We thank Dawid and Senn (2026) for clarifying why marginal interventional probabilities suffice when the objective is to maximize average recovery. We argue that the debate over counterfactuals in personalized decision-making persists because it conflates two levels of decision-making: individual bedside consultations and population policymaking. In a bedside consultation, a patient may be averse to treatment-caused harm relative to standard care, and the clinician’s role is to combine marginal study evidence with consultation-specific context to calibrate a subjective belief about joint potential outcomes. At the population level, a policymaker relies on marginal recovery rates conditional on measured study covariates, leaving counterfactual harm only partially identified. Using a three-arm extension of Dawid and Senn (2026), we characterize when marginal evidence suffices to uniquely determine an optimal treatment under harm aversion, when it yields a partial order that eliminates inferior options while leaving non-dominated treatments for clinical consultation, and how population harm budgets induce priority rules via lexical leapfrogging.