Machine learning platforms have emerged as a new promissory technology that some argue will revolutionize work practices across a broad range of professions, including medical care. During the past few years, IBM has been testing its Watson for Oncology platform at several oncology departments around the world. Published reports, news stories, as well as our own empirical research show that in some cases, the levels of concordance over recommended treatment protocols between the platform and human oncologists have been quite low. (...) Other studies supported by IBM claim concordance rates as high as 96%. We use the Watson for Oncology case to examine the practice of using concordance levels between tumor boards and a machine learning decision-support system as a form of evidence. We address a challenge related to the epistemic authority between oncologists on tumor boards and the Watson Oncology platform by arguing that the use of concordance levels as a form of evidence of quality or trustworthiness is problematic. Although the platform provides links to the literature from which it draws its conclusion, it obfuscates the scoring criteria that it uses to value some studies over others. In other words, the platform “black boxes” the values that are coded into its scoring system. (shrink)

Many ethical concerns have been voiced about Clinical Decision Support Systems (CDSSs). Special attention has been paid to the effect of CDSSs on autonomy, responsibility, fairness and transparency. This journal has featured a discussion between Rosalind McDougall and Ezio Di Nucci that focused on the impact of IBM’s Watson for Oncology (Watson) on autonomy. The present article elaborates on this discussion in three ways. First, using Jonathan Pugh’s account of rational autonomy we show that how Watson presents its results might (...) impact decisional autonomy, while how Watson produces knowledge might affect practical autonomy. Second, by drawing an analogy with patient decision aids we identify an empirical way of estimating Watson’s impact on autonomy (ie, value-congruence). Lastly, McDougall introduced the notion of value-flexible design as a way to account for the diverging preferences patients hold. We will clarify its relation with the established domain of value-sensitive design. In terms of the tripartite methodology of value-sensitive design, we offer aconceptualclarification using Pugh’s account of rational autonomy, anempiricaltool to evaluate Watson’s impact on autonomy and situate a group oftechnicaloptions to incorporate autonomy in Watson’s design. (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)

The use of opaque, uninterpretable artificial intelligence systems in health care can be medically beneficial, but it is often viewed as potentially morally problematic on account of this opacity—because the systems are black boxes. Alex John London has recently argued that opacity is not generally problematic, given that many standard therapies are explanatorily opaque and that we can rely on statistical validation of the systems in deciding whether to implement them. But is statistical validation sufficient to justify implementation of these (...) AI systems in health care, or is it merely one of the necessary criteria? I argue that accountability, which holds an important role in preserving the patient‐physician trust that allows the institution of medicine to function, contributes further to an account of AI system justification. Hence, I endorse the vanishing accountability principle: accountability in medicine, in addition to statistical validation, must be preserved. AI systems that introduce problematic gaps in accountability should not be implemented. (shrink)