Recommender systems (RS) have become an integral component of the web and central to producing massive commercial value gains in e-commerce (e.g. Amazon), social media (e.g. Facebook), entertainment media (e.g. Netflix and Spotify), and online advertising (e.g. Google and Criteo). However, although RS have been introduced in news publishing similar gains have not yet materialized. The reason is that efficient recommendation in news publishing face additional technical and ethical challenges that must be solved.
Hence, our project, Responsible Recommender Systems for Danish News Publishing (RRS-DK), involves the development of responsible RS designed specifically for news based on artificial intelligence (AI), i.e. deep neural networks. Accordingly, to enable ethically responsible RS, RRS-DK also include development of methods for explainable AI (XAI), allowing diverse mix of RS to prevent adverse effects such as filter bubbles, echo chambers, and rabbit holes.
All technologies, systems, and methods that are developed will be tested against traditional academic benchmarks as well as through live tests on Ekstra Bladets digital platform which has approximately 1 million daily users and 15 million daily page views. By doing so, RRS-DK will advance academic research and real-world implementations of RS in fast moving settings in general and in Danish news publishing.
The goal is for granular matching of users with news content in a responsible manner while increasing the core value of a news offering to readers. In a longer-term perspective, RRS-DK contributes to making the Danish news publishing industry ready for a future where AI gives rise to new dominant news products and business models with AI as the core mode of operating. Thus, RRS-DK hope to solve several news domain specific technical and ethical problems that have held back the use of recommender systems in news publishing.