The next speakers in this session at the 2026 International Communication Association conference in Cape Town are Pascal Schneiders and Andreas Riedl, whose interest is in diversity-oriented news recommender systems. Such ‘responsible’ recommender systems are being promoted as algorithmic solutions to ensuring that users receive a diverse diet of news content; they might pick up on popularity, content, and collaboratively created cues.
The aim here is to nudge audiences towards certain content, breaking through their ideologically shaped, one-sided news exposure and resulting in more diverse news consumption. Attitudes towards such systems depend on technological optimism, feelings of information overload, and media trust. But there is a risk of a Matthew effect, where audiences with already diverse news exposure benefit the most from such systems, while others perceive such interventions as overly paternalistic and patronising.
Potential users’ attitudes towards such systems were explored via a survey, which tested interest in diversity-oriented, popularity-based, collaborative, and content-based recommendation systems. This was correlated with participants’ existing ideological news exposure diversity, as measured by the variety, disparity, and balance of the news outlets participants reported using within a given week. Levels of media trust, news overload, and technology optimism were also tested through a series of questions.
Diversity- and popularity-based recommender systems received the highest level of support overall, content-based and collaborative systems were les popular. This is strongly correlated with an interest in diverse news recommendation, unsurprisingly, while users with a low existing diversity of news consumption are also least interested in such recommender systems. News overload and technology optimism are also significant drivers here, as is overall interest in the news, while media trust plays no major role.
Diversity-oriented news recommender systems hardly appeal to their key target audience, then; this means that such systems likely miss their mark. This might be able to be addressed by strengthening users’ agency and accountability in the design of such systems.
