TV Tinder Exploring how the use of a video match-maker system can potentially improve co-viewing experiences of online public television

This research investigated to what extent our mock-up version of a group recommender system – that was created for the purpose of this study – positively contributed to the co-viewing experience of two people (a dyad), regarding public service media. Through the theory on Social Uses of TV by Lull (1980) we have shed light on the oftentimes forgotten social context in the development of group recommender systems, and their underlying algorithms to predict accurate viewing suggestions. By building on this theory with empirical research in the field, we have arrived at our hypotheses. A quantitative cross-sectional survey research was conducted to gather our data from a sample that matches the target audience of the Dutch public broadcaster NPO. We have taken some inspiration from experimental research in designing our methodology to ensure that the presentation of our mock-up group recommender system (GRS) was as closely linked to reality as possible (in our case duplicating the existing interface of the NPO online streaming platform). We foresaw that a match or mismatch in the viewing interests of our dyads, and the indication of their social relationship quality would influence the intention to follow up on the recommendations produced by our GRS. However, as appeared from our results, these associations were not significant. In this study, perceived usefulness of the endorsed videos by the GRS was measured looking at three constructs that recur in the debate on creating recommender system algorithms, being accuracy, novelty, and diversity of recommendations. A Hayes parallel multiple mediation model was conducted on the mediating power of perceived usefulness on the relationship between the perceived relevance of the GRS (captured by further use intentions) and the watching intentions of the suggestions put forward by this GRS. The results from our mediation analysis tied in closely with the prevailing accuracy-diversity debate as it complicates the creation of a diversityaware recommendation model serving the societal function of the public broadcaster. Only accuracy was found to fully mediating this relationship, however, we argue in our discussion that although the concept of diversity is harder to measure in terms of audience acceptance, it should not be underestimated for its valuable contribution to the richness of the media landscape in the Netherlands.