James Edgar Lim

Democratic decision-making is difficult. Representatives often fail to represent the preferences of their constituents, and directly consulting members of the public can be costly. Inspired by these difficulties, several scholars have discussed the use of artificial intelligence (AI) models to support democratic decision-making. One such particular application is the use of AI to represent public policy preferences by predicting them. In this paper, we perform an analysis on the different ways AI models can be used to represent public policy preferences. We make distinctions between using AI as epistemic tools and for procedure; group and individual predictions; and predictions about preferences and inferences about values. We also describe how AI models can help policymakers screen policies for potential worries and objections, double-check any beliefs they have about the acceptability of their policies, and justify policy proposals. We also consider a number of worries about the use of AI in policymaking. We argue that these worries, while legitimate, can be mitigated or avoided in the way we have proposed the use of AI.

Why is transparency important for the use of AI in healthcare? Responses to this question typically claim that transparency is something owed to the patient – because it is a condition for informed consent, legitimacy, accountability to the patient, etc. In this paper, we draw attention to why transparency can be valuable for medical practitioners. We claim that transparent AI models facilitate critical engagement by medical practitioners with AI models that they use. That is, they enable practitioners to assess why AI models make the recommendations they do, think about how those reasons affect their own beliefs and judgments, and make reasoned decisions about whether to maintain or change their own judgments. Via this process, AI models can help medical practitioners to improve their practice in a distinctly valuable way. In turn, this benefits both medical practitioners and their patients. This conclusion has important implications for AI design in healthcare: if AI models are to be used in healthcare, they should be designed in ways which allow medical practitioners to understand how the models arrive at their recommendations, and engage with them critically.

Why is transparency important for the use of AI in healthcare? Responses to this question typically claim that transparency is something owed to the patient – because it is a condition for informed consent, legitimacy, accountability to the patient, etc. In this paper, we draw attention to why transparency can be valuable for medical practitioners. We claim that transparent AI models facilitate _critical engagement_ by medical practitioners with AI models that they use. That is, they enable practitioners to assess why AI models make the recommendations they do, think about how those reasons affect their own beliefs and judgments, and make reasoned decisions about whether to maintain or change their own judgments. Via this process, AI models can help medical practitioners to improve their practice in a distinctly valuable way. In turn, this benefits both medical practitioners and their patients. This conclusion has important implications for AI design in healthcare: if AI models are to be used in healthcare, they should be designed in ways which allow medical practitioners to understand how the models arrive at their recommendations, and engage with them critically.

What, if anything, can justify public shaming? Philosophers who have written on this topic have pointed out the role of public shaming in enforcing valuable social norms. In this paper, I defend an alternate, supplementary justification for public shaming: as a form of moral self-defence. Moral self-defence is the defence of one’s moral standing – being recognized as an equal in the eyes of oneself and others – rather than the defence of one’s physical body or rights. Agents can engage in moral self-defence by publicly criticizing, blaming, or expressing other negative feelings towards wrongdoers, which sometimes constitutes contributions to public shaming.

Social punishments– the informal penalties imposed by private individuals, rather than formal authorities– like the practice of online public shaming have attracted attention from philosophers, other academics, and journalists. Several have emphasized the harmful nature of social punishments, and the tendency of practices like public shaming to be disrespectful and disproportionate. So, what (if anything) justifies practices of social punishment like public shaming? Some authors have pointed out the important role social punishments play in enforcing morally valuable and authoritative norms, like norms against racist or sexist behaviour. In this paper, I argue that while these authors are correct in claiming that social punishments help to enforce valuable norms, their accounts fall short of explaining why private individuals are justified in imposing harm on others. Some obvious and intuitive explanations– like the claim that wrongdoers forfeit their rights against being harmed– fail to show that private individuals are permitted to harm others. As a result, more work needs to be done to develop our accounts of how and why social punishments can be justified.

Democratic decision-making is difficult. Representatives often fail to represent the preferences of their constituents, and directly consulting members of the public can be costly. Inspired by these difficulties, several scholars have discussed the use of artificial intelligence (AI) models to support democratic decision-making. One such particular application is the use of AI to represent public policy preferences by predicting them. In this paper, we perform an analysis on the different ways AI models can be used to represent public policy preferences. We make distinctions between using AI as epistemic tools and for procedure; group and individual predictions; and predictions about preferences and inferences about values. We also describe how AI models can help policymakers screen policies for potential worries and objections, double-check any beliefs they have about the acceptability of their policies, and justify policy proposals. We also consider a number of worries about the use of AI in policymaking. We argue that these worries, while legitimate, can be mitigated or avoided in the way we have proposed the use of AI.

Black-box AI cannot provide causal explanations for the decisions it makes, but medical AI has shown great promise as an accurate and reliable technology that both improves the quality of patient care and provides better access to healthcare for more patients. There is an ethical argument that to meet the informational requirements of patient autonomy, medical decision-making ought to be explainable to the patient. As such, there have been claims that black-box AI ought to be only minimally used in healthcare. This paper seeks to argue that black-box AI within the clinical context does not necessarily undermine patient autonomy, as defined in standard or relational accounts. Rather, patient autonomy is affected primarily by how AI tools are used.

Introduction Nonconsequentialist ethicists have noted that small harms, goods, or claims should not count against large claims. For example, given a choice between saving one life and a large group of people with minor headaches, we ought to save the one life, no matter how large the group is. This principle has been called limited aggregation. The principle of limited aggregation might have implications on public health policy, given that public health policy involves weighing the claims of individuals against one another. Aims and Methods I aim to show that limited aggregation has implications for policies on e-cigarettes and alternative nicotine delivery systems. The methodology used in this study involves critical analysis of existing literature and pre-theoretical moral views. Results This study does not use empirical research. Conclusions In deciding to allow or encourage the use of e-cigarettes or alternative nicotine delivery systems, we sometimes must weigh benefits to each existing smoker affected by the policy against risks to each nonsmoker affected. I argue in this paper that when these risks, to each individual nonsmoker, are sufficiently small, we ought not to count them against more significant benefits to smokers. This applies even when the number of nonsmokers affected by a policy exceeds the number of smokers. Implications This paper implies that policymakers ought to be sensitive to the scale of benefits or risks introduced by a policy on individuals. If the negative side effects, on each affected individual, of a proposed policy are sufficiently small, they do not count against the beneficial goals of that policy. Depending on the expected effects of each given e-cigarette policy, this may give policymakers defeasible reasons to prioritize the needs of current smokers, who may each gain a lot from various means of smoking cessation, over nonsmokers, who each may only have a small chance of picking up smoking.

Limited Aggregation is the view that when there are competing moral claims that demand our attention, we should sometimes satisfy the largest aggregate of claims, depending on the strength of the claims in question. In recent years, philosophers such as Patrick Tomlin and Alastair Norcross have argued that Limited Aggregation violates a number of rational choice principles such as Transitivity, Separability, and Contraction Consistency. Current versions of Limited Aggregation are what may be called Comparative Approaches because they involve assessing the relative strengths of various claims. In this paper, we offer a non-comparative version of Limited Aggregation, what we call the Threshold Approach. It states that there is a non-relative threshold that separates various claims. We demonstrate that the Threshold Approach does not violate rational choice principles such as Transitivity, Separability, and Contraction Consistency, and we show that potential concerns regarding such a view are surmountable.