One of the most contentious ethical issues in the neonatal intensive care unit is the withdrawal of life-sustaining treatment from infants who may otherwise survive. In practice, one of the most important factors influencing this decision is the prediction that the infant will be severely intellectually disabled. Most professional guidelines suggest that decisions should be made on the basis of the best interests of the infant. It is, however, not clear how intellectual disability affects those interests. Why should intellectual disability (...) be more important than physical disability to the future interests of an infant? Is it discriminatory to base decisions on this? This paper will try to unravel the above questions. It seems that if intellectual disability does affect the best interests of the child it must do so in one of three ways. These possibilities will be discussed as well as the major challenges to the notion that intellectual disability should have a role in such decisions. The best interests of the child can be affected by severe or profound intellectual disability. It is, though, not as clear-cut as some might expect. (shrink)

We argue that artificial networks are explainable and offer a novel theory of interpretability. Two sets of conceptual questions are prominent in theoretical engagements with artificial neural networks, especially in the context of medical artificial intelligence: Are networks explainable, and if so, what does it mean to explain the output of a network? And what does it mean for a network to be interpretable? We argue that accounts of “explanation” tailored specifically to neural networks have ineffectively reinvented the wheel. In (...) response to, we show how four familiar accounts of explanation apply to neural networks as they would to any scientific phenomenon. We diagnose the confusion about explaining neural networks within the machine learning literature as an equivocation on “explainability,” “understandability” and “interpretability.” To remedy this, we distinguish between these notions, and answer by offering a theory and typology of interpretation in machine learning. Interpretation is something one does to an explanation with the aim of producing another, more understandable, explanation. As with explanation, there are various concepts and methods involved in interpretation: Total or Partial, Global or Local, and Approximative or Isomorphic. Our account of “interpretability” is consistent with uses in the machine learning literature, in keeping with the philosophy of explanation and understanding, and pays special attention to medical artificial intelligence systems. (shrink)

In a recent reply to our article, “What is Interpretability?,” Prasetya argues against our position that artificial neural networks are explainable. It is claimed that our indefeasibility thesis—that adding complexity to an explanation of a phenomenon does not make the phenomenon any less explainable—is false. More precisely, Prasetya argues that unificationist explanations are defeasible to increasing complexity, and thus, we may not be able to provide such explanations of highly complex AI models. The reply highlights an important lacuna in our (...) original paper, the omission of the unificationist account of explanation, and affords us the opportunity to respond. Here, we argue that artificial neural networks are explainable in a way that should satisfy unificationists and that interpretability methods present ways in which ML theories can achieve unification. (shrink)

When making substituted judgments for incapacitated patients, surrogates often struggle to guess what the patient would want if they had capacity. Surrogates may also agonize over having the (sole) responsibility of making such a determination. To address such concerns, a Patient Preference Predictor (PPP) has been proposed that would use an algorithm to infer the treatment preferences of individual patients from population-level data about the known preferences of people with similar demographic characteristics. However, critics have suggested that even if such (...) a PPP were more accurate, on average, than human surrogates in identifying patient preferences, the proposed algorithm would nevertheless fail to respect the patient’s (former) autonomy since it draws on the ‘wrong’ kind of data: namely, data that are not specific to the individual patient and which therefore may not reflect their actual values, or their reasons for having the preferences they do. Taking such criticisms on board, we here propose a new approach: the Personalized Patient Preference Predictor (P4). The P4 is based on recent advances in machine learning, which allow technologies including large language models to be more cheaply and efficiently ‘fine-tuned’ on person-specific data. The P4, unlike the PPP, would be able to infer an individual patient’s preferences from material (e.g., prior treatment decisions) that is in fact specific to them. Thus, we argue, in addition to being potentially more accurate at the individual level than the previously proposed PPP, the predictions of a P4 would also more directly reflect each patient’s own reasons and values. In this article, we review recent discoveries in artificial intelligence research that suggest a P4 is technically feasible, and argue that, if it is developed and appropriately deployed, it should assuage some of the main autonomy-based concerns of critics of the original PPP. We then consider various objections to our proposal and offer some tentative replies. (shrink)

There is growing evidence that dopamine replacement therapy (DRT) used to treat Parkinson’s Disease can cause compulsive behaviours and impulse control disorders (ICDs), such as pathological gambling, compulsive buying and hypersexuality. Like more familiar drug-based forms of addiction, these iatrogenic disorders can cause significant harm and distress for sufferers and their families. In some cases, people treated with DRT have lost their homes and businesses, or have been prosecuted for criminal sexual behaviours. In this article we first examine the evidence (...) that these disorders are caused by DRT. If it is accepted that DRT cause compulsive or addictive behaviours in a significant minority of individuals, then the following ethical and clinical questions arise: Under what circumstances is it ethical to prescribe a medication that may induce harmful compulsive behaviours? Are individuals treated with DRT morally responsible and hence culpable for harmful or criminal behaviour related to their medication? We conclude with some observations of the relevance of DRT-induced ICDs for our understanding of addiction and identify some promising directions for future research and ethical analysis. (shrink)

