RESUMO
To ensure equitable quality of care, differences in machine learning model performance between patient groups must be addressed. Here, we argue that two separate mechanisms can cause performance differences between groups. First, model performance may be worse than theoretically achievable in a given group. This can occur due to a combination of group underrepresentation, modeling choices, and the characteristics of the prediction task at hand. We examine scenarios in which underrepresentation leads to underperformance, scenarios in which it does not, and the differences between them. Second, the optimal achievable performance may also differ between groups due to differences in the intrinsic difficulty of the prediction task. We discuss several possible causes of such differences in task difficulty. In addition, challenges such as label biases and selection biases may confound both learning and performance evaluation. We highlight consequences for the path toward equal performance, and we emphasize that leveling up model performance may require gathering not only more data from underperforming groups but also better data. Throughout, we ground our discussion in real-world medical phenomena and case studies while also referencing relevant statistical theory.
RESUMO
There are high hopes that clinical decisions can be improved by adopting algorithms trained to estimate the likelihood that a patient suffers a condition C. Introducing work on the epistemic value of purely statistical evidence in legal epistemology I show that a certain type of AI devices for making medical decisions about persons rely on purely statistical evidence and that it raises an important question about the appropriateness of relying on such devices for allocating health resources. If the argument I present is sound, then it suggests a radical rethinking of the use of prevalent types of AI devices as well as the use of statistical evidence in medical practice more generally.
RESUMO
When is it justified to use opaque artificial intelligence (AI) output in medical decision-making? Consideration of this question is of central importance for the responsible use of opaque machine learning (ML) models, which have been shown to produce accurate and reliable diagnoses, prognoses, and treatment suggestions in medicine. In this article, I discuss the merits of two answers to the question. According to the Explanation View, clinicians must have access to an explanation of why an output was produced. According to the Validation View, it is sufficient that the AI system has been validated using established standards for safety and reliability. I defend the Explanation View against two lines of criticism, and I argue that within the framework of evidence-based medicine mere validation seems insufficient for the use of AI output. I end by characterizing the epistemic responsibility of clinicians and point out how a mere AI output cannot in itself ground a practical conclusion about what to do.
RESUMO
This article's main contributions are twofold: 1) to demonstrate how to apply the general European Union's High-Level Expert Group's (EU HLEG) guidelines for trustworthy AI in practice for the domain of healthcare and 2) to investigate the research question of what does "trustworthy AI" mean at the time of the COVID-19 pandemic. To this end, we present the results of a post-hoc self-assessment to evaluate the trustworthiness of an AI system for predicting a multiregional score conveying the degree of lung compromise in COVID-19 patients, developed and verified by an interdisciplinary team with members from academia, public hospitals, and industry in time of pandemic. The AI system aims to help radiologists to estimate and communicate the severity of damage in a patient's lung from Chest X-rays. It has been experimentally deployed in the radiology department of the ASST Spedali Civili clinic in Brescia, Italy, since December 2020 during pandemic time. The methodology we have applied for our post-hoc assessment, called Z-Inspection®, uses sociotechnical scenarios to identify ethical, technical, and domain-specific issues in the use of the AI system in the context of the pandemic.
RESUMO
Explainability for artificial intelligence (AI) in medicine is a hotly debated topic. Our paper presents a review of the key arguments in favor and against explainability for AI-powered Clinical Decision Support System (CDSS) applied to a concrete use case, namely an AI-powered CDSS currently used in the emergency call setting to identify patients with life-threatening cardiac arrest. More specifically, we performed a normative analysis using socio-technical scenarios to provide a nuanced account of the role of explainability for CDSSs for the concrete use case, allowing for abstractions to a more general level. Our analysis focused on three layers: technical considerations, human factors, and the designated system role in decision-making. Our findings suggest that whether explainability can provide added value to CDSS depends on several key questions: technical feasibility, the level of validation in case of explainable algorithms, the characteristics of the context in which the system is implemented, the designated role in the decision-making process, and the key user group(s). Thus, each CDSS will require an individualized assessment of explainability needs and we provide an example of how such an assessment could look like in practice.
