It ain't easy: using normatively motivated news diversification to facilitate policy support, tolerance, and political participation

Abstract

Proponents of ‘democratic news recommender design’ argue that algorithmic news diversification may facilitate democratic participation. However, while various news diversification metrics have been proposed in recent years, few of them have been put to the test with real users. To assess the promises and pitfalls of algorithmic news diversification, we conduct a 2 (low vs. high levels of activating language) by 3 (low vs medium vs high levels of alternative voices) between subjects experiment with N = 715 respondents to test how normatively driven news diversification affects readers' (a) policy support, (b) outcome tolerance, (c) outgroup tolerance, and (d) political participation. Results show that in a one-off experiment, exposure diversity has at best very small effects on the dependent variables when demographic and attitudinal characteristics are controlled for. We also find that extreme forms of news diversification may impede user satisfaction.