Oliver Buchholz

The uptake of social science methods by bioethics significantly expanded its methodological spectrum, raising new theoretical, methodological, and practical questions. Recently, we are witnessing another trend, adding advanced data science methods to bioethics’ toolkit to aid, for example, in online data analysis, support scholarly writing, and inform clinical ethics. This article explores the emerging field of Digital Bioethics across its dimensions by analysing the tangled relationship between topics and methods, highlighting intersections between Digital Bioethics and Bioethics of the Digital, and advocating for a methods-based definition of the field. The use of advanced data science methods within bioethics must be interpreted in the context of the use of Artificial Intelligence (AI) in health care. At the same time, it presents unique opportunities and challenges. Defining, and thus demarcating, Digital Bioethics can create support for the new field but also requires navigating trade-offs. To do so, we take four kindred academic fields as points of comparison (Digital Humanities, Experimental Philosophical Bioethics, computational medicine and digitised biology) to analyse what each of them teaches for critically assessing and further developing Digital Bioethics. The article discusses potential pitfalls and concludes with recommendations on how the field can fully develop its potential to promote bioethical research and argument. Furthermore, the article discusses how a critical reflection of the use of AI methods within bioethics itself will also contribute to the ethical oversight of increasingly AI-driven branches of healthcare.

Which kind of artificial intelligence do we want to live with? Should machines explain themselves to us? Machine learning techniques are developing at a rapid pace and find applications not only in banal everyday uses, but also in high-stake situations, including science, medicine, banking, law, and business. But it is impossible to reconstruct how they reach their results and to judge whether they reach their results in the intended way. The mechanism is entirely opaque. This prompts a lot of justified skepticism and criticism of these computer programs. By closely investigating the foundations of opacity and explanations from a philosophy of science and epistemological perspective, Buchholz comes to more optimistic conclusions. This book derives practical consequences from a rigorous conceptual analysis of opacity, paving the way to an effective regulation of machine learning, and will advance the debate about the nature of explanation in the philosophy of science.

The increasing relevance of AI systems paired with their repeatedly observed opacity gave rise to the field of explainable artificial intelligence (XAI). Methods of XAI are being developed and evaluated based on whether they overcome said opacity by providing explanations, thereby apparently pursuing an epistemic end. What is commonly sidestepped, however, is the distinction between the ability and the obligation to explain: In which specific cases and under what specific circumstances is there a moral obligation to give an explanation? We address this issue by first arguing that apart from their epistemic virtues, explanations can be ethically relevant and hence subject to moral considerations. Further, we build on Kantian theories to provide a framework that systematizes in which instances there is a moral obligation to explain and in which instances there is none. We relate our framework both to Forst's "right to justification" and to the "right to explanation" as codified in recent legislation. Finally, we show that our framework does not only close a gap in clarifying when explanations of AI systems are required, but that it also paves the way for concrete suggestions regarding the design of XAI methods.

Machine learning operates at the intersection of statistics and computer science. This raises the question as to its underlying methodology. While much emphasis has been put on the close link between the process of learning from data and induction, the falsificationist component of machine learning has received minor attention. In this article, we argue that the idea of falsification is central to the methodology of machine learning. It is commonly thought that machine learning algorithms infer general prediction rules from past observations. This is akin to a statistical procedure by which estimates are obtained from a sample of data. But machine learning algorithms can also be described as choosing one prediction rule from an entire class of functions. In particular, the algorithm that determines the weights of an artificial neural network operates by empirical risk minimization and rejects prediction rules that lack empirical adequacy. It also exhibits a behaviour of implicit regularization that pushes hypothesis choice towards simpler prediction rules. We argue that taking both aspects together gives rise to a falsificationist account of artificial neural networks.

Die Technologie des maschinellen Lernens spielt inzwischen eine bedeutende Rolle im Forschungsprozess verschiedener wissenschaftlicher Disziplinen. Das vorliegende Kapitel untersucht diese Rolle des maschinellen Lernens, vor allem hinsichtlich ihrer methodologischen und wissenschaftstheoretischen Implikationen. Ein besonderes Augenmerk liegt dabei auf der Frage, inwiefern eine Kontinuität zwischen der bislang in der empirischen Forschung dominierenden klassischen Statistik und dem maschinellen Lernen besteht.

