‍

The third reason why news can be more challenging to personalize than other industries is what we know as the filter bubble. Numerous experts have mentioned it in the media landscape, or you may know it because of the famous TED Talk of Eli Pariser in 2011. 

‍

You expect relevance from your entertainment provider, while you expect curation from your newspaper.

It seems pretty acceptable on a very recurrent basis to receive recommendations about sports documentaries if I effectively watch them. I would feel somewhat happy to discover these novelties without having to search by myself easily. 

However, having my newsfeed being entirely filled with football articles on a champions’ league evening would not really please me, even if I read about Cristiano Ronaldo testing positive for covid-19. With news, you actually do not expect consistency in recommendations, but rather diversity. Not a random list of articles from diverse sections, of course. You expect a handpicked selection of relevant articles broadly covering your different interests. 

‍

Win against the filter bubble‍

To achieve this curation mission, we use two main techniques. First of all, we apply a diversity re-ranked as the last step of the recommendation process. That means we make sure that the recommendation list that will be presented to the user contains sufficient diversity in terms of topics, and we rerank if necessary. Additionally, we use our very own “Shared Account” algorithm to detect the multiple personality facet one user can have. By ranking your reading history in subgroups and selecting them as different accounts, we make sure to propose relevant content but only from relevant topics. 

To show diverse articles is the one and only solution.

To sum up, we already count for three reasons, the model challenging events, the short shelf life, and in this episode, we discovered the filter bubbles. The filter bubbles are dangerous because they limit your point of view and your capacity to learn other perspectives and opinions.

Nowadays, the algorithms tend to show you random articles, so you don’t read the same thing repeatedly. That’s what we call diversity.