This project is a collaboration with Finnish Broadcasting Company Yleisradio (YLE). The client is interested in how the use of algorithms is visible to the users of their online news services and how they can develop their digital news service algorithm visibility from the user perspective.
The client described their need for transparent use of algorithms and the need to explain the use of algorithms in an understandable manner to the users of their digital news services. The users of YLE news services must be informed of the use of algorithms in the services. There are multiple algorithms that are used to recommend the most relevant news for the specific user. The use of these algorithms will be researched more closely in this project to enable explaining their use to the users.
According to Lempinen (2019) the algorithm-based recommender system in YLE is meant to improve the ability to perform the basic task of presenting content in the services and applications to users. It is important to YLE that their recommender services are built on an ethically sustainable and transparent basis. The recommendations are based on cookies set by the browser, device identifiers set by mobile applications or Yle ID. YLE also uses journalistic recommendations in addition to algorithm-based recommendations.
The recommendation services consist of four main categories:
- content that may interest the user
- more content on the same subject
- user-controlled recommendations
- recommended content
Lempinen explains that all recommendation algorithms are not used in one service, but some services use more than one category of recommendations. YLE excludes certain topics from personalisation for ethical and technical reasons. Therefore, all content is not automatically personalised, or personalisation is not generated for all users. The topics that are excluded from news personalisation are content-specifically important news chosen via journalistic choice. YLE Arena streaming service’s news content is not personalised in Arena but only through the news service’s content personalisation. Children’s content in Arena is not personalised because data processing for minors is sensitive. (Lempinen, 2019.)
Lätti et al. (2020) define that YLE’s calls their algorithmic recommendations “journalistic referral system” and it combines importance, significance and interest. YLE is currently developing a new algorithm of journalistic referral. The goals and policies of the referral system are to increase the transparency of the recommendation. The special features of the news referral algorithm are based directly on the principles of news operations.
Four key things are considered:
- exceptional news situations
- mechanisms that ensure the main news are noticed by users
- newsroom’s ability to influence the recommendations if necessary
- the user’s ability to influence their own recommendations
Lätti et al. (2020) describe that YLE strives to improve the transparency of algorithmic decision-making and referral systems. Thereby they are developing new ways of communicating about referral to journalists and citizens. The development of YLE’s new referral algorithms has started from the mobile browser view because it is the most used system for the users. The focus is on recommending news articles which are the number one digital news broadcasting product in YLE. The recommendations will be extended into other content types in the future. This would include for example recommendations of Arena documents or podcasts.
The process is being evaluated by conducting user testing and qualitative research methods. The new journalistic referral system begins YLE’s common referral system’s development process. It is referred as an important factor on building a digital service experience in YLE. The journalistic recommendation algorithm development process will be evaluated throughout the process.
References
Lempinen, J. (2019) Miten ja missä Yle käyttää suosittelualgoritmeja. https://yle.fi/aihe/artikkeli/2019/11/27/miten-ja-missa-yle-kayttaa-suosittelupalveluita Last viewed 4.12.2021.
Lätti, R., Lempinen, J. & Kekäläinen, A. (2020) Miten rakennamme journalistisia suosittelujärjestelmiä ja miksi. Yle News Lab. https://newslab.yle.fi/blog/71ev58uiSMgpqz7uSVPa0J Last viewed 10.12.2021.