This study advances a model of how messages on groups one personally dislikes impact individual attitudes, using message persuasion and out-group admission theories to account for evidence type (numerical vs. narrative), easing conditions (fostering empathy vs. objectivity), and the underlying mechanisms (immersion). This model is tested in a pretest-posttest experiment, in which a sample of 601 Americans read counter-attitudinal observations below articles proposing either narrative or numerical evidence on illegal immigrants or same-sex couples. Narratives led to greater message admission and greater immersion, particularly in the empathetic condition. In turn, numerical messages caused self-perceived attitude shift in the objective condition. Immersion mediated narratives' persuasive impact in the empathetic, but not in the objective condition.