问责
公共行政
政治
水准点(测量)
生成语法
完备性(序理论)
政治学
公共政策
公共关系
经验证据
政策分析
语调(文学)
公务员
遗产管理(遗嘱认证法)
批评
社会学
法律与经济学
政策制定
实证研究
公共经济学
管理科学
最佳实践
测量数据收集
经济
标识
DOI:10.1093/scipol/scag075
摘要
Abstract This Letter responds to Nzobonimpa, Savard and Lawarée's article on a fine-tuned large language model for policy briefing notes. While the article offers an important empirical benchmark for task-specific GenAI in public administration, I argue that better drafting performance should not be equated with policy capacity. Its Alpha-Beta distinction suggests that GenAI is stronger in clarity, tone and structure than in evidence integration, contextual completeness and decision justification. I propose a layered evaluation framework that distinguishes artefact quality, analytical capacity, organizational capacity and political capacity, especially for developing-country administrations where data infrastructure, accountability and political contestation shape AI-assisted policy work.
科研通智能强力驱动
Strongly Powered by AbleSci AI