RESUMO
In this article I consider two pertinent questions that practitioners must consider when they deploy an algorithmic system as support in clinical shared decision-making. The first question concerns how to interpret and assess the significance of different performance measures for clinical decision-making. The second question concerns the professional obligations that practitioners have to communicate information about the quality of an algorithm's output to patients in light of the principles of autonomy, beneficence, and justice. In the article I review the four standard performance measures used to evaluate and validate algorithms, outline their role in the discussion of algorithmic fairness, and discuss the professional responsibilities that practitioners face when communicating information about these measures to patients.
Assuntos
Tomada de Decisão Clínica , Justiça Social , Inteligência Artificial , Beneficência , Tomada de Decisões , Humanos , Autonomia PessoalRESUMO
It has been argued that the precautionary principle is incoherent and thus useless as a guide for regulatory policy. In a recent paper in Bioethics, Wareham and Nardini propose a response to the 'precautionary paradox' according to which the precautionary principle's usefulness for decision making in policy and regulation contexts can be justified by appeal to a probability threshold discriminating between negligible and non-negligible risks. It would be of great significance to debates about risk and precaution if there were a sound method for determining a minimum probability threshold of negligible risk. This is what Wareham and Nardini aim to do. The novelty of their approach is that they suggest that such a threshold should be determined by a method of public deliberation. In this article I discuss the merits of Wareham and Nardini's public deliberation method for determining thresholds. I raise an epistemic worry about the public deliberation method they suggest, and argue that their proposal is inadequate due to a hidden assumption that the acceptability of a risk can be completely analysed in terms of its probability.
Assuntos
Política Pública , Biologia Sintética , Tomada de Decisões , Humanos , ProbabilidadeRESUMO
Gene-editing technology, such as CRISPR/Cas9, holds great promise for the advancement of science and many useful applications technology. This foundational technology enables modification of the genetic structure of any living organisms with unprecedented precision. Yet, in order to enhance its potential for societal benefit, it is necessary to adapt rules and produce adequate regulations. This requires an interdisciplinary effort in legal thinking. Any legislative initiative needs to consider both the benefits and the problematic aspects of gene editing, from a broader societal and value-based perspective. This paper stems from an interdisciplinary research project seeking to identify and discuss some of the most pressing legal implications of gene-editing technology and how to address these. While the questions raised by gene editing are global, laws and regulations are to a great extent bound by national borders. This paper presents a European perspective, written for a global audience, and intends to contribute to the global debate. The analysis will include brief references to corresponding USA rules in order to place these European debates in the broader international context. Our legal analysis incorporates interdisciplinary contributes concerning the scientific state of the art, philosophical thinking regarding the precautionary principle and dual-use issues as well as the importance of communication, social perception, and public debate. Focusing mainly in the main regulatory and patent law issues, we will argue that (a) general moratoriums and blank prohibitions do a disservice to science and innovation; (b) it is crucial to carefully consider a complex body of international and European fundamental rights norms applicable to gene editing;
RESUMO
A widespread and influential characterization of synthetic biology emphasizes that synthetic biology is the application of engineering principles to living systems. Furthermore, there is a strong tendency to express the engineering approach to organisms in terms of what seems to be an ontological claim: organisms are machines. In the paper I investigate the ontological and heuristic significance of the machine analogy in synthetic biology. I argue that the use of the machine analogy and the aim of producing rationally designed organisms does not necessarily imply a commitment to mechanical biology. The ideal of applying engineering principles to biology is best understood as expressing recognition of the machine-unlikeness of natural organisms and the limits of human cognition. The paper suggests an interpretation of the identification of organisms with machines in synthetic biology according to which it expresses a strategy for representing, understanding, and constructing living systems that are more machine-like than natural organisms.
