Daniel Weissglass

Tracking patient preferences is vital to medical decision-making, but evidence suggests that the standard method for tracking the preferences of incapacitated or incompetent patients (ie, surrogates) is inaccurate. Recent proposals suggest that artificial intelligence preference predictors (AIPPs) can improve preference tracking for these patients, but have faced significant objections. While many of these objections depend on unsettled empirical or technical assumptions, one prominent objection—that AIPPs rely inappropriately on impersonal information—seems to be an in-principle challenge to AIPPs. In this paper, we show that even granting an implausibly strong version of this objection, AIPPs may provide value to clinical decision-making. To show this, we develop suggestions that AIPPs may support best interest decision-making (BIDM) by improving the accuracy, consistency and speed of BIDM, and show that the prevalence of BIDM in the intensive care unit (ICU) grants this application of AIPPs significant moral and practical consequence. This not only clears a path to improve BIDM but also establishes a safe harbour—a relatively uncontroversial yet impactful space—in which proponents may develop AIPPs sufficently to resolve empirical and technical questions about their potential. We conclude by highlighting key questions for the application of AIPPs to BI determinations, setting an agenda for the deeper examination of a largely overlooked application of these tools.

Medical artificial intelligence (MAI) creates an opportunity to radically expand access to healthcare across the globe by allowing us to overcome the persistent labor shortages that limit healthcare access. This democratization of healthcare is the greatest moral promise of MAI. Whatever comes of the enthusiastic discourse about the ability of MAI to improve the state‐of‐the‐art in high‐income countries (HICs), it will be far less impactful than improving the desperate state‐of‐the‐actual in low‐ and middle‐income countries (LMICs). However, the almost exclusive development of MAI in HICs risks this promise being thwarted by contextual bias, an algorithmic bias that arises when the context of the training data is significantly dissimilar from potential contexts of application, which makes the unreflective application of HIC‐based MAI in LMIC contexts dangerous. The use of MAI in LMICs demands careful attention to context. In this paper, I aim to provide that attention. First, I illustrate the dire state of healthcare in LMICs and the hope that MAI may help us to improve this state. Next, I show that the radical differences between the health contexts of HICs and those of LMICs create an extraordinary risk of contextual bias. Then, I explore ethical challenges raised by this risk, and propose policies that will help to overcome those challenges. Finally, I sketch a wide range of related issues that need to be addressed to ensure that MAI has a positive impact on LMICs—and is able to improve, rather than worsen, global health equity.

The One Health revolution has recentered global health discourse so as to emphasize the impact of the non-human on human health, but has so far largely neglected the vegetal. In this chapter, I address the impact of vegetal turns on global health and the ethical issues which surround it. I review three ways that a vegetal turn might emerge in global health: emphasis on the instrumental value of the vegetal, recognition that interactions with the vegetal may inculcate virtues and vices in humans, and consideration that plants may themselves have intrinsic value. I review broadly, admittedly sacrificing depth, to suggest the scope and scale of the impacts that each of these would have to global health and related ethical considerations—presenting a sketch of the space of problems rather than offering solutions. Nonetheless, I hope that this chapter may serve as an effective call to action in illustrating a range of open and important questions at the intersection of the vegetal and global health.

As globalization and artificial intelligence technologies become increasingly intertwined, the use of generative AI (GenAI)—especially large language models (LLMs)—has made it possible to construct new forms of intercultural communication tools, but it has also raised significant ethical concerns. This paper aims to systematically explore the potential, challenges, and ethical framework surrounding “Generative Intercultural Communication Assistants” (GICAs). It begins by arguing for the technical feasibility and practical promise of GICAs, while also highlighting their potential risks. It then analyzes how LLMs represent cultural information, their inherent limitations, and the issue of “contextual bias,” identifying cultural misrepresentation as a serious ethical challenge. Next, it examines methods for adapting cultural representations—such as prompt engineering, retrieval-augmented generation (RAG), and model fine-tuning—and their limitations, while further discussing how data inequality and commercial incentives affect the fairness of cultural representation. On this basis, the paper proposes a set of ethical principles encompassing accuracy, non-stigmatization, participatory evaluation, goal supervision, and purpose-appropriateness to guide the responsible development and deployment of GICAs. Finally, it outlines directions for future research, emphasizing the need to address these dilemmas at both theoretical and practical levels, to grapple with power dynamics, and to foster interdisciplinary collaboration.

A key question about the value of truth in epistemology is whether the truthfulness of some proposition is a factor in our evaluation of beliefs. The traditional view—evidenced in introductory texts and academic journals :349–369, 2002, p. 350)—is that the truth of a belief should not impact our evaluations of it. Recent work has raised empirical objections to this default position of truth-insensitivity by suggesting that our ordinary belief evaluations assign considerable weight to the truth value of the believed proposition. If correct, this work grants truth a larger role in our epistemology—suggesting that it serves both as an element of our ordinary epistemic evaluations and as an intuitive norm for belief evaluation :4009–4030, 2015. https://doi.org/10.1007/s11229-015-0727-z, Philos Phenomenol Res 93:348–367, 2016. https://doi.org/10.1111/phpr.12218). However, this position is not convincing. In this paper, I argue that this empirical work, while revelatory of some interesting features of belief evaluation, falls far short of demonstrating that belief evaluation is truth-sensitive, or that truth-sensitivity is intuitive.

Causal theories of content, a popular family of approaches to defining the content of mental states, commonly run afoul of two related and serious problems that prevent them from providing an adequate theory of mental content—the misrepresentation problem and the disjunction problem. In this paper, I present a causal theory of content, built on information theoretic tools, that solves these problems and provides a viable model of mental content. This is the greatest surprise reduction theory of content, which identifies the content of a signal as the event the surprisal of which is most reduced by that signal. Conceptually, this amounts to the claim that the content of a signal is the event the probability of which has increased by the largest proportion, or the event that the signal makes the most less surprising to us. I develop the greatest surprise reduction theory of content in four stages. First, I introduce the general project of causal theories of content, and the challenges presented to this project by the misrepresentation and disjunction problems. Next, I review two recent and prominent causal theories of content and demonstrate the serious challenges faced by these approaches, both clarifying the need for a solution to the misrepresentation and disjunction problems and providing a conceptual background for the greatest surprise reduction theory. Then, I develop the greatest surprise reduction theory of content, demonstrate its ability to resolve the misrepresentation and disjunction problems, and explore some additional applications it may have. Finally, I conclude with a discussion of a particularly difficult challenge that remains to be addressed—the partition problem—and sketch a path to a potential solution.