Published: March 13th, 2026 by Dr. Hengyi Fu

My recent publication, titled “Understanding Community Dynamics of Peer Production Through Contradictions: An Activity Theory Analysis of Mathematics Stack Exchange Q&A Community” appeared in the International Journal of Human–Computer Interaction.  This study investigates how internal contradictions shape the development and sustainability of online peer production communities, using the Mathematics Stack Exchange (Math SE) as an in-depth case. Drawing on Activity Theory, the paper conceptualizes Math SE as a socio-technical activity system in which subjects (users), objects (community goals), tools, rules, and divisions of labor interact dynamically over time. Through a 20-month mixed-methods field study, including participant observation, analysis of 1,942 governance discussion threads, and interviews with 28 contributors and moderators, the study identifies how tensions within this system do not merely hinder community functioning but actively drive adaptation, innovation, and institutional change .

The analysis reveals four categories of contradictions: between tools and objectives, between rules and objectives, between division of labor and objectives, and within the objectives themselves. Tool–object contradictions emerge from limitations in search, duplicate detection, question rediscovery, and moderation support, particularly in a domain where mathematical notation complicates algorithmic indexing. Rule–object contradictions arise from opaque question-closing processes, reputation-driven incentives, and inconsistent moderation practices, often leading to conflicts such as “close–reopen wars.” Labor-related contradictions stem from mismatches between reputation-based authority and effective community leadership, while object-level contradictions reflect competing visions among researchers, educators, and students regarding the community’s scope and purpose.

Rather than treating these tensions as failures, the study shows how Math SE has partially resolved them through three mechanisms: user-developed tools that compensate for platform limitations, flexible governance practices (notably tagging systems) that allow multiple sub-communities to coexist, and evolving moderation roles that balance quality control with newcomer support. These findings demonstrate Activity Theory’s analytical strength in explaining not only how communities break down, but how they reorganize themselves in response to structural strain.

The significance of this work is amplified in the context of the generative AI age. As AI systems increasingly mediate knowledge production, moderation, and search, there is a growing tendency to frame community challenges as technical problems solvable through automation. This study cautions against that reduction. Many of the contradictions identified—such as tensions between expert knowledge and accessibility, or between efficiency and inclusiveness—are fundamentally social and normative rather than computational. Generative AI may assist with duplicate detection, content summarization, or moderation triage, but it cannot resolve disagreements about community purpose, fairness, or legitimate participation without human governance and collective negotiation.

By foregrounding contradictions as engines of change, the study offers a timely reminder that sustainable AI-augmented communities must preserve space for human judgment, disagreement, and adaptation. Its design implications—such as graduated moderation systems, AI-assisted but human-interpretable governance tools, and role structures that recognize community maintenance work—are directly relevant to the design of future AI-integrated knowledge platforms.