Genetic variation can have important consequences for populations: high population genetic diversity is typically associated with ecological success. Some mechanisms that account for these benefits assume that local social groups with high genetic diversity are more successful than low-diversity groups. At the same time, active decision-making by individuals can influence group genetic diversity, a behavioral process not generally incorporated into discussions of population-level diversity effects. Here, we examine how maternal decisions that determine group genetic diversity influence the viability of Drosophila melanogaster larvae. Our groups contained wildtype larvae, whose genetic diversity we manipulated; and genetically-marked “tester” larvae, whose genotype and frequency were identical in all trials. We measured wildtype and tester viability for each group. Surprisingly, the viability of wildtype larvae did not depend on group genetic diversity. However, the viability of the tester genotype was substantially depressed in large, high-diversity groups. Further, not all high-diversity groups produced this effect: certain combinations of wildtype genotypes were deleterious to tester viability, while other groups of the same diversity—but containing different wildtype genotypes—were not deleterious. These deleterious combinations of wildtype genotypes could not be predicted by observing the performance of the same tester and wildtype genotypes in low-diversity groups. Taken together, these results suggest that non-additive interactions among genotypes, rather than genetic diversity per se, account for between-group differences in viability in D. melanogaster; and that predicting the consequences of genetic diversity at the population level may not be straightforward.