•  
  •  
 

Jurnal Ekonomi dan Studi Pembangunan

Abstract

This study explores the use of large language models (LLMs) as AI-assisted tools for evaluating sustainable blue economy policy scenarios in Indonesia. A snapshot prospective design was employed to compare five LLMs (ChatGPT, Gemini, Claude, DeepSeek, and Meta AI), across five blue economy scenarios: overfishing, coastal zone management, marine ecotourism, sustainable fisheries, and marine debris and circular economy implementation. The generated policy recommendations were assessed using a newly developed Blue Economy Policy Evaluation Framework (BEPEF), comprising eight dimensions: economic viability, environmental sustainability, social inclusiveness, governance quality, innovation and technology, alignment with blue economy principles, Sustainable Development Goals integration, and Indonesian policy context. The five LLMs produced different aggregate assessments, with mean scores ranging from 28.72 to 35.64 on a maximum scale of 40. The evaluation also showed high internal consistency and inter-model agreement under the specified assessment protocol. However, because the analysis involved only five LLM evaluators (N = 5), the findings are interpreted as exploratory rather than confirmatory. The observed score differences should therefore not be interpreted as evidence of objective model superiority or as direct measurements of real-world policy effectiveness. Furthermore, BEPEF has not yet undergone independent human-expert validation. The study positions LLM-assisted evaluation as a complementary policy-screening and scenario exploration instrument that may help policymakers and development economists identify multidimensional policy trade-offs and generate hypotheses for subsequent empirical and expert validation.

First Page

50

Last Page

64

Download

Share

COinS