Assuntos
Filosofia/história , Biologia Sintética/história , Bioengenharia , História do Século XVII , História do Século XVIII , História do Século XIX , História do Século XX , História do Século XXI , Biologia de SistemasRESUMO
Theorists analyzing the concept of disease on the basis of the notion of dysfunction consider disease to be dysfunction requiring. More specifically, dysfunction-requiring theories of disease claim that for an individual to be diseased certain biological facts about it must be the case. Disease is not wholly a matter of evaluative attitudes. In this paper, I consider the dysfunction-requiring component of Wakefield's hybrid account of disease in light of the artifactual organisms envisioned by current research in synthetic biology. In particular, I argue that the possibility of artifactual organisms and the case of oncomice and other bred or genetically modified strains of organism constitute a significant objection to Wakefield's etiological account of the dysfunction requirement. I then develop a new alternative understanding of the dysfunction requirement that builds on the organizational theory of function. I conclude that my suggestion is superior to Wakefield's theory because it (a) can accommodate both artifactual and naturally evolved organisms, (b) avoids the possibility of there being a conflict between what an organismic part is supposed to do and the health of the organism, and (c) provides a nonarbitrary and practical way of determining whether dysfunction occurs.
Assuntos
Temas Bioéticos , Doença , Filosofia Médica , Biologia Sintética/ética , Animais , Animais Endogâmicos , HumanosRESUMO
This article considers naturalistic analyses of the concepts of health and disease in light of the possibility of constructing novel living systems. The article begins by introducing the vision of synthetic biology as the application of engineering principles to the construction of biological systems, the main analyses of the concepts of health and disease, and the standard theories of function in artefacts and organisms. The article then suggests that reflection on the possibility of artefactual organisms amounts to a challenge to the functional theories of health and disease proposed by Wakefield and Boorse. More specifically, Wakefield and Boorse's theories are reconstructed as responses to a dilemma concerning how to allow for the ascription of health and disease to artefactual organisms without at the same time opening up the possibility of diseased nonliving artefacts such as cars and computers. It is argued that neither response will enable us to ascribe health and disease to artefactual organisms, because both theories, in order to rule out the possibility of ascribing health and disease to nonliving artefacts, make such ascriptions conditional on having a natural-selection history or being part of a species which has been designed by evolution.
Assuntos
Saúde , Biologia Sintética , Animais , Doença , Saúde/ética , Humanos , Modelos Biológicos , Biologia Sintética/éticaRESUMO
Synthetic biology is an emerging discipline that aims to apply rational engineering principles in the design and creation of organisms that are exquisitely tailored to human ends. The creation of artificial life raises conceptual, methodological and normative challenges that are ripe for philosophical investigation. This special issue examines the defining concepts and methods of synthetic biology, details the contours of the organism-artifact distinction, situates the products of synthetic biology vis-à-vis this conceptual typology and against historical human manipulation of the living world, and explores the normative implications of these conclusions. In addressing the challenges posed by emerging biotechnologies, new light can be thrown on old problems in the philosophy of biology, such as the nature of the organism, the structure of biological teleology, the utility of engineering metaphors and methods in biological science, and humankind's relationship to nature.
Assuntos
Artefatos , Biotecnologia , Vida , Biologia Sintética , HumanosRESUMO
In this paper I assess the explanatory powers of theories of function in the context of products that may result from synthetic biology. The aim is not to develop a new theory of functions, but to assess existing theories of function in relation to a new kind of biological and artifactual entity that might be produced in the not-too-distant future by means of synthetic biology. The paper thus investigates how to conceive of the functional nature of living systems that are not the result of evolution by natural selection, or instantly generated by cosmic coincidence, but which are products of intelligent design. The paper argues that the aetiological theory of proper functions in organisms and artifacts is inadequate as an account of proper functions in such 'Paley organisms' and defends an alternative organisational approach. The paper ends by considering the implications of the discussion of biological function for questions about the interests and moral status of non-sentient organisms.