Abstract

Intuitively, an agent who does harm behaves differently from an agent who allows harm to happen. This thesis examines the distinction between doing harm and merely allowing it to occur. I argue that this distinction is morally relevant, and doing harm is harder to justify than allowing harm, but that there is not always a fact of the matter how the distinction ought to be drawn. In Chapters 1 and 2, I survey the main alternative accounts for explaining the difference between “doing” and “allowing”. I compare causal approaches, which distinguish doing and allowing on the basis of how an agent caused an outcome, with “norm-based” accounts, which explain the distinction appealing to independent moral features. I conclude that a “mixed” causal account, such as Hitchcock's self-contained network model, is the most promising for tracking doing/allowing classifications. I then examine whether this distinction is morally relevant. I outline two theoretical hypotheses, the “positive” and the “negative” theses. The former argues that there is a fact of the matter whether an action is “doing” or “allowing”, and this classification is morally significant; the latter that there might or might not be such a fact of the matter, but in any case this distinction is not morally relevant. In Chapter 3, I critique an influential strategy for settling this issue, that is, comparing “fully-equalized cases”. In Chapter 4, I consider the import of “framing effects”. Despite attempts to use fully-equalized cases or evidence of framing in support of either thesis, these strategies are not compelling. In Chapters 5 and 6, I present my alternative thesis, which relies on the self-contained network model. I define “doings” as instances where an outcome counterfactually depends on the agent, within a “self-contained” network, and “allowings” as instances where the outcome depends on the agent, within a “non-self-contained” network. This classification captures whether an agent is causally relevant to an outcome in a specific way; nonetheless, the identification of “selfcontained” networks incorporates agents' empirical and normative expectations. The distinction is thus morally relevant, as it captures moral considerations, but may be ultimately ambiguous, as there may not always be a fact of the matter as to how the distinction should be drawn.