How to channel knowledge coproduction behavior in an online community: Combining machine learning and narrative analysis

With ambiguous role definitions and without traditional organizational control or coordination mechanisms, how could online community channel knowledge coproduction behaviors among participants with heterogeneity roles? By using a novel theory-building method combining machine learning and narrative analysis, we inves-tigate the heterogeneity of participant roles and knowledge coproduction channeling behaviors in a Chinese self -organizing online community-Miui.com, launched by Xiaomi Inc. Using latent Dirichlet allocation topic modeling analysis and a complementary qualitative analysis of community participants' online forum posts, we initially identify three types of participant role in the community: leaders, supporters, and integrators, and find that (a) conceptualizing narrative practice is adopted to promote knowledge collaboration for leaders; (b) serializing and anthologizing narrative practices are used to promote knowledge collaboration for supporters; and (c) anthologizing narrative practice is adopted to promote knowledge collaboration for integrators. Our study advances the theoretical understanding of the knowledge coproduction and value creation in online communities.