?
Mediating complexity: computational models as tools for developing systemic thinking about urban education
Background. Urban educational systems are complex, characterized by nonlinear interactions between multiple stakeholders – students, teachers, administrators, families, and policymakers. These systems involve interconnected dynamics: school choice based on incomplete information shapes school composition and resources; competition between schools produces both improvements and stratification; teacher satisfaction directly affects educational quality and school stability. Traditional analytical approaches rely on linear models or reductionist analysis. They fail to capture how individual decisions aggregate into system-level patterns, how competitive dynamics reshape educational opportunities, or how system properties feed back into learning outcomes. Digital transformation in education extends beyond technology implementation. It involves creating cognitive tools that make complex systems visible, enable systematic experimentation, externalize assumptions, reveal emergent patterns, and support reproducible practice.
Purpose. To investigate how agent-based modeling combined with semantic media wikis functions as a mediating tool for developing educators' systemic understanding of urban educational complexity, and to demonstrate that properly designed computational tools operationalize theoretical principles from Vygotsky, Davydov, Papert, and Turkle into educational practice that transforms how educators think about complex systems.
Materials and methods. This study employs design-based research implementing three interconnected agent-based models in NetLogo examining urban educational complexity: school choice dynamics with information asymmetries, competitive dynamics between urban schools, and systemic factors affecting teacher retention and satisfaction. Models are integrated into a semantic media wiki platform (digida.mgpu.ru) serving as collaborative documentation space. The research follows a seven-phase collaborative learning pipeline: model preparation and web conversion, platform integration using mw-embedNetlogo extension, initial guided exploration and discussion, systematic experimentation and data collection via BehaviorSpace, data organization on external repositories with integration through External Data extension, documentation of experimental processes and interpretation using visualization tools, and peer discussion and collaborative knowledge building in MediaWiki Talk spaces. Data include documented experimental work, discussion records, and evolution of understanding across wiki contributions. Analysis focuses on qualitative coding recognizing system properties (emergence, feedback loops, nonlinear causality, distributed causality) and tracking conceptual shifts from linear to systemic thinking, reductionist to holistic understanding, and policy naïveté to policy sophistication.
Results. Engagement with agent-based models of urban school systems leads to recognition of previously invisible system properties. Educators understand emergence - how segregation patterns, school rankings, and teacher turnover arise from individual decisions and institutional behaviors. They recognize feedback loops: how low teacher satisfaction leads to turnover which increases workload and further reduces satisfaction. They perceive nonlinear relationships: how small changes in information availability create dramatic shifts in school choice. They experience qualitative cognitive shifts: from linear causal thinking toward systemic thinking grasping circular causality; from reductionist analysis toward holistic understanding recognizing how multiple factors interact; toward policy sophistication recognizing cascading effects and unintended consequences. Model adaptation from research to educational versions demonstrates that reducing parameters, streamlining outputs, and optimizing procedures while maintaining validity creates epistemic tools supporting particular kinds of learning. Systematic data generation through BehaviorSpace reveals that single policy interventions produce limited effects while combined interventions addressing multiple dimensions produce substantial effects, demonstrating that urban education system outcomes emerge from how structural parameters interact. Students discover that outcomes involve threshold effects and nonlinearities invisible to intuition. Documentation within semantic wikis and peer discussion make learning processes visible and collective. The transformation from intuitive to systemic understanding constitutes the central result: students systematically engaging with computational models develop capacity for systemic analysis essential for educational leadership.
The scientific novelty of this study lies in its theoretical and empirical justification that computational models embedded in semantic wikis function as mediating tools for developing educators’ systemic thinking about the complexity of urban education systems, rather than merely serving as illustrative digital resources. The study translates general theoretical positions of educational psychology into concrete educational practices of collaborative work with agent-based models in a wiki environment. In doing so, it refines the understanding of digitalization of education as the design of tools that support a shift from empirical description toward systemic and theoretically grounded analysis. The practical significance of this study lies in its proposal of a reproducible model of learning design in which educators develop a systemic understanding of urban education through collaborative work with computational models and documentation of results in a wiki environment. This approach can be directly integrated into pre-service and in-service teacher education programmes, enabling educators to independently explore the consequences of policy and management decisions within a safe digital environment.