Investigating Crowd Creativity in Online Music Communities
This post summarizes our research paper on success factors in collaborative music communities, which will be presented at the ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSWC 2018) on November 6th.
Musicians who want to collaborate online with others to co-create music have quite a few options today. Collaborative songwriting communities like Songtree, Splice, and ccMixter provide features that allow their members to collaborate by reusing (extending) the music they share. Imagine that you are a talented singer: Why would you pick that specific instrumental song and extend it by singing over? Is it for the song? — This song rocks, it has so many likes and comments! Look, everybody is reusing their songs! — Or is it because of the talent of the musician who recorded it? — This musician is so gifted! Anything she touches becomes gold! Or, maybe, it’s both?!?
We have conducted a study aimed at understanding the factors influencing the success of collaboration in creative music communities. We built statistical regression models by analyzing about 45K songs created and reused by 7K authors in the three communities mentioned above. Then, we also interviewed 49 songwriters from Songtree, collecting feedback about their creative activity and how the platform supports it. Here is what we found.
Findings
Overall, our statistical models unveiled several features that can successfully predict song reuse across the three platforms.
We found that the most ‘popular’ ones, that is, those that receive more likes, bookmarks, and comments, have significantly more chances to be picked and reused by other musicians on the three platforms. We also found that when musicians decide to reuse a song, they select more often a new song rather than dig out an old release. We also observed that songs lose their ability to attract contributions over time, that is, the more a song has been reused, the fewer chances it has to be further extended.
As for authors, we observed that songwriters with higher ‘ranking,’ that is, those who are more active and embedded in a community, have higher odds to see their songs reused by fellow musicians. Also, this is true for those songwriters who customize their avatar image, replacing the default one with something that is easier to spot in the feed of activity.
According to Signaling theory, all these features are used by community members as assessment signals, that is, observable metadata from which musicians can more easily infer the hidden quality of the songs they want to reuse as well as the talent of their authors.
Implications for Collaborative Songwriting Platform Designers
Finally, we interviewed 49 Songtree users who raised interesting suggestions and insights that are potentially useful to designers of collaborative music platforms. Most notably, they highlighted: i) the need for using custom tags to define a folksonomy, thus providing a more flexible and efficient way to organize and seek content; ii) the need for implementing project management features to better support collaborative songwriting as a virtual band; iii) the flipside of gamification and contests, which foster competition instead of collaboration within the community, where the latter is expected to be the primary goal of such platforms.
For more details about the study and its findings, please refer to the full paper.
Fabio Calefato, Giuseppe Iaffaldano, Filippo Lanubile, and Federico Maiorano. 2018. Investigating Crowd Creativity in Online Music Communities. In Proceedings of the ACM on Human-Computer Interaction, Vol. 2, CSCW, Article 27 (November 2018). ACM, New York, NY. 22 pages. https://doi.org/10.1145/3274296
To find and read similar posts, follow @ACM_CSCW
