemic activities are situated. The Internet can thus be seen as an important part of the 'cognitive ecology' that helps to shape, support and (on occasion) realize aspects of human cognizing. Much of the previous philosophical work in this area has sought to analyze the cognitive significance of the Internet from the perspective of human cognition. There has, as such, been little effort to assess the cognitive significance of the Internet from the perspective of 'machine cognition'. This is unfortunate, because the Internet is likely to exert a significant influence on the shape of machine intelligence. The present paper attempts to evaluate the extent to which the Internet serves as a form of cognitive ecology for synthetic (machine-based) forms of intelligence. In particular, the phenomenon of Internet-situated machine intelligence is analyzed from the perspective of a number of approaches that are typically subsumed under the heading of situated cognition. These include extended, embedded, scaffolded and embodied approaches to cognition. For each of these approaches, the Internet is shown to be of potential relevance to the development and operation of machine-based cognitive capabilities. Such insights help us to appreciate the role of the Internet in advancing the current state-of-the-art in machine intelligence. [ABSTRACT FROM AUTHOR]

This paper revisits the often debated question Can machines think? It is argued that the usual identification of machines with the notion of algorithm has been both counter-intuitive and counter-productive. This is based on the fact that the notion of algorithm just requires an algorithm to contain a finite but arbitrary number of rules. It is argued that intuitively people tend to think of an algorithm to have a rather limited number of rules. The paper will further propose a modification of the above mentioned explication of the notion of machines by quantifying the length of an algorithm. Based on that it appears possible to reconcile the opposing views on the topic, which people have been arguing about for more than half a century. [ABSTRACT FROM AUTHOR]

In their joint paper entitled “ The Replication of the Hard Problem of Consciousness in AI and BIO- AI” (Boltuc et al. Replication of the hard problem of conscious in AI and Bio- AI: An early conceptual framework ), Nicholas and Piotr Boltuc suggest that machines could be equipped with phenomenal consciousness, which is subjective consciousness that satisfies Chalmer’s hard problem (We will abbreviate the hard problem of consciousness as “H-consciousness”). The claim is that if we knew the inner workings of phenomenal consciousness and could understand its’ precise operation, we could instantiate such consciousness in a machine. This claim, called the extra- strong AI thesis, is an important claim because if true it would demystify the privileged access problem of first-person consciousness and cast it as an empirical problem of science and not a fundamental question of philosophy. A core assumption of the extra-strong AI thesis is that there is no logical argument that precludes the implementation of H-consciousness in an organic or in-organic machine provided we understand its algorithm. Another way of framing this conclusion is that there is nothing special about H-consciousness as compared to any other process. That is, in the same way that we do not preclude a machine from implementing photosynthesis, we also do not preclude a machine from implementing H-consciousness. While one may be more difficult in practice, it is a problem of science and engineering, and no longer a philosophical question. I propose that Boltuc’s conclusion, while plausible and convincing, comes at a very high price; the argument given for his conclusion does not exclude any conceivable process from machine implementation. In short, if we make some assumptions about the equivalence of a rough notion of algorithm and then tie this to human understanding, all logical preconditions vanish and the argument grants that any process can be implemented in a machine. The purpose of this paper is to comment on the argument for his conclusion and offer additional properties of H-consciousness that can be used to make the conclusion falsifiable through scientific investigation rather than relying on the limits of human understanding. [ABSTRACT FROM AUTHOR]

From an object-oriented perspective, this paper investigates the interdisciplinary aspects of problem representation as well the differences between representation of problems in the mind and that in the machine. By defining an object as a combination of a symbol-structure and its associated operations, it shows how the representation of problems can become related to control, which conducts the search in finding a solution. Different types of representation of problems in the machine are classified into four categories, and in a similar way four distinct models are distinguished for the representation of problems in the mind. The concept of layered hierarchies, as the main theme of the object-oriented paradigm, is used to examine the implications of problem representation in the mind for improving the representation of problems in the machine. [ABSTRACT FROM AUTHOR]

In this paper I argue that Turing’s responses to the mathematical objection are straightforward, despite recent claims to the contrary. I then go on to show that by understanding the importance of learning machines for Turing as related not to the mathematical objection, but to Lady Lovelace’s objection, we can better understand Turing’s response to Lady Lovelace’s objection. Finally, I argue that by understanding Turing’s responses to these objections more clearly, we discover a hitherto unrecognized, substantive thesis in his philosophical thinking about the nature of mind. [ABSTRACT FROM AUTHOR]

