Choice architecture interventions are now being increasingly used in deploying behavioral interventions on a large scale, with the goal of helping citizens make better decisions. Many of these interventions are designed by mimicking other successful projects, or by conceptually extending ideas from published academic research. In this talk, Dr. Soman describes two such specific interventions in the area of financial wellbeing. The first intervention is an initiative by the government in South Korea on helping consumers make better spending decisions while using credit cards by sending text messaging alerts after every instance of credit card use. The second set of interventions in Mexico uses a redesigned pension account statement and text messaging alerts designed to motivate and remind citizens to make (recommended) voluntary contributions to their pension accounts. Both sets of interventions were designed on the basis of a conceptual analysis of evidence published in the behavioral sciences.
Results in both cases show that while the intervention worked for a subset of the target population, it did not work (and in some cases, actively backfired) for others. In South Korea, the intervention was only successful for 15% of the recipients. For the other 85%, a different psychological process (which arose from the particular design features of the intervention) played a stronger role and resulted in an opposite effect. The results point to the importance of pre-testing interventions, especially in domains where there is an expectation of strong context dependence. In Mexico, the strength of the effect varied as a function of age and gender. The researchers were able to use machine learning to detect the effects of heterogeneity. In ongoing work, Dr. Soman plans to customize the choice architecture interventions to account for heterogeneity.