Using a randomised controlled trial to study behaviour in social networks

Around the world, social networks, digital ones in particular, appear to have increasing relevance in peoples’ everyday lives. They impact attitudes and behaviours in manifold ways, and they compete and often overlay news and information provided by public authorities and traditional media. This has stirred lively discussions of their role and relevance in a variety of social and political developments. Issues of concern stretch from democratic backlash, intolerance and political fanatism, to (un)healthy eating habits, vaccine uptake, and the spread of conspiracy theories. Research suggests, however, that the degree of influence between persons can vary depending on factors such as how people are connected and how similar they are to each other. To better understand the various dynamics occurring within social networks, researchers have experimented with controlled networks, manipulating factors like the number or kinds of ties. This increases the chances of identifying causal effects of network structures, which is important to social science since decade-old long questions of social phenomena can now be experimentally investigated. Yet, these strengths have to be weighed against effects on the external validity of the results since the design is very artificial. Especially for social science research, where variables and dynamics of interest tend to be particularly context dependent, this can constitute a problem. The paper presents a study design devised to reduce artificiality by studying real-life behavior and behavior change, while also including controlled treatments to compare variations across different contexts. Specifically, we present conclusions from a 4-month randomized controlled trial conducted with a random selection of adults living in Sweden from October 2022 to February 2023. Focusing on food intake as a behavior which is particularly common, frequent, and embedded in social contexts, we use a mobile phone application to investigate if and how real-life, self-reported intakes of plant- and animal-based foods diffuse under certain social network conditions. The trial includes two treatment groups and one control group exposed to (1) different social network conditions (with or without informational exposure), and (2) different source-types of information about food (by public authorities versus social). As we show, the design allows for studying diffusion of behaviors in social groups given certain ways in which people are connected and kinds of information reaching them. Hence, the design can also be applied in other contexts where the ambition is to understand the role and nature of social network dynamics for distinct social and political developments.