Despite early emphasis on decentralized organization and individual contributions, information commons such as online encyclopedias and open-source projects increasingly rely on users working in groups to collectively generate content as a shared resource. However, current insights into institutions as a driver of group performance are limited. First, the effect of institutions is often assessed in the aggregate rather than delineated by type. This ignores the fact that different rules govern different aspects of group behavior, leading to potential discrepancies in their relative effect on group outcomes. Second, groups vary not only in the individual rules they adopt but also in global attributes. Some groups exhibit greater institutional instability than others. Groups also differ in how diverse and how balanced the rules are.

Guided by the IAD approach, we develop and test several propositions linking rules to outcomes in an information commons. As the largest online encyclopedia in China, Baidu Baike operates as an open platform and relies on individual contributors to build and update encyclopedic entries. Based on a new dataset tracking changes in the rules and performance of 40 self-organized user groups over an eight-year period, our analysis shows that the effects on group performance are typologically imbalanced, in that significant variation in group performance is attributed to only some rule types. As for the global attributes, having a larger and more diverse set of rules generally corresponds with better group performance, while institutional instability does not seem to exert any discernible impact.

This dataset contains all protest-relevant posts on the LIHKG forum between June 10 and July 11, 2019. The dataset comprises a substantial corpus of 2,389,590 individual posts that are organized into 49,658 threads and were contributed by 12,624 distinct users. Note: all data could be publicly accessible in the LIHKG forum.

Data key fields: thread_id: Unique identifier for a thread. cat_id: Identifier for thread category. user_id: User ID who created the thread. item_data_reply_time: Date and time of the reply to the post within the thread data. item_data_user_id: ID of the user who posted within the thread data. post_text_token: Token of the thread data. push_count: Whether contain any of the following terms: “push”, “pish”, “posh”, “pash”, “psuh”, “up”, “tui”, “推”, or “幫推”. issues_pred: Strategic framing identified in the thread by the Bayesian algorithm. topic: Substantive topics identified in the thread by the LDA model.

How do individuals self-organize information provision in a community without centralized leadership? Protesters involved in the 2019 anti-extradition movement made the unusual decision to organize themselves in a “leaderless” structure, thus disabling many low-cost strategies that large groups use to apportion limited collective attention across competing priorities. Drawing on data from an online forum popular among movement participants, we empirically examine propositions that (1) users are sensitive to group-level consumption of information and seek to influence it and (2) their individual efforts contribute to global efficiency in information consumption.