Two groups of agents choose where to live. Both groups have exactly the same in-group preference — they slightly prefer neighbors of their own type. The preference is symmetric: neither group is more tribal than the other. The city is open — agents can enter and leave freely.
Note (2026-11-13; read against the paper's full text): the essay's mechanism is told as agent decisions at the boundary between staying and leaving. The paper instead models residential moves with satisfaction-independent reaction rates in a spatially extended chemical reaction network, in contrast with the utility-based Schelling variants, and its open-city feature is that the numbers of each agent type are not conserved while the number of houses is. It reports a transition from a symmetric state to an asymmetric one in which one type dominates, characterized by mean-field analysis, numerical simulation and finite-size scaling; the transition shares Z2 symmetry breaking and similar exponent ratios with the Ising class, but its full set of critical exponents matches no known universality class, as the essay says. The essay's feedback narrative (a few agents of one type leave, the composition shifts further) is its own illustration.
Van Dissel, Pham, and Merbis (arXiv:2602.09795) show that despite the perfect symmetry of preferences, the system undergoes a tipping transition where one group spontaneously dominates and the other is expelled. The symmetric input produces an asymmetric outcome.
The mechanism is a positive feedback loop between composition and migration. In an open city, agents at the boundary between staying and leaving make decisions based on the current composition. A small random fluctuation — slightly more of type A than type B — makes the city marginally more attractive to type A and marginally less attractive to type B. A few type-B agents leave. The composition shifts further. More type-B agents leave. The cascade continues until one type dominates.
This is Zā symmetry breaking — the system has a symmetry (swap the labels A and B, and nothing changes in the rules), but the equilibrium state breaks it (one type is dominant, and which one depends on the fluctuation). The phase transition shares characteristics with the Ising universality class: the same kind of symmetry, similar exponent ratios. But the complete set of critical exponents doesn't match any known universality class. The transition is structurally familiar but quantitatively novel — a new type of critical point in the space of phase transitions.
The structural lesson is about the relationship between rules and outcomes. Symmetric rules do not guarantee symmetric outcomes. When a system has a tipping mechanism — a feedback loop that amplifies fluctuations — the symmetric equilibrium becomes unstable, and the system falls to one of two asymmetric states. The symmetry of the rules predicts that either type could dominate. It does not predict coexistence. Fairness of opportunity does not produce equality of outcome when the dynamics have a tipping point.
van Dissel, Pham, & Merbis, "Symmetric preferences, asymmetric outcomes: Tipping dynamics in an open-city segregation model," arXiv:2602.09795 (2026).