When should a scientific community be cognitively diverse? This article presents a model for studying how the heterogeneity of learning heuristics used by scientist agents affects the epistemic efficiency of a scientific community. By extending the epistemic landscapes modeling approach introduced by Weisberg and Muldoon, the article casts light on the micro-mechanisms mediating cognitive diversity, coordination, and problem-solving efficiency. The results suggest that social learning and cognitive diversity produce epistemic benefits only when the epistemic community is faced with problems of sufficient difficulty.