Workshop on Surprise, Opposition, and Obstruction in Adaptive and Personalized Systems (SOAP)

The phenomenon often referred to as the “filter-bubble,” i.e., the effect that collaborative, as well as content-based recommender systems keep making obvious, predictable, redundant, uninspiring, and therefore disengaging suggestions based on previous interactions, has emphasized the value of system qualities beyond pure accuracy, e.g., diversity, novelty, serendipity, or unexpectedness, to keep the user satisfied, e.g., [1, 2, 4]. Apart from the obvious use case in commercial systems (where this user satisfaction directly translates to revenue), these additional qualities become even more important in other areas. For instance, in creative domains, such as music production, we find that similaritybased, “more of the same” recommendations have basically no relevance, as illustrated by a quote from a professional music producer on the use of recommender systems that could predict his behavior in the process of music making: “I would be more interested in something that made me sound like the opposite of me [...] cause I can’t do that on my own” (anonymous, during interview on location at the Red Bull Music Academy 2014, cf. [3]). In fact, for creators and artists — as opposed to consumers — the idea of simulating a user based on the user’s past activity provides no added value, as the goal of creative work is something new that challenges and questions expectations and past behavior rather than reproducing it. Thus, opposite goals to those used in typical recommender systems matter when collaborating with an intelligent system for creative work: change of context instead of contextual preservation, defamiliarization [9] instead of predictability, or even explainability, opposition instead of imitation, and obstruction instead of automation. Still, these goals reside within the artist’s idea of personal style, putting a need for personalization of systems at the core of these concepts, cf. [7].