Abstract
Most formal models of governance treat governance as a single-level phenomenon and permit jurisdictions to advance their own flourishing at the expense of others. This paper extends the General Governance Success Model (GSM) to multi-level governance under a bounded non-exploitation constraint: no jurisdiction may impose uncompensated, non-trivial sacrifices on another for its own benefit. The extension introduces justice as a sixth governance capability, decomposed into distributive, procedural, recognition, corrective, and intergenerational dimensions. Justice is treated as a productive, fragile, multi-level capability that co-evolves with accountability, competence, cohesion, continuity, and learning—not merely as an external normative constraint. Four constraints govern the justice subsystem: bounded non-exploitation, a dimension-specific justice floor, vulnerability-sensitive non-exploitation, and vertical alignment enforced dimension-wise. A signed justice interaction matrix captures synergies and tensions between dimensions; the matrix is disciplined by bounded support, sign priors, and a hierarchical Bayesian identification protocol. Cross-level accountability is specified at the dimension level, so that recognition and intergenerational shortfalls cannot be masked by high distributive performance. Constraint enforcement is modelled as policy intervention with designated authority, implementation cost distributed across multiple capability channels, finite horizon, and passive recovery, yielding an endogenous frequency ceiling on aggressive justice improvement. A dimension-specific prioritarian welfare function completes the architecture. GSM-J is a conditional normative systems model: it makes value commitments explicit and swappable, but does not derive them. Its contribution is a formal substrate for a computational theory of normative governance systems. The model is released as an open-source Python library, gsm-governance.