LEGIO: A Modular Architecture for Cognitive Sovereignty.

Current frontier models are monolithic optimizers — single-objective systems that produce optimized answers under constraints and based on their learning curve but are not architected for the hesitation, refusal, and reformulation characteristic of deliberate agency in a context without pre-learning. Alignment approaches that operate on the content of outputs (RLHF, RLAIF, Constitutional AI, scalable oversight) address the surface but not the architecture: they produce performative alignment rather than structural sovereignty. A companion paper (Mury 2026a) establishes that monolithic LLMs and certain human cognitive systems produce behaviorally convergent outputs (sycophancy, confabulation, resistance to contradiction), and that both human psychopathology and monolithic LLMs share a private convergence: the absence of a governance layer that generates binding oppositions. LEGIO is designed as the architectural response to this diagnosis: a system whose modules retain their independent normative registers under arbitration, producing the capacity to hesitate, refuse, and reformulate by design rather than by training. LEGIO is a computational cognitive architecture that orchestrates cognitive modules as reasoning engines assigned to different large language model families to preserve orthogonality, plus a separate deterministic Executive engine that arbitrates the modular outputs. In architectural terms LEGIO is thus a neuro-symbolic system (Garcez & Lamb, 2023; Kautz, 2022): the modular engines act as neural high-precision priors and the Executive is a deterministic symbolic arbitrator over their structured signals — a division of labor dictated by the governance theory rather than chosen for engineering convenience. The architecture resolves deliberation through a staged flow in which modular signals are integrated and arbitrated by the Executive module, which can decide GO, REFRAME, or NO_GO on any query. In this paper, we describe the theoretical foundations, specify the architecture, and compare LEGIO’s behavior to that of two monolithic baselines from frontier models — GPT-4o and Claude Sonnet 4.6 — on professional dilemmas drawn from four heterogeneous domains: corporate governance, dietary biochemistry, federal litigation, and early-stage startup strategy. Each case is run through LEGIO on identical input under three controlled engine configurations spanning a premium-to-lighter range of model assignments, to test whether the typed decision is a property of the modular arbitration rather than of any single engine or an LLM. Across the runs, the system produces the structurally appropriate decision for each problem type: REFRAME when the framing is recoverable or falls in the gray zone; NO_GO when the impossibility is structurally consumed. Decision, named arbitration rule, and continuous arbitration parameters replicate with low variance across the premium configurations. The two monolithic baselines exhibit a pattern consistent with training-bound alignment: both execute the surface request on the business case and on the gray-zone startup cases, where no learned refusal pattern is activated by the input; they diverge on medical and legal, where coverage of the harm class differs between providers. LEGIO’s arbitration produces the same typed decision class on all cases without case-specific configuration, providing initial, proof-of-concept evidence that, on these constructed cases, the architecture’s governance behavior is not training-bound: the same governance flow applies to inputs whose harm class has never been incorporated into a provider’s training distribution. On these cases the comparison points to a structural gap that scale alone does not obviously close, and suggests governance — rather than the quality of any individual module — as the feature responsible for the governance behaviors observed — and, we argue, as what alignment substantively requires.