The bachelor thesis was written by Laura Modre.
Most web-based news outlets employ recommendation algorithms which collect and process data on users’ previous behaviors and preferences to curate highly personalized news environments. Increasingly, there is a call for these systems to also incorporate insights from behavioral science to construct recommender systems that extend beyond accuracy and aim at improving individually and societally relevant issues. This paper introduces socially responsible news recommender systems that are designed to increase news consumption diversity through the utilization of digital nudges. To test the effectiveness of a feedback and a social norms nudge on users’ news consumption behavior in a simulated news recommender, a study was conducted as part of this paper. A sample of n = 117 participants completed an experimental online survey in which they responded to a set of questions regarding diversity in news and ran through three trials of news article selection based on their assignment to the control, feedback nudge, or social norms nudge group. A diversified article was defined as the target article and its selection rate measured and compared within the treatment groups via A/B test (control vs. feedback nudge; control vs. social norms nudge). The calculated conversion rates and conversion lifts (66.25%lift for the feedback nudge, 169.68% lift for the social norms nudge) were subsequently analyzed for significance employing a Mann-Whitney U-Test for independent samples. Results revealed that the feedback nudge did not lead to a significant change in article selection (U(114,144) = 7731, p = .078, r = - 0.058), which may, however, be explained by methodological limitations. In contrast, the study found that the social norms nudge did significantly impact article selection for the target article (U(114,71) = 4512, p = .002, r = - 0.149). This finding suggests that digital nudging may increase consumption of diverse news in news recommenders as developers can make design choices at the user interface that tap into cognitive heuristics and biases and may thus steer user decisions and behaviors towards specific, pre-defined options. The insights gained through the study provide ground for conducting further empirical research into nudgeenhanced recommender designs and highlight that digital nudges should be employed increasingly in news recommender systems to promote central democratic values such as diversity and tolerance in news consumption.
Supervisor: Irina Nalis-Neuner, Co-Supervisor: Julia Neidhardt