Data, Model, Interpretation, Decision: An Accountability Framework for AI-Assisted Educational Planning

Educational planning increasingly combines forecasting models with generative artificial intelligence. Forecasting models estimate future demand, supply, enrolment, staffing, or resource needs, while generative AI can translate those estimates into explanations, briefing notes, scenarios, and policy recommendations. Existing governance approaches usually assess a predictive model or an automated decision system as a single object. They give less attention to the additional interpretation stage created when a large language model converts a conditional forecast into policy language. This article develops a four-stage accountability framework for that combined decision chain. The framework assigns responsibilities for data, models, interpretation, and decisions; distinguishes model outputs from validated evidence and policy recommendations; establishes validation and authorization gates; and scales review requirements to the consequences of the proposed action. A published GM(1,1) forecast of teacher numbers and student-teacher ratios in Malaysia provides an illustrative application. The case illustrates how a numerical trend could be transformed into stronger causal and prescriptive claims that the forecasting model did not establish. The framework treats generative AI as a useful drafting and synthesis tool while retaining identifiable human and institutional ownership of verification, authorization, and correction. It offers education authorities a practical structure for using AI-assisted forecasts without obscuring who must explain the evidence and answer for the resulting policy.