The proteasome family of proteases comprises oligomeric assemblies of very different symmetry. In different sizes, it features ring‐like oligomers with dihedral symmetry that allow the stacking of further rings of regulatory subunits as observed in the modular proteasome system, but also less symmetric helical assemblies. Comprehensive sequence and structural analyses of proteasome homologs reveal a parsimonious scenario of how symmetry may have emerged from a monomeric ancestral precursor and how it may have evolved throughout the proteasome family. The four characterized (...) representatives—ancestral β subunit (Anbu), HslV, betaproteobacterial proteasome homolog (BPH), and the 20S proteasome—are outlasting cornerstones in the family's evolutionary history, each marking a transition in symmetry. This article contextualizes the evolutionary and functional key aspects of these symmetry transitions, explaining how they facilitated the diversification and concurrent evolution of independent proteolytic systems side by side, each with its customized network of auxiliary interactors. -/- . (shrink)

Abductive reasoning is central to reconstructing the past in the geosciences. This paper outlines the nature of the abductive method and restates it in Bayesian terms. Evidence plays a key role in this working method and, in particular, traces of the past are important in this explanatory framework. Traces, whether singularly or as groups, are interpreted within the context of the event for which they have evidential claims. Traces are not considered as independent entities but rather as inter-related pieces of (...) information concerning the likelihood of specific events. Exemplification of the use of such traces is provided by dissecting an example of their use in the environmental reconstruction of mountain climate. (shrink)

Recent studies using functional magnetic resonance imaging of patients in a vegetative state have raised the possibility that such patients retain some degree of consciousness. In this paper, the ethical implications of such findings are outlined, in particular in relation to decisions about withdrawing life-sustaining treatment. It is sometimes assumed that if there is evidence of consciousness, treatment should not be withdrawn. But, paradoxically, the discovery of consciousness in very severely brain-damaged patients may provide more reason to let them die. (...) Although functional neuroimaging is likely to play an increasing role in the assessment of patients in a vegetative state, caution is needed in the interpretation of neuroimaging findings. (shrink)

Dreaming in sleep must depend on the activity of the brain as does cognition and memory in wakefulness. Yet our understanding of the physiological subtleties of state differences may still be too primitive to guide theories adequately in these areas. One can state nonetheless unequivocally that the brain in REM is poorly equipped to practice for eventualities of wakefulness through dreaming, or for consolidating into memory the complex experiences of that state. [Hobson et al., Nielsen, Solms, Vertes & Eastman, Revonsuo].

Obtaining informed consent from patients prior to a medical or surgical procedure is a fundamental part of safe and ethical clinical practice. Currently, it is routine for a significant part of the consent process to be delegated to members of the clinical team not performing the procedure (eg, junior doctors). However, it is common for consent-taking delegates to lack sufficient time and clinical knowledge to adequately promote patient autonomy and informed decision-making. Such problems might be addressed in a number of (...) ways. One possible solution to this clinical dilemma is through the use of conversational artificial intelligence using large language models (LLMs). There is considerable interest in the potential benefits of such models in medicine. For delegated procedural consent, LLM could improve patients’ access to the relevant procedural information and therefore enhance informed decision-making. In this paper, we first outline a hypothetical example of delegation of consent to LLMs prior to surgery. We then discuss existing clinical guidelines for consent delegation and some of the ways in which current practice may fail to meet the ethical purposes of informed consent. We outline and discuss the ethical implications of delegating consent to LLMs in medicine concluding that at least in certain clinical situations, the benefits of LLMs potentially far outweigh those of current practices. (shrink)

Here we describe how more important findings were obtained in a delirium study by using an informal assessment of mental capacity, and, in those who lacked capacity, obtaining consent later when or if capacity returned or a proxy was found. From a total of 233 patients 23 patients lacked capacity as judged by our informal capacity judgment and 210 did not. Of those who lacked capacity, 13 agreed to enter in the study. Six of them regained capacity later. When these (...) 13 participants were excluded from analysis, significant findings were no longer evident. These results show that by the inclusion of subjects who lacked capacity the results of analyses of the condition from whish they suffer are altered. We suggest that this approach to the study of delirium is more ethical than the usual system of strict exclusion of people who lack capacity to give consent and for whom assent is not available. (shrink)

Kidneys for transplantation are scarce, and many countries give priority to children in allocating them. This paper explains and criticizes the paediatric priority. We set out the relevant ethical principles of allocation, such as utility and severity, and the relevant facts to do with such matters as sensitization and child development. We argue that the facts and principles do not support and sometimes conflict with the priority given to children. We next consider various views on how age or the status (...) of children should affect allocation. Again, these views do not support priority to children in its current form. Since distinctions based on age ought to be positively justified, the failure of all these attempts at justification implies that the priority to children is ethically mistaken. Finally, the paper points to evidence that the paediatric priority reduces the overall supply of kidneys, at least in the United States. Paediatric priority is a real-world policy that seems discriminatory, in some places probably reduces the supply of organs, has no robust official defence, and is unsupported by mainstream ethical principles. Consequently, it should be ended. (shrink)

The age of minerals determined by radioactive methods agrees roughly with the age of the galaxy. This suggests that radioactive alpha-decay constants have probably changed by less than a factor of 3 or 4 during the last 3 or 4 × l09 years. It is shown that this implies that many ‘Constants of Nature’, particularly e, h and c probably change more slowly than 10−12 parts per year. The meson coupling constant also appears to change more slowly than this and (...) the beta-coupling constants by less than 10−9 parts per year. (shrink)