Explainable artificial intelligence (XAI) is an emerging, multidisciplinary field of research that seeks to develop methods and tools for making AI systems more explainable or interpretable. XAI researchers increasingly recognise explainability as a context-, audience- and purpose-sensitive phenomenon, rather than a single well-defined property that can be directly measured and optimised. However, since there is currently no overarching definition of explainability, this poses a risk of miscommunication between the many different researchers within this multidisciplinary space. This is the problem we (...) seek to address in this paper. We outline a framework, called Explanatory Pragmatism, which we argue has two attractive features. First, it allows us to conceptualise explainability in explicitly context-, audience- and purpose-relative terms, while retaining a unified underlying definition of explainability. Second, it makes visible any normative disagreements that may underpin conflicting claims about explainability regarding the purposes for which explanations are sought. Third, it allows us to distinguish several dimensions of AI explainability. We illustrate this framework by applying it to a case study involving a machine learning model for predicting whether patients suffering disorders of consciousness were likely to recover consciousness. (shrink)

There are some philosophical questions that can be answered without attention to the social context in which evidence is produced and distributed.ing away from social context is an excellent way to ignore messy details and lay bare the underlying structure of the limits of inference. Idealization is entirely appropriate when one is essentially asking: In the best of all possible worlds, what am I entitled to infer? Yet, philosophers’ concerns often go beyond this domain. As an example I examine the (...) debate on mechanistic evidence and then reevaluate a canonical case study in this debate. I show that for the assessment of actual evidence, produced in a world that is far from ideal, omission of the social aspects of medical epistemology leads philosophers to draw the wrong lessons from cases they take as paradigmatic cases for their views. I close by arguing that social epistemology provides an avenue to incorporate these complications and provides the necessary framework to understand medical evidence. (shrink)

In recent years, bariatric surgery has become an increasingly popular treatment of obesity. The amount of resources spent on this kind of surgery has led to a heated debate among health care professionals and the general public, as each procedure costs at minimum $14,500 and thousands of patients undergo surgery every year. So far, no substantial argument for or against giving this treatment a high priority has, however, been presented. In this article, I argue that regardless which moral perspective we (...) consider—greatest need, utility or personal responsibility—the conclusion is that we should give bariatric surgery a high priority when allocating scarce resources in health care. (shrink)

In current philosophical discussions on evidence in the medical sciences, epidemiology has been used to exemplify a specific version of evidential pluralism. According to this view, known as the Russo–Williamson Thesis, evidence of both difference-making and mechanisms is produced to make causal claims in the health sciences. In this paper, I present an analysis of data and evidence in epidemiological practice, with a special focus on research on the exposome, and I cast doubt on the extent to which evidential pluralism (...) holds in this case. I start by focusing on the claim that molecular data allows for the production of mechanistic evidence. On the basis of a close look at the ways in which molecular data is used in exposome research, I caution against interpretations in terms of mechanistic evidence. Secondly, I expand my critical remarks on the thesis by addressing the conditions under which data is categorised as evidence in exposome research. I argue that these show that the classification of a dataset as a type of evidence is dependent on the ways in which the data is used. This is in contrast with the approach of evidential pluralism, where evidence is classified in different types on the basis of its intrinsic properties. Finally, I come back to what I consider the core of the thesis and suggest that the epidemiological research analysed in the paper indicates different interpretations of evidential pluralism and its applicability in the health sciences. (shrink)

Kevin Elliott and others separate two common arguments for the legitimacy of societal values in scientific reasoning as the gap and the error arguments. This article poses two questions: How are these two arguments related, and what can we learn from their interrelation? I contend that we can better understand the error argument as nested within the gap because the error is a limited case of the gap with narrower features. Furthermore, this nestedness provides philosophers with conceptual tools for analyzing (...) more robustly how values pervade science. (shrink)

Over the past 25 years, several new “medicines” have come screeching onto health care’s various platforms, including narrative medicine, personalized medicine, precision medicine and person-centred medicine. Philosopher Miriam Solomon calls the first three of these movements different “ways of knowing” or “methods,” and argues that they are each a response to shortcomings of methods that came before them. They should also be understood as reactions to the current dominant model of medicine. In this article, I will describe our dominant model, (...) which I call “the new medical model.” I will argue that several towering problems in modern medicine can be traced to its philosophical foundations, which calls for philosophical analysis. (shrink)

