In experiments in isolated villages in Honduras, researchers evaluated a new strategy for identifying individuals that could be targeted for effective information spreading. Their approach – more effective than random targeting, and also less time-requisite than approaches that require a complete understanding of the relevant social network – could have far-reaching policy implications in lower and middle-income countries. Understanding the structure and function of human social networks has yielded insights for exploiting social contagion – the spread of behaviors, attitudes, and practices through the members of a group. Such an approach could be used to disseminate important information, including public health interventions. However, deliberately fostering social contagion in face-to-face social networks requires identifying the structurally influential individuals, or “seeds,” to maximize information spillover. Although previous research has suggested several ways to identify these individuals, all existing methods have generally required mapping the entire social network’s structure, which is often expensive, time-consuming, and infeasible in real-world face-to-face situations. Through various field experiments, Edoardo Airoldi and Nicholas Christakis evaluated whether it’s possible to identify the best “seeds” within a group without having to map the entire network. Airoldi and Christakis performed a large, randomized controlled trial of network targeting among 24,702 people in 176 isolated villages in Honduras. The authors randomly assigned villages to friendship targeting methods, varying the fractions of households receiving a 22-month health education package and the method by which these households were chosen. According to the authors, a friendship targeting strategy leveraging the so-called “friendship paradox” of human social networks, which states that, on average, the friends of randomly selected individuals are more central to the social network than those who identify them, was able to substantially reduce the number of households that needed to be targeted to attain a specified level of village-wide uptake. “Deploying interventions through network targeting, without increasing the number of people targeted or the expense incurred, may enhance the adoption and spread of the interventions and thereby improve human welfare,” write Airoldi and Christakis.
For news outlets interested in building their own data visualizations, data usable to draw network pictures for 11 of the involved villages is available here: https://yale.app.box.com/s/9q1r2grrugtlr5yqtsm9wdhjdx9wgflk. The authors will be happy to provide guidance regarding what subset of the data to use.
