We started out our project with developing a new method for studying how social influence works in adolescent social networks. For this we have set-up a mobile lab to do research at schools. This consists of 64 android tablets, 4 wifi routers, and dividers shield to ensure privacy (see Figure 1). With this mobile lab, we went to schools to collect data in two waves. First, we established the social network in a classroom based on sociometric data (e.g. friend ratings). At the same time, we measured behavior of interest on an experimental task. In the second session, we would return, and the pupils would do the same task but this time they would be confronted with earlier answers of their peers. Together with the detailed information about the social networks this allowed us to ask specific scientific questions regarding the impact of social information in relation to specific peer relations, peer status or position of pupils in the network. Finally, for our work at schools we also have developed a web-portal for pupils, parents and teachers, where they can find information about our work and sign up for studies (see Figure 1).
We have studied the role of social impact in several domains across different studies such as pro-social behavior (paper under revision), decisions under uncertainty (under revision), social norms (Pinho et al., 2021) and perception (Gradassi et al., 2021). With these studies we have also looked at different social relationships and have already gained a few key insights about social learning in adolescence. First, we have established that the impact of peers is generally declining with age across adolescence. Second, we have been able to confirm that high status individuals indeed have outsized social impact, even on those who are not directly friends. However, we have also found that those who are perceived as smart also have significantly more influence than others. The latter is hopeful because this already indicates that social information use in adolescence is to some extent a rational choice.
The perceptual estimation task has been central in the development of our conceptual (Molleman et al., 2019) and computational framework (Molleman et al., 2020) and nicely illustrates our approach. In this simple task people must quickly estimate the number of animals they see on the screen. Once they have done this, they are presented with the estimate of another person who also saw the same picture (see Figure 2). Now they can decide to change their initial estimate, the extent to which they change is our measure of social impact. They are rewarded for accuracy on both initial and second estimate, so should only change their mind when they think it really improves their results. This task was developed to be easily done by children as well as adolescents. The estimation task has now also been deployed in the two Science Museums: The Humboldt Forum in Berlin, and the NEMO science center in Amsterdam. In Berlin, the task is part of the permanent science exhibition (Humboldt Labor; currently data of 2000+ visitors, age ranges 10-90 from 25+ countries) and in Amsterdam we have used it to collect data for 8 weekends on social learning within families (135 families participated fig 3).
