Cognitive differences can catalyze social learning through the process of one-to-one social influence. Yet the learning benefits of exposure to the ideas of cognitively dissimilar others often fail to materialize. Why do cognitive differences produce learning from interpersonal influence in some contexts but not in others? To answer this question, we distinguish between cognition that is expressed — one’s public stance on an issue and the way in which supporting arguments are framed — and cognition that is latent — the semantic associations that underpin these expressions. We theorize that, when latent cognition is obscured, one is more likely to be influenced to change one’s mind on an issue when exposed to the opposing ideas of cognitively dissimilar, rather than similar, others. When latent cognition is instead observable, a subtle similarity-attraction response tends to counteract the potency of cognitive differences — even when social identity cues and other categorical distinctions are inaccessible. To evaluate these ideas, we introduce a novel experimental paradigm in which participants: (a) respond to a polarizing scenario; (b) view an opposing argument by another whose latent cognition is either similar to or different from their own and is either observable or obscured; and (c) have an opportunity to respond again to the scenario. A pre-registered study (N=1,000) finds support for our theory. A supplemental study (N=200) suggests that the social influence of latent cognitive differences operates through the mechanism of argument novelty. We discuss implications of these findings for research on social influence, collective intelligence, and cognitive diversity in groups.