While much of politics is about making decisions, machine learning is about making predictions. Despite this diverging focus, machine learning is increasingly deployed in politically sensitive fields. This chapter focuses on the tension arising from this deployment. We specifically address the question of whether machine learning-based political decision making can be justified and how this question of justification interrelates with considerations about epistemological aspects of machine learning. To that end, we evaluate the epistemic state that individual decision makers can reach when relying on machine learning systems and assess when this state can be considered justified. We then broaden this first-person perspective and turn to political justification. By employing the so-called public justification principle, we investigate whether epistemic justification transmits to political justification in applications of machine learning. We argue that in light of the epistemological considerations, a series of problems arise concerning the political justification of deploying these models.

Machine learning (ML) models recently led to major breakthroughs in predictive tasks in the natural sciences. Yet their benefits for the social sciences are less evident, as even high-profile studies on the prediction of life trajectories have shown to be largely unsuccessful – at least when measured in traditional criteria of scientific success. This paper tries to shed light on this remarkable performance gap. Comparing two social science case studies to a paradigm example from the natural sciences, we argue that, in addition to explanation, prediction is an important goal of social science – and we identify constraints that impede pure ML prediction from being successful in that field. As a remedy, we outline elements of an integrative modelling approach that combines explanatory models and predictive ML models.

Explainable artificial intelligence (XAI) seeks to produce explanations for those machine learning methods which are deemed opaque. However, there is considerable disagreement about what this means and how to achieve it. Authors disagree on what should be explained (topic), to whom something should be explained (stakeholder), how something should be explained (instrument), and why something should be explained (goal). In this paper, I employ insights from means-end epistemology to structure the field. According to means-end epistemology, different means ought to be rationally adopted to achieve different epistemic ends. Applied to XAI, different topics, stakeholders, and goals thus require different instruments. I call this the means-end account of XAI. The means-end account has a descriptive and a normative component: on the one hand, I show how the specific means-end relations give rise to a taxonomy of existing contributions to the field of XAI; on the other hand, I argue that the suitability of XAI methods can be assessed by analyzing whether they are prescribed by a given topic, stakeholder, and goal.

Methods of machine learning (ML) are gradually complementing and sometimes even replacing methods of classical statistics in science. This raises the question whether ML faces the same methodological problems as classical statistics. This paper sheds light on this question by investigating a long-standing challenge to classical statistics: the reference class problem (RCP). It arises whenever statistical evidence is applied to an individual object, since the individual belongs to several reference classes and evidence might vary across them. Thus, the problem consists in choosing a suitable reference class for the individual. I argue that deep neural networks (DNNs) are able to overcome specific instantiations of the RCP. Whereas the criteria of narrowness, reliability, and homogeneity, that have been proposed to determine a suitable reference class, pose an inextricable tradeoff to classical statistics, DNNs are able to satisfy them in some situations. On the one hand, they can exploit the high dimensionality in big-data settings. I argue that this corresponds to the criteria of narrowness and reliability. On the other hand, ML research indicates that DNNs are generally not susceptible to overfitting. I argue that this property is related to a particular form of homogeneity. Taking both aspects together reveals that there are specific settings in which DNNs can overcome the RCP.

Machine learning operates at the intersection of statistics and computer science. This raises the question as to its underlying methodology. While much emphasis has been put on the close link between the process of learning from data and induction, the falsificationist component of machine learning has received minor attention. In this paper, we argue that the idea of falsification is central to the methodology of machine learning. It is commonly thought that machine learning algorithms infer general prediction rules from past observations. This is akin to a statistical procedure by which estimates are obtained from a sample of data. But machine learning algorithms can also be described as choosing one prediction rule from an entire class of functions. In particular, the algorithm that determines the weights of an artificial neural network operates by empirical risk minimization and rejects prediction rules that lack empirical adequacy. It also exhibits a behavior of implicit regularization that pushes hypothesis choice toward simpler prediction rules. We argue that taking both aspects together gives rise to a falsificationist account of artificial neural networks.