Commentary: Advancing Applied Behavioral Science with Larger and Longer Field Partnerships

Abstract

T he applied behavioral science community currently faces two important and related calls to action. The first is to better explain how important findings work in different ways across different contexts and populations. The second is to harness these richer insights to more reliably scale applications to address important real world challenges. Larger and longer collaborations between researchers and field partners are one way to facilitate more sustained and coordinated testing of important insights across contexts, and to more efficiently address both. The last decade has seen a surge in interest for applied behavioral science, with a growing community of practitioners, policymakers, and researchers responding to exciting headlines claiming new cost-effective and choice-preserving ways to address important real world challenges (Thaler and Sunstein 2008; Soman and Leung 2020). However, this excitement was often followed by disappointment as insights failed to have the same effects when scaled across different contexts and populations (List 2022; Mažar and Soman 2022). There are now growing calls from researchers and practitioners to direct more efforts toward field research that can accelerate our understanding of how different interventions work across contexts and populations (Bryan, Tipton, and Yeager 2021; Goodyear, Hossain, and Soman 2022). To do so efficiently, research teams would benefit from more programs that systematically and iteratively test different ideas across contexts over longer time lines in coordinated ways. However, behavioral science field research practices are not typically set up with these critical elements. A more common approach sees researchers working independently on narrow or siloed programs (Milkman et al. 2021) with short-term field partners, often using opportunistic samples (Bryan et al. 2021) or biased sites (Allcott 2015) andwithout access to sufficient contextual detail (Szazi et al. 2017).