Crowd Development: The Interplay between Crowd Evaluation and Collaborative Dynamics in Wikipedia

Abstract

Collaborative crowdsourcing is an increasingly common way of accomplishing work in our economy. Yet, we know very little about how the behavior of these crowds changes over time and how these dynamics impact their performance. In this paper, we take a group development approach that considers how the behavior of crowds change over time in anticipation and as a result of their evaluation and recognition. Towards this goal, this paper studies the collaborative behavior of groups comprised of editors of articles that have been recognized for their outstanding quality and given the Good Articles (GA) status and those that eventually become Featured Articles (FA) on Wikipedia. The results show that the collaborative behavior of GA groups radically changes just prior to their nomination. In particular, the GA groups experience increases in the level of activity, centralization of workload, and level of GA experience and decreases in conflict (i.e., reverts) among editors. After being promoted to GA, they converge back to their typical behavior and composition. This indicates that crowd behavior prior to their evaluation period is dramatically different than behavior before or after. In addition, the collaborative behaviors of crowds during their promotion to GA are predictive of whether they are eventually promoted to FA. Our findings shed new light on the importance of time in understanding the relationship between crowd performance and collaborative measures such as centralization, conflict and experience.