Is doing harm morally worse than allowing it to occur? Our every-day intuitions, supported by a long-standing tradition in moral philosophy, suggest that this is the case. Nonetheless, the study of framing effects and cognitive biases has pointed out that our intuitions over the doing/allowing distinction are far from robust and reliable. This line of research casts doubts over the adequacy of our intuitions in grounding the moral principle “doing is worse than allowing” and seems to downplay the doing/allowing distinction as a cognitive bias or as a byproduct of our flawed reasoning skills. In this paper, I take evidence about framing and biases as a serious threat to the doing/allowing distinction. However, if we aim to explain common-sense morality, we need to account for its widespread use. To keep these two insights together, I build a causal model of the distinction, based on Christopher Hitchcock’s self-contained network account, which explains instances of attributions of these two labels. I conclude that the doing/allowing distinction can be better understood as a heuristic: in most cases, it helps us delivering quick moral judgements, but it can also misfire when cases are unfamiliar, underdescribed, or controversial.