Introduction: Clinical practice guidelines (CPGs) are an important source of justification for clinical decisions in modern evidence-based practice. Yet, we have given little attention to how they argue their evidence. In particular, how do CPGs argue for treatment with long-term medications that are increasingly prescribed to older patients? Approach and rationale: I selected six disease-specific guidelines recommending treatment with five of the medication classes most commonly prescribed for seniors in Ontario, Canada. I considered the stated aims of these CPGs and (...) the techniques employed towards those aims. Finally, I reconstructed and logically analysed the arguments supporting recommendations for pharmacotherapy. Analysis: The primary function of CPGs is rhetorical, or persuasive, and their means of persuasion include both a display of their credibility and their argumentation. Arguments supporting pharmacotherapy recommendations for the target population follow a common inductive pattern: statistical generalization from randomized controlled trial (RCT) and meta-analysis evidence. Two of the CPGs also argue their treatment recommendations for older patients in this style, while three fail to justify pharmacotherapy specifically for the older population. Discussion: The arguments analysed lack the auxiliary assumptions that would warrant making a generalization about the clinical effectiveness of medications for the older population. Guidelines reason using simple induction, while ignoring important inferential gaps. Future guidelines should aspire to be well-reasoned rather than simply evidence-based; argue from a plurality of evidence; be wary of hasty inductions; appropriately limit the scope of their recommendations; and avoid making law-like, prescriptive generalizations. (shrink)

Over the past several decades, we devoted much energy to generating, reviewing and summarizing evidence. We have given far less attention to the issue of how to thoughtfully apply the evidence once we have it. That’s fine if all we care about is that our clinical decisions are evidence-based, but not so good if we also want them to be well-reasoned. Let us not forget that evidence based medicine (EBM) grew out of an interest in making medicine ‘rational’, with the (...) idea that rational clinical evaluations should be evidence-based. I agree with the uncontroversial statement that the best decision is supported, at least in part, by the best available evidence. Rationality, however, is constituted by reasoning, not evidence. Complete arguments are necessary for rational evaluations, arguments that begin with general evidence and end in a conclusion about a particular patient. In order to traverse these inferential gaps, medicine must address the issue of how to establish, as an intermediate premise, what the evidence has to say about the efficacy of an intervention for particular patients in a particular practice setting. (shrink)

This paper argues that there exists a collective epistemic state of ‘Broad Medical Uncertainty’ regarding the effectiveness of many medical interventions. We outline the features of BMU, and describe some of the main contributing factors. These include flaws in medical research methodologies, bias in publication practices, financial and other conflicts of interest, and features of how evidence is translated into practice. These result in a significant degree of uncertainty regarding the effectiveness of many medical treatments and unduly optimistic beliefs about (...) the benefit/harm profiles of such treatments. We argue for an ethical presumption in favour of openness regarding BMU as part of a ‘Corrective Response’. We then consider some objections to this position, including concerns that public honesty about flaws in medical research could undermine trust in healthcare institutions. We suggest that, as it stands, the Anti-Corrective Response is unconvincing. (shrink)

The article addresses the question of how should scientific peers revise their beliefs (if at all) upon recognized disagreement. After presenting the basics of peer disagreement in sections 1 and 2, we focus, in section 3, on a concrete case of scientific disagreement, to wit, the dispute over the evidential status of randomized control trials in medical practice. The examination of this case motivates the idea that some scientific disagreements permit a steadfast reaction. In section 4, we support this conclusion (...) by providing a normative argument in the same direction; if we are correct, typical reasons for conciliation are absent in this kind of scientific disagreements. (shrink)

Within the evidence-based medicine construct, clinical expertise is acknowledged to be both derived from primary experience and necessary for optimal medical practice. Primary experience in medical practice, however, remains undervalued. Clinicians’ primary experience tends to be dismissed by EBM as unsystematic or anecdotal, a source of bias rather than knowledge, never serving as the “best” evidence to support a clinical decision. The position that clinical expertise is necessary but that primary experience is untrustworthy in clinical decision-making is epistemically incoherent. Here (...) we argue for the value and utility of knowledge gained from primary experience for the practice of medicine. Primary experience provides knowledge necessary to diagnose, treat, and assess response in individual patients. Hierarchies of evidence, when advanced as guides for clinical decisions, mistake the relationship between propositional and experiential knowledge. We argue that primary experience represents a kind of medical knowledge distinct from the propositional knowledge produced by clinical research, both of which are crucial to determining the best diagnosis and course of action for particular patients. (shrink)

The themes of this issue—which include the meaning of our health and disease concepts, the so-called “medical gaze” and its embedded power relations, and the epistemic value of mixing therapy with research—are introduced by reflecting on a case about an infant girl whose head is observed to be somewhat flat.

