Abstract

The accelerating closure of schools in depopulating regions is leaving a growing surplus of public assets whose reuse must be prioritised, yet systematic and transferable tools for supporting such decisions in advance remain scarce. This study proposes an artificial-intelligence-augmented multi-criteria framework that prioritises the adaptive reuse of closed schools using only openly available demographic and spatial data. Four criteria—regional ageing, building floor area, and proximity to administrative and transport infrastructure—were evaluated for 121 closed schools in Chungbuk Province, South Korea, under two weighting schemes, the subjective analytic hierarchy process and the objective entropy method, with a large-language-model agent added as an explanatory layer to interpret context and recommend reuse types. The data-driven weights proved liable to a structural distortion, elevating the single largest building to first place in urgency on the strength of its size alone, a misjudgement the agent corrected through contextual reasoning over the same data. Examining the schools from the opposed standpoints of intervention urgency and reuse potential further revealed a near-perfect inversion between them (Spearman ρ = −0.998), indicating that the most urgent schools are systematically those least able to sustain a market-led conversion. The framework addresses this dilemma not through a single optimal ranking but through spatially differentiated, agent-generated recommendations and is formulated for transfer to other middle-income economies approaching the same demographic transition.

IPC Classification

Keywords