The next speaker in this session at ECREA 2018 is Anne-Marie in der Au, who notes evidence that individual selection of media content may foster polarisation; however, there is also suspicion that algorithmic selection may foster such polarisation by building on and reinforcing such selective exposure. But empirical evidence on this is divided; several studies show no algorithmic impact or even demonstrate a negative correlation. What is going on here, and are there other variables that may interfere?
The present study examined these dynamics for the case of Germany, building on a representative phone survey. This measured the polarisation of opinions on the refugee issue in Germany, by assessing participants’ response to Angela Merkel’s statement Wir schaffen das! at the height of the refugee crisis. It correlated this with measures on the intensity of Internet and social media use, and the diversity of social media platforms used (some 44% of respondents were non-users of social media, notably), and with other sociodemographic factors.
Social media users showed greater levels of extreme agreement as well as strong levels of non-agreement; non-users were less likely to show extreme agreement, but similarly strong non-agreement, and the latter is an effect of their generally older age rather than of their non-use. Extreme disagreement is also more likely amongst participants without a high-school degree (who are also less active Internet and social media users, and may have a lower media literacy).
Overall, then, increased social media use seems to lead to more extreme opinions, but mainly for those users without a high-school degree; for more educated users, social media use actually appears to have a depolarising effect. However, it must be noted that these results are for a single topic, and on self-reported social media usage levels. Further work will extend this analysis by conducting direct tracking studies to observe actual social media usage patterns.