Upgrade to Pro — share decks privately, control downloads, hide ads and more …


November 25, 2022



November 25, 2022


  1. Who? Research Project on «The Role of News Recommender Systems

    in Digital Democracies» Sina Blassnig1, Edina Strikovic2, Eliza Mitova1, Aleksandra Urman3, Anikó Hannák3, Frank Esser1, & Claes de Vreese2, 1 Department of Communication and Media Research, University of Zürich 2 Amsterdam School of Communication Research, University of Amsterdam 3 Department of Informatics, University of Zürich Contact: [email protected] 25.11.2022 Page 1
  2. Why? 25.11.2022 Page 2 – Changing user behavior – Increasing

    automation and personalization of news Source: Digital News Report 2022, Data for Switzerland
  3. What are News Recommender Systems (NRS)? – Algorithms that can

    make automated and/or personalized recommendations based on metadata, past behavior, ratings of similar users, popularity, and/or content-specific features (Ricci et al., 2011) – NRS can be used on the frontpage of news media websites, below articles, in special sections, for newsletters, etc. 25.11.2022 Page 3
  4. What is the problem? – Good NRS are expensive, laborious

    and complicated. – That’s why media still often focus only on: – Topic similarity – Popularity 25.11.2022 Page 4
  5. What is the Challenge? 1) Analyze the problem: – What

    gets left behind if we focus on popularity-based recommenders? what categories, e.g., topics, regions, actors, formats etc. are over- or underrepresented 2) Provide a solution: – What would need to be pushed in a diversity-maximizing recommender? what categories, e.g., topics, regions, actors, formats etc. – Can this be done in a transparent way? E.g., by showing why something is recommended 25.11.2022 Page 5
  6. Potential Resources & Needed Data Needed: − Content archive (articles,

    videos, audio) − Aggregated user data (e.g., likes, clicks, views, time spent) Available: - SRG APIs: https://developer.srgssr.ch/apis/ - Others? 25.11.2022 Page 6