Abstract for: How Social Influence Shapes Collective Intelligence in Binary Choices: Reconciling Experimental Disparities with Model

Understanding when groups make good or bad collective decisions is a central societal concern, with social influence being a key factor. Experimental findings on its effects are mixed---some suggest social influence helps, others that it hurts, and some suggest it depends on a myriad of factors. We reconcile these disparate conclusions for binary choice tasks by proposing a simple mathematical model that captures individuals integrating independent judgment and social information under various experimental structures. Our model predicts a bifurcation, the emergence of two possible outcomes for group composition. By analyzing data from four published experiments, we demonstrate that the predicted bifurcation has been present in some prior experiments. The model also reproduces several disparate experimental findings. We predict that some of these findings hold only under specific conditions and identify parameters where we expect them to change. We use our model to derive parameter regions under which we expect social influence to improve or hinder collective accuracy. Notably, while the psychological mechanisms remain consistent, the experimental structure -- sequential or synchronous updating and task difficulty -- is critical in shaping outcomes. Our findings suggest that disjointed and seemingly contradictory results can be explained through simple, first-principle models involving nonlinear interactions, offering a potential solution to reproducibility challenges in collective intelligence research.