What makes people accept recommendations? A comparison of different recommender types on music streaming platforms

This research aims to contribute to the development of a better understanding of how users of online services react to recommendations. To do so, it examines how the acceptance of a recommendation on a music streaming platform differs depending on the source of the recommendation, thereby drawing a comparison between algorithms, experts and peers as recommenders. Data for this research was gathered by conducting a between-subjects online survey experiment with three experimental conditions according to the three recommenders under study. The research used a vignette and a fabricated music recommendation to elicit the reactions of users to recommendations on streaming platforms. Based on the concept of algorithm appreciation by Logg, Minson and Moore (2019), it was hypothesised that algorithms will be perceived more positively as recommenders than peers, but less positive than experts. Changes in the outcome variables attitudes and intended behaviours were expected to be caused by the type of recommender, and – in consideration of the concept of source credibility – also by the degree to which participants perceived the recommender as trustworthy and knowledgeable. This relation in turn was expected to be influenced by eWOM scepticism. To analyse the considered effects, the research used moderated mediation. As the results show, both recommender type and source credibility had an impact on the acceptance of a recommendation and algorithm appreciation was confirmed as the algorithm recommender was perceived significantly better than the expert and the peer recommender. However, the effects under study occurred independently of each other, thus the level of source credibility attributed to a recommender does not serve as an explanation for why the recommender type affected the acceptance of a recommendation. eWOM scepticism was not confirmed as a moderator but did have a negative impact on source credibility. Perceived personalisation is discussed as a potential alternative explanation and it is suggested to future research to further explore the conceptual connections of perceived personalisation and algorithm appreciation. This research provides insights into how practices of music discovery are evolving on streaming platforms under the prevalent influence of technologies such as machine learning. In a broader sense, this thesis is a contribution to the larger theoretical framework of understanding the implications of artificial intelligence for society and culture.