Much of the recent work on the epistemology of causation has centered on two assumptions, known as the Causal Markov Condition and the Causal Faithfulness Condition. Philosophical discussions of the latter condition have exhibited situations in which it is likely to fail. This paper studies the Causal Faithfulness Condition as a conjunction of weaker conditions. We show that some of the weaker conjuncts can be empirically tested, and hence do not have to be assumed a priori. Our results lead to two methodologically significant observations: (1) some common types of counterexamples to the Faithfulness condition constitute objections only to the empirically testable part of the condition; and (2) some common defenses of the Faithfulness condition do not provide justification or evidence for the testable parts of the condition. It is thus worthwhile to study the possibility of reliable causal inference under weaker Faithfulness conditions. As it turns out, the modification needed to make standard procedures work under a weaker version of the Faithfulness condition also has the practical effect of making them more robust when the standard Faithfulness condition actually holds. This, we argue, is related to the possibility of controlling error probabilities with finite sample size (“uniform consistency”) in causal inference. [ABSTRACT FROM AUTHOR]

Computational philosophy (CP) aims at investigating many important concepts and problems of the philosophical and epistemological tradition in a new way by taking advantage of information-theoretic, cognitive, and artificial intelligence methodologies. I maintain that the results of computational philosophy meet the classical requirements of some Peircian ‘pragmatic ambitions’. Indeed, more than a 100 years ago, the American philosopher C.S. Peirce, when working on logical and philosophical problems, suggested the concept ofpragmatism(‘pragmaticism’, in his own words) as a logical criterion to analyze what words and concepts express through their practical meaning. Many words have been spent on creative processes and reasoning, especially in the case of scientific practices. In fact, many philosophers have usually offered a number of ways of construing hypotheses generation, but they aim at demonstrating that the activity of generating hypotheses is paradoxical, obscure, and thus not analyzable. Those descriptions are often so far from Peircian pragmatic prescription and so abstract to result completely unknowable and obscure. To dismiss this tendency and gain interesting insight about the so-called ‘logic of scientific discovery’ we need to build constructive procedures, which could play a role in moving the problem-solving process forward by implementing them in some actual models. The ‘computational turn’ gives us a new way to understand creative processes in a strictly pragmatic sense. In fact, by exploiting artificial intelligence and cognitive science tools, computational philosophy allows us to test concepts and ideas previously conceived only in abstract terms. It is in the perspective of theseactual computational modelsthat I find the central role ofabductionin the explanation of creative reasoning in science. I maintain that the computational philosophy analysis ofmodel-basedandmanipulativeabduction and ofexternalandepistemic mediatorsis important not only to delineate the actual practice of abduction, but also to further enhance the development of programs computationally adequate in rediscovering, or discovering for the first time, for example, scientific hypotheses or mathematical theorems. The last part of the paper is devoted to illustrating the problem of the extra-theoretical dimension of reasoning and discovery from the perspective of some mathematical cases derived from calculus and geometry. [ABSTRACT FROM AUTHOR]

This paper argues that chance (risk or opportunity) discovery is challenging, from a reasoning point of view, because it represents a dilemma for inductive reasoning. Chance discovery shares many features with the grue paradox. Consequently, Bayesian approaches represent a potential solution. The Bayesian solution evaluates alternative models generated using a temporal logic planner to manage the chance. Surprise indices are used in monitoring the conformity of the real world and the assessed probabilities. Game theoretic approaches are proposed to deal with multi-agent interaction in chance management. [ABSTRACT FROM AUTHOR]

The purpose of this work is to analyse the cognitive process of the domain theories in terms of the measurement theory to develop a computational machine learning approach for implementing it. As a result, the relational data mining approach, the authors proposed in the preceding books, was improved. We present the approach as an implementation of the cognitive process as the measurement theory perceived. We analyse the cognitive process in the first part of the paper and present the theory and method of the logically most powerful empirical theory discovery in the second. The theory is based on the notion of ‘law-like’ rules, which conform to all the properties of laws of nature, namely generality, simplicity, maximum refutability and minimum number of parameters. This notion is defined for deterministic and probabilistic cases. Based on the method, the ‘discovery’ system is developed. The system was successfully applied to many practical tasks. [ABSTRACT FROM AUTHOR]