Abstract: This article makes a recommendation for improving our ethical judgments. Working with the background assumption that many of these judgments are based on emotions, I argue that we predictably tend to underestimate the ethical significance of certain kinds of problem. Specifically, due to a lack of emotional connection, we will tend to underestimate the badness of problems that are impersonal, stochastic, large-scale and multiply realizable (e.g. climate breakdown, the threat of non-aligned AI). In response to this, I recommend we adopt a practice of *moral magnifying*. This involves (i) upping our estimations of the badness of problems with these features, and (ii) striving to increase the emotional vividness of such problems. More broadly, I argue that moral epistemologists should devote more effort to making practical recommendations of this kind, a project I call *ameliorative moral epistemology​*.