Generalizability of Choice Architecture Interventions

Although a given choice architecture intervention (‘nudge’) can be highly effective in some conditions, it may be ineffective in others and counter-productive in yet others. Critically, one cannot reliably predict which of these outcomes will happen. In this review, we argue that the average effectiveness of choice architecture interventions in influencing behavior is modest, and there is substantial heterogeneity in their impact. The complex interaction of multiple moderators and their dynamic change over time, makes it difficult to learn about when and to what extent choice architecture interventions work. We outline the obstacles to understanding generalizability, clarify the dimensions of generalizability and review the research practices (systematic exploration and measurement of moderators; sampling, designing, analyzing and reporting for generalizability) that could help the field more efficiently accumulate evidence. We conclude that adopting these practices is essential for advancing nuanced theories and for more accurately predicting the effectiveness of choice architecture interventions across diverse populations, settings, treatments, outputs and analytical approaches.