When the full-scale russian invasion struck Ukraine in February 2022, volunteers showed up before the government did. In every documented case, spontaneous civilian mobilization preceded institutional coordination. Ukraine’s wartime experience has since been described as a tension between the Jungle, the chaotic, adaptive, decentralized innovation ecosystem that emerged from below, and the Zoo, the orderly institutional machinery of the state (Koshman, cited in Sweijs et al., 2026). This paper formalizes that tension computationally. Using Bevzenko’s crisis-agency framework, which distinguishes institutional and self-organizational pathways of social change, an agent-based model is built from 37 interviews and a 35-organization survey. The model formalizes 27 adaptation mechanisms as 38 computational rules within a BDI-inspired agent architecture organized around the Coleman Boat. Approximately 22,200 simulation runs yield three findings. First, volunteer mobilization is the only mechanism group without which the system fails. Second, decentralized configurations outperform centralized ones at every level of government capacity, revealing a paradox: strong institutions suppress the self-organization they intend to support. Third, the war leaves an asymmetric scar: the economy overshoots its pre-war level, but solidarity and cultural diversity are permanently consumed. The Jungle saves the Zoo, but the Zoo cannot return the favor.

The talk will present an outline of a research project of modelling mechanisms of securizitation and desecurizitation of civil society in Kosovo. In the project it is planned to simulate practices of local democracy and civil society in the case of Kosovo in which the war of the end 1990s is still a traumatic experience. This question implies an investigation of the mutual relation between trust in institutions and (development of) civil society. The modelling will be a participatory process and apply the methodology of interpretive agent-based modelling, which will result in narrative scenarios which then will be discussed in policy workshops.

How does a group of climate justice activists respond to escalating state-level repression aimed at preventing their activities? How does financial, emotional, and practical social support inside and outside the group help prevent individuals and the group from burning out? We address these questions with an agent-based model (ABM) grounded in extended fieldwork with a social movement organisation (SMO) that experienced escalating repression after blockading fossil fuel infrastructure. An LLM-aided analysis of 41 interviews and field notes highlighted social support inside and outside the SMO as central to preventing burnout. Building on this, the ABM captures the dynamic interplay of repression, social support, other causes of burnout, and resilience. We find that as individuals burn out and new individuals join, the distribution of external social support within the group changes. Over time, only individuals with high external social support can survive in the group. If repression decreases before both individuals and the group burn out, and internal social support drops to zero, the group can recover. Otherwise, the group undergoes a quick cycle of new members joining and burning out, in an increasingly taxing environment, before slowly dying out when repression decreases and there are less members joining.

Flooding is a complex socio-environmental challenge shaped by interactions between physical processes, human behaviour, and institutional responses. Agent-based modelling (ABM) offers a promising approach for capturing these dynamics. Participatory approaches and citizen science can improve behavioural realism, data availability, and stakeholder relevance, yet their integration into ABM remains limited. This contribution examines how participatory approaches are used in ABM for flood risk through a targeted scoping review of 46 studies. While participation, behaviour, and decision support are frequently referenced, these elements are sparsely integrated into modelling workflows. Key components such as data collection, model development, validation, and application are often developed in isolation. Building on these findings, we present an ongoing project developing a participatory ABM framework for flood risk management. The approach integrates stakeholder engagement, empirical data, and iterative model development within a connected modelling pipeline, linked to a citizen science platform and decision support system.