Governing Human Capital Depreciation in Urban Automation via Capability-Constrained Dispatch
Year: 2026
Type: preprint
Abstract: Municipal services are transitioning from passive data analytics to agentic dispatch, employing software agents for cognitive tasks and autonomous robots for physical field operations. Current dispatch systems settle assignments through narrow operational metrics (speed, immediate accuracy, and unit cost) while deferring structural repercussions to post-deployment audits. This bifurcation masks a profound systemic hazard. Novice-level tasks are simultaneously the most economical to automate and the primary training ground where municipalities nurture future domain experts. An unbroken sequence of locally defensible dispatches can thus dismantle an urban professional talent pipeline without any single decision appearing flawed. We introduce CivicWorkOS, a governance-integrated allocation framework that embeds this long-term liability directly into real-time dispatch decisions. Central to our formulation is Civic Automation Debt, which prices five deferred institutional liabilities: developmental skill atrophy, emergency fallback degradation, accountability dilution, proprietary vendor lock-in, and labor transition friction. This penalty is integrated into a Sustainable Civic Value objective where it holds primary weight. Four operational constraints enforce sustainability: a qualified-practice capability budget, a demographic access rule guaranteeing equitable practice allocation across worker cohorts, a group displacement ceiling, and an N−1/N−2 contingency reserve ensuring critical service survivability. We formulate the municipal assignment program as a binary optimization problem, derive an online Lagrangian relaxation equipped with a feasibility-restoration mechanism for deficit states, and demonstrate via a bridge-inspection case study that the shadow price of practice hours flips the selected mode away from unconstrained automation toward capability-preserving collaboration.
Source: Preprints.org
Language: English
Subfield: Media Technology
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223178232