Research

Digitalization has increased the importance of algorithms in society as well as in the production and consumption of news. However, current news recommendation systems are overly simplistic as they do not take into account content-related metrics such as the level of controversy or sentiment. Instead, they recommend news based on popularity metrics. The interplay between human agency and algorithms leads to less diversity in the views the audience is exposed to – commonly referred to as the ‘filter bubble’. This is problematic as diverse news consumption leads to well-informed citizens and to a well-balanced public debate.

Drawing on an multidisciplinary partnership between communication sciences, linguistics, law an computational sciences, this project aims to develop an algorithm that brings personalization and privacy in balance with each other, promoting news diversity. The research project will therefore draw on text- and communication scientific-based inquiries together with the development of technological applications and procedures.

These goals are reflected by five work packages. A first work package deals with the operationalization and algorithmization of news diversity metrics (WP1). The following two work packages work in close interaction with each other: the models developed for fine-grained automatic content analysis (WP2) will be incorporated into the development of a recommendation algorithm (WP3). A fourth work package, audience research, produces several in-depth user profiles in terms of diversity of news consumption, whereon the algorithm will be tested and validated. Last, but not least, we will assess how a diversity algorithm could be embedded in policy and regulation (WP5).

Back