After describing the disorder of psychopathy, I examine the theories and the evidence concerning the psychopaths’ deficient moral capacities. I first examine whether or not psychopaths can pass tests of moral knowledge. Most of the evidence suggests that they can. If there is a lack of moral understanding, then it has to be due to an incapacity that affects not their declarative knowledge of moral norms, but their deeper understanding of them. I then examine two suggestions: it is their deficient (...) practical reason or their stunted emotions that are at fault. The evidence supports both explanations. I conclude with an overview of the debate concerning whether they are morally or legally responsible for their actions. (shrink)

Cartwright and Munro argued that extrapolation of findings from randomized controlled trials to other settings can be difficult because information about the underlying causal structure and subgroups is often not available. They advocated the use of ‘capacities’ – that is fixed causal contributions – in predicting effects of interventions. In psychiatry, it is often not possible to determine what the fixed causal contributions are and one can only establish ‘approximate capacities’. However, using ‘approximate capacities’ does imply a different way of (...) evaluating health services, especially combined interventions. In health service research, if different studies, randomized controlled trials or other designs, have given different outcomes, the best way to investigate the effectiveness of a particular way of service organization is not to conduct more randomized controlled trials. It is preferable to study the effects of certain elements of the complex intervention, which have been tested before in other settings, that is investigating ‘approximate capacities’. One should check whether the separate elements do form a part of the complex intervention in practice and whether they have the same effect as in other studies and if not, why not. This enhances knowledge about the underlying causal structure and increases the possibility of extrapolation of the findings. (shrink)

What clinicians, biomedical scientists, and other health care professionals know as individuals or as groups and how they come to know and use knowledge are central concerns of medical epistemology. Activities associated with knowledge production and use are called epistemic practices. Such practices are considered in biomedical and clinical literatures, social sciences of medicine, philosophy of science and philosophy of medicine, and also in other nonmedical literatures. A host of different kinds of knowledge claims have been identified, each with different (...) uses and logics of justification. A general framework is needed to situate these diverse contributions in medical epistemology, so we can see how they fit together. But developing such a framework turns out to be quite tricky. In this survey, three possible frameworks are considered along with the difficulties associated with each of them. The essay concludes with a fourth framework, which considers any epistemology as part of a practice that is oriented toward overcoming errors that emerge in antecedently given practices where knowledge is developed and used. As medicine indirectly advances health by directly mitigating disease, so epistemology indirectly advances knowledge by directly mitigating error. (shrink)

The UK’s National Institute for Health and Care Excellence is responsible for conducting health technology assessment on behalf of the National Health Service. In seeking to justify its recommendations to the NHS about which technologies to fund, NICE claims to adopt two complementary ethical frameworks, one procedural—accountability for reasonableness —and one substantive—an ‘ethics of opportunity costs’ that rests primarily on the notion of allocative efficiency. This study is the first to empirically examine normative changes to NICE’s approach and to analyse (...) whether these enhance or diminish the fairness of its decision-making, as judged against these frameworks. It finds that increasing formalisation of NICE’s approach and a weakening of the burden of proof laid on technologies undergoing HTA have together undermined its commitment to EOC. This implies a loss of allocative efficiency and a shift in the balance of how the interests of different NHS users are served, in favour of those who benefit directly from NICE’s recommendations. These changes also weaken NICE’s commitment to AfR by diminishing the publicity of its decision-making and by encouraging the adoption of rationales that cannot easily be shown to meet the relevance condition. This signals a need for either substantial reform of NICE’s approach, or more accurate communication of the ethical reasoning on which it is based. The study also highlights the need for further empirical work to explore the impact of these policy changes on NICE’s practice of HTA and to better understand how and why they have come about. (shrink)

Arguments that medical decision making should rely on a variety of evidence often begin from the claim that meta-analysis has been shown to be problematic. In this paper, I first examine Stegenga’s argument that meta-analysis requires multiple decisions and thus fails to provide an objective ground for medical decision making. Next, I examine three arguments from social epistemologists that contend that meta-analyses are systematically biased in ways not appreciated by standard epistemology. In most cases I show that critiques of meta-analysis (...) fail to account for the full range of meta-analytic procedures. In the remainder of cases, I argue that the critiques identify problems that do not uniquely cut against meta-analysis. I close by suggesting one reason why it may be pragmatically rational to violate the principle of total evidence and by outlining the criteria for a successful argument against meta-analysis. A set of criteria I contend remain unmet. (shrink)

Hierarchies of evidence are an important and influential tool for appraising evidence in medicine. In recent years, hierarchies have been formally adopted by organizations including the Cochrane Collaboration [1], NICE [2,3], the WHO [4], the US Preventive Services Task Force [5], and the Australian NHMRC [6,7]. The development of such hierarchies has been regarded as a central part of Evidence-Based Medicine, a movement within healthcare which prioritises the use of epidemiological evidence such as that provided by Randomised Controlled Trials. Philosophical (...) work on the methodology of medicine has so far mostly focused on claims about the superiority of RCTs, and hence has largely neglected the questions of what hierarchies are, what assumptions they require, and how they affect clinical practice. This thesis shows that there is great variation in the hierarchies defended and in the interpretations they are, and can be, given. The interpretative assumptions made in using hierarchies are crucial to the content and defensibility of the underlying philosophical commitments concerning evidence and medical practice. Once this variation is been identified, it becomes clear that the little philosophical work that has been done so far affects only some hierarchies, under some interpretations. Modest interpretations offered by La Caze [11], conditional hierarchies like GRADE [12-14], and heuristic approaches such as that defended by Howick et al. [15,16] all survive previous philosophical criticism. This thesis extends previous criticisms by arguing that modest interpretations are so weak as to be unhelpful for clinical practice; that GRADE and similar conditional models omit clinically relevant information, such as information about variation in treatments’ effects and the causes of different responses to therapy; and that heuristic approaches lack the necessary empirical support. The conclusion is that hierarchies in general embed untenable philosophical assumptions: principally that information about average treatment effects backed by high-quality evidence can justify strong recommendations, and that the impact of evidence from individual studies can and should be appraised in isolation. Hierarchies are a poor basis for the application of evidence in clinical practice. The Evidence-Based Medicine movement should move beyond them and explore alternative tools for appraising the overall evidence for therapeutic claims. (shrink)

Summary of Transformative Experience by L.A. Paul and replies to symposiasts. Discussion of undefined values, preference change, authenticity, experiential value, collective minds, mind control.

Male circumcision—partial or total removal of the penile prepuce—has been proposed as a public health measure in Sub-Saharan Africa, based on the results of three randomized control trials showing a relative risk reduction of approximately 60 per cent for voluntary, adult male circumcision against female-to-male human immunodeficiency virus transmission in that context. More recently, long-time advocates of infant male circumcision have argued that these findings justify involuntary circumcision of babies and children in dissimilar public health environments, such as the USA, (...) Australasia and Europe. In this article, we take a close look at the necessary ethical and empirical steps that would be needed to bridge the gap between the African RCTs and responsible public health policy in developed countries. In the course of doing so, we discuss some of the main disagreements about the moral permissibility of performing a nontherapeutic surgery on a child to benefit potential future sexual partners of his. In this context, we raise concerns not only about weaknesses in the available evidence concerning such claims of benefit, but also about a child’s moral interest in future autonomy and the preservation of his bodily integrity. We conclude that circumcision of minors in developed countries on public health grounds is much harder to justify than proponents of the surgery suggest. (shrink)

Simple extrapolation is the orthodox approach to extrapolating from clinical trials in evidence-based medicine: extrapolate the relative effect size from the trial unless there is a compelling reason not to do so. I argue that this method relies on a myth and a fallacy. The myth of simple extrapolation is the idea that the relative risk is a ‘golden ratio’ that is usually transportable due to some special mathematical or theoretical property. The fallacy of simple extrapolation is an unjustified argument (...) from ignorance: we conclude that the relative effect size is transportable in the absence of evidence to the contrary. In short, simple extrapolation is a deeply problematic solution to the problem of extrapolation. (shrink)

Arguments in favor of greater research-practice integration in medicine have tended to be ethical, political, or pragmatic. There are good epistemic reasons to pursue greater integration, and it is important to think through these reasons in order to avoid inadvertently designing new systems in ways that replicate the epistemic elitism common within current systems. Meaningful transformation within health care is possible with close attention to all reasons in favor of greater research-practice integration, including epistemic reasons.

Evidence, as viewed through the lens of statistical significance, is not always as it appears! In the investigation of clinical research findings arising from statistical analyses, a fundamental initial step for the emerging fraud detective is to retrieve the source data for cross-examination with the study data. Recognizing that source data are not always forthcoming and that, realistically speaking, the investigator may be uninitiated in fraud detection and investigation, this paper will highlight some key methodological procedures for providing a sounder (...) evidence base for withdrawing from a study on grounds of integrity. The promotion of patient safety is paramount. However, there is a broader rationale for disseminating these ideas. This includes empowering researchers to optimize their personal integrity, make informed choices regarding membership of future research collaborations and successfully voice their concerns to journal editors, particularly where a conflict of interests can render such dialogues particularly difficult. Recommendations will be supported by topical case studies and practical steps involving data exploration, testing of baseline data and application of Benford’s Law. While this paper has a clinical focus, the advice provided is transferrable to a wide range of multidisciplinary research settings outside of Medicine. (shrink)