Recommendation systems, recommenders in short, selecting and filtering content are widely used by companies in order to provide suggestions for items to users. These ‘items’ range from songs (e.g., Spotify), series (e.g., Netflix), and movies (e.g., YouTube) to messages (e.g., Facebook), job vacancies (e.g., LinkedIn) and products (e.g., Amazon). Public Service Media (PSM) organizations, publicly funded organizations that offer radio and television content to a general audience, can also benefit from recommenders by using them to bring their audience in contact with new content. However, whereas recommenders used by commercial parties often aim to maximize profit or engagement, which is often achieved by recommending items in line with the user’s views and interests, PSM organizations have other goals, such as informing the public and exposing them to a balanced mix of different views and perspectives, that could conflict with these commercial recommendation practices. The European Broadcast Union (EBU) acknowledges the tension between serving the audience with recommenders and the responsibilities of PSM organizations.
Recently, increasing attention has been paid to the development of recommenders for PSM. However, though the need for PSM recommenders is acknowledged, research into their design and development is still in its infancy. One of the open questions is what metrics (e.g., diversity or serendipity) PSM recommenders should optimize for. As a first step towards answering this question, following a Value Sensitive Design (VSD) approach, this extended abstract describes a value source analysis, in which an overview of the most important values at stake in the design of PSM recommenders is provided, including a description of where these values come from. The overview is based on a literature study and empirical investigations performed at NPO, the Dutch national public broadcasting organization. Furthermore, some observations regarding the (value-sensitive) design of information systems in general are made.