Burden-of-Proof Governance for Bullshit-Task Reduction in Digitally Governed Organizations

Digitally governed organizations can still accumulate persistent, high-burden workflows that do not materially improve outcomes participants care about—even under major technological progress. This paper formalizes that failure mode for settings where task-relevant actions are natively logged, governance is auditable, and decisions must be reproducible from committed observations. To avoid a single fixed welfare function, we introduce Mutable Subjective Objective Registries (MSORs): committed, versioned, anti-retcon dashboards computed from observables. The minimal MSOR schema emphasizes leisure, self-preference fulfillment, and social-structural cohesion, alongside protected experienced-meaning telemetry that is explicitly not used as an individual performance score. Task value is defined by causal contribution to the committed MSOR outcomes. Estimation follows a tiered identification stack designed for real organizations: (i) randomized ablation when safe, (ii) quasi-experiments when randomization is infeasible, and (iii) log-only observational learning strengthened by proximal causal bridges, with weak-proxy fallback to partial identification. Decisions maintain a Bayesian state with operational, essential, experiential, and value-certification posteriors (pbull, pₑss, pₑxp, pᵥal). Instead of hard auto-deletion, we replace “kill switches” with a burden-of-proof governance regime: hold → experiment → justify/substitute → reversible throttle. The governance layer includes challenge-bonded contestable ontology updates, anti-collusion risk-bounded substitution contracts, reserve and deficit-ledger solvency controls under systemic shocks, and constitutional sunset/renewal with auditable no-meta self-mutation of hyperparameters. On task-dependency graphs, the framework provides safe SCC intervention with fallback artifacts and dense-SCC edge controls that reduce cyclic compliance burden while preserving essential downstream continuity and innovation capacity. The framework is deployment-oriented: it maps to process-mining event logs (e. g. , XES/JXES), private-log / public-verifiability releases, and attested local execution (TEE/MPC with spot recomputation), with machine-readable claim dependencies for audit and crawler consumption. Intended audience: researchers and practitioners in digital governance, organizational/process analytics, causal inference for interventions, and robust mechanism design under